WO2018184202A1

WO2018184202A1 - Inference engine training method and device, upgrade method, device and storage medium

Info

Publication number: WO2018184202A1
Application number: PCT/CN2017/079710
Authority: WO
Inventors: 徐亮禹; 马忠伟; 胡鹏; 刘志强
Original assignee: 北京悦琦创通科技有限公司
Priority date: 2017-04-07
Filing date: 2017-04-07
Publication date: 2018-10-11
Also published as: CN110235206B; CN110235206A

Abstract

Disclosed are an inference engine training method and device, an upgrade method and device for an auxiliary diagnostic system and a storage medium. The training method comprises: step S2010: using multiple groups of blood flow data stored in a knowledge base as samples for training an inference engine, wherein each group of the blood flow data stored in the knowledge base comprises blood flow feature parameters and corresponding known conclusions; step S2020: selecting at least one group of blood flow data from the knowledge base; step S2030: inferring by using the inference engine, with respect to each group of blood flow data in the at least one group of blood flow data, blood flow feature parameters in the group of blood flow data to obtain corresponding test conclusions; step S2040: performing hemodynamic checks on the basis of the blood flow feature parameters and the corresponding test conclusions of each group of blood flow data in the at least one group of blood flow data; and step S2050: determining whether the inference engine is up to a standard on the basis of a result of the hemodynamic checks. By using the hemodynamic checks to verify whether the trained inference engine is up to a standard, the training level of the current inference engine can be accurately known.

Description

Inference engine training method and device, upgrade method and device, and storage medium

Technical field

The present invention relates to the field of artificial intelligence, and more particularly to an inference engine training method and apparatus, an upgrade method and apparatus for an auxiliary diagnostic system, and a computer readable storage medium.

Background technique

Transcranial Doppler (TCD) blood flow analysis is a method for evaluating the physiological characteristics of different blood flow states through non-invasive examination. The Ultrasound Transcranial Doppler Flow Analyzer (referred to as Transcranial) is a customized ultrasound device designed for transcranial ultrasound examination. Transcranial is a product that appeared in the early 1980s to diagnose cerebrovascular disease and help to examine conditions such as narrowing of blood vessels, obstruction, poor blood flow, or cerebral hemorrhage. Doppler spectrum analysis technology can provide blood flow waveform, blood flow velocity (peak velocity, average velocity), blood flow disorder and frequency width under eddy current state, blood flow volume and other information for clinical diagnosis. Early detection is very important.

Ultrasound transcranial Doppler flowmetry uses an in vitro ultrasound probe to transmit ultrasound to the cerebral vessels through the gap or "window" of the skull. A Doppler effect (Doppler shift) is generated between the ultrasonic wave and the blood flow, and the reflected ultrasonic wave returns to the probe, and the data is processed by the processor in the analyzer to obtain corresponding information. Using the Doppler effect, the ultrasound transcranial Doppler flow analyzer can detect information such as blood flow velocity in blood vessels. When angiogenesis occurs, such as stenosis, obstruction, etc., its hemodynamics will change significantly.

In the existing transcranial Doppler equipment, after the spectrum is generated, the operator evaluates the spectrum to give a diagnosis. First, this increases the operator's workload, and the operator needs to manually identify the features one by one. Secondly, operator feature recognition is greatly affected by mental state, and leakage recognition may occur when fatigue or mood is low. Thirdly, the clinical diagnosis problem is very complicated, and the technical requirements of the operator are high, which requires more clinical training. Therefore, the artificial intelligence-based computer-aided diagnosis technology can be applied to the transcranial Doppler examination to assist the operator in diagnosis and improve the diagnosis efficiency. Auxiliary diagnostic systems generally include a knowledge base, an inference engine, and the like. There is currently a lack of training methods for inference engines for assisted diagnostic systems for transcranial Doppler examination.

Summary of the invention

The present invention has been made in consideration of the above problems. The invention provides an inference engine training method and device, an upgrade method and device for the auxiliary diagnosis system, and a computer readable storage medium.

According to an aspect of the present invention, an inference engine training method is provided. The method includes: Step S2010: using a plurality of sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes a blood flow characteristic parameter and a corresponding known conclusion; step S2020 Selecting at least one set of blood flow data from the knowledge base; step S2030: for each set of blood flow data in the at least one set of blood flow data, using an inference engine to infer blood flow characteristic parameters in the set of blood flow data, Obtaining corresponding test conclusions; step S2040: performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data; and step S2050: based on blood flow force school The test result judges whether the inference engine is qualified.

Illustratively, step S2050 includes: if it is determined that the inference engine fails, then go to step S2060; the inference engine training method further includes: step S2060: outputting indication information indicating that the inference engine has not passed the hemodynamic check.

Illustratively, after step S2060, the inference engine training method further includes: step S2070: receiving a model for modifying the model used to implement the inference engine, parameters of the inference engine, blood flow data stored in the knowledge base, and a hemodynamic checkout station. a first modification instruction of one or more of the employed hemodynamic models; and step S2080: performing a corresponding modification action according to the first modification instruction and returning to step S2010.

Illustratively, step S2050 includes: if it is determined that the inference engine is qualified, then go to step S2090; the inference engine training method further includes: step S2090: updating the known rules stored in the knowledge base with rules provided by the inference engine.

Exemplarily, step S2050 includes: if it is determined that the inference engine is qualified, go to step S2100; the inference engine training method further includes: step S2100: determining whether the rule provided by the inference engine conflicts with a known rule stored in the knowledge base, if Go to step S2110; and step S2110: output indication information indicating that the rule conflicts.

Illustratively, after step S2110, the inference engine training method further includes: step S2120: receiving one or more of modifying the model for implementing the inference engine, the parameters of the inference engine, and the blood flow data stored in the knowledge base. a second modification instruction; and step S2130: performing a corresponding modification action according to the second modification instruction and returning to step S2010.

Exemplarily, step S2100 includes: if the rule provided by the inference engine does not conflict with the known rule stored in the knowledge base, then go to step S2140; the inference engine training method further includes: step S2140: update the knowledge with the rule provided by the inference engine Known rules stored in the library.

Exemplarily, after step S2010, the inference engine training method further includes: step S2012: cross-validating the inference engine with the plurality of sets of blood flow data to evaluate the correct rate of the inference engine.

Illustratively, step S2012 includes: if the correct rate is less than the preset threshold, then go to step S2014; the inference engine training method further includes: step S2014: outputting indication information indicating that the inference engine has not passed the cross-validation.

Illustratively, after step S2014, the inference engine training method further comprises: step S2016: receiving one or more of modifying the model for implementing the inference engine, the parameters of the inference engine, and the blood flow data stored in the knowledge base. a third modification instruction; and step S2018: performing a corresponding modification action according to the third modification instruction and returning to step S2010.

Illustratively, the inference engine is implemented using a Bayesian classifier, a support vector machine, or a deep neural network.

Illustratively, step S2040 includes: establishing a predefined hemodynamic model; and determining whether blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data conform to predefined blood The physical laws specified by the flow dynamics model to obtain hemodynamic verification results.

According to another aspect of the present invention, there is provided a method for upgrading an auxiliary diagnostic system, comprising: inferring one or more sets of actual blood flow characteristic parameters using at least an inference engine trained by the above-described inference engine training method using an auxiliary diagnostic system; Obtaining one or more actual conclusions that correspond one-to-one with one or more sets of actual blood flow characteristic parameters; outputting one or more actual conclusions; receiving an indication of at least some of the actual conclusions of the one or more actual conclusions And storing, based on the indication information, at least a portion of the actual conclusions and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with at least a portion of the actual conclusions into a knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

Illustratively, the database of the auxiliary diagnostic system is used to store actual blood flow characteristic parameters previously processed by the auxiliary diagnostic system and previous conclusions obtained from inference of previously processed actual blood flow characteristic parameters, one or more sets of actual blood flow characteristic parameters Is the actual processed blood flow characteristic parameter of the database from the secondary diagnostic system.

Illustratively, at least part of the actual conclusion is the actual conclusion corresponding to the manually selected atypical case.

Illustratively, inference of one or more sets of actual blood flow characteristic parameters using at least an inference engine trained by the inference engine training method as described above using at least an auxiliary diagnostic system includes: rules provided by the integrated inference engine and knowledge of the auxiliary diagnostic system Known rules stored in the library to obtain a consolidated rule; and reasoning one or more sets of actual blood flow characteristic parameters based on the integrated rules.

Illustratively, the database of the auxiliary diagnostic system is used to store feedback information about whether the previous conclusion is adopted, and the upgrading method further includes: determining a specific type according to the feedback information stored in the database If the number of times the previous conclusion is not adopted exceeds the threshold of the number of times, the prompt information about the possible conclusion of the particular type of previous conclusion is output; the model for modifying the model used to implement the inference engine, the parameters of the inference engine, and the known rules stored in the knowledge base are received. Modifying one or more of the instructions; and performing a corresponding modification action according to the modified instruction.

Illustratively, the upgrading method further includes: for each of the plurality of blood vessels, using at least an inference engine to infer a set of actual blood flow characteristic parameters associated with the blood vessel to obtain a blood vessel related conclusion corresponding to the blood vessel; The blood vessel related conclusion corresponding to the blood vessel is output in real time; and feedback information on whether or not the blood vessel related conclusion corresponding to the blood vessel is accepted is received and stored in real time.

Illustratively, the upgrading method further includes: inferring, by using at least an inference engine, a plurality of sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence to obtain a plurality of blood vessel correlation conclusions corresponding to the plurality of blood vessels one by one; Multiple vessel related conclusions determine the overall conclusion; output the overall conclusion; and receive and store feedback information as to whether the overall conclusion was adopted.

Illustratively, one or more sets of actual blood flow characteristic parameters are blood flow characteristic parameters that the auxiliary diagnostic system currently needs to process, and the upgrading method further includes: generating a new interpreter in the auxiliary diagnostic system based on the knowledge base and/or the inference engine .

According to another aspect of the present invention, an inference engine training apparatus is provided, comprising: a training module, configured to use a plurality of sets of blood flow data stored in a knowledge base as a sample training inference engine, wherein each set of blood flow stored in the knowledge base The data includes blood flow characteristic parameters and corresponding known conclusions; a selection module for selecting at least one set of blood flow data from the knowledge base; and an inference module for each set of blood flow data in the at least one set of blood flow data, Using the inference engine to process the blood flow characteristic parameters in the blood flow data to obtain corresponding test conclusions; the hemodynamic verification module is configured to be based on each set of blood flow data in the at least one set of blood flow data The blood flow characteristic parameter and the corresponding test conclusion are subjected to hemodynamic verification; and the qualification judgment module is configured to judge whether the inference engine is qualified based on the hemodynamic verification result.

According to another aspect of the present invention, there is provided an apparatus for upgrading an auxiliary diagnostic system, comprising: a first inference module, configured to use at least one of the inference engine trained by the inference engine training method of the auxiliary diagnostic system; The actual blood flow characteristic parameters are inferred to obtain one or more actual conclusions corresponding to one or more sets of actual blood flow characteristic parameters; a first output module for outputting one or more actual conclusions; a receiving module, configured to receive indication information about at least part of the actual conclusions of the one or more actual conclusions; and a first storage module, configured to, based on the indication information, at least part of the actual conclusion and at least one of the actual conclusions A set of actual blood flow characteristic parameters are stored in the knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

According to another aspect of the present invention, there is provided an inference engine training apparatus comprising: a memory for storing a program; a processor for running a program; wherein, when the program is run in the processor, the method is the following: Step S2010 : using the plurality of sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions; Step S2020: selecting from a knowledge base At least one set of blood flow data; step S2030: for each set of blood flow data in the at least one set of blood flow data, using an inference engine to infer blood flow characteristic parameters in the blood flow data to obtain a corresponding test conclusion; Step S2040: performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data; and step S2050: determining whether the inference engine is based on the hemodynamic check result qualified.

According to another aspect of the present invention, an apparatus for upgrading an auxiliary diagnostic system includes: a memory for storing a program; a processor for running a program; wherein, when the program is run in the processor, the following steps are performed: At least the inference engine obtained by the above-mentioned inference engine training method using the auxiliary diagnosis system infers one or more sets of actual blood flow characteristic parameters to obtain one-to-one correspondence with one or more sets of actual blood flow characteristic parameters. One or more actual conclusions; outputting one or more actual conclusions; receiving indication information regarding at least a portion of the actual conclusions in the one or more actual conclusions; and based on the indication information at least a portion of the actual conclusions and at least some of the actual conclusions Corresponding at least one set of actual blood flow characteristic parameters are stored in a knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

According to another aspect of the present invention, a computer readable storage medium is provided. A program is stored on a storage medium, and the program is executed at runtime to perform the following steps: Step S2010: using a plurality of sets of blood flow data stored in the knowledge base as a sample training An inference engine, wherein each set of blood flow data stored in the knowledge base includes a blood flow characteristic parameter and a corresponding known conclusion; step S2020: selecting at least one set of blood flow data from the knowledge base; and step S2030: for at least one group of blood Each set of blood flow data in the flow data is inferred by the inference engine to determine blood flow characteristic parameters in the blood flow data to obtain a corresponding test conclusion; step S2040: based on each set of blood in at least one set of blood flow data The blood flow characteristic parameter in the flow data and the corresponding test conclusion are subjected to hemodynamic check; and step S2050: determining whether the inference engine is qualified based on the hemodynamic check result.

According to another aspect of the present invention, a computer readable storage medium is provided, on which a program is stored, the program being used at runtime to perform the following steps: at least using the auxiliary diagnosis system and the reasoning obtained by the above-described inference engine training method training Performing inference on one or more sets of actual blood flow characteristic parameters to obtain one or more actual conclusions that correspond one-to-one with one or more sets of actual blood flow characteristic parameters; outputting one or more actual conclusions; receiving At least some of the actual conclusions of one or more actual conclusions Instructing information; and storing, based on the indication information, at least a portion of the actual conclusions and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with at least a portion of the actual conclusions into a knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

The inference engine training method and device, the upgrade method and device for the auxiliary diagnosis system, and the storage medium according to the embodiment of the present invention use the hemodynamic check to verify whether the inference engine after the training is qualified, thereby accurately knowing the current inference engine Training level. If it is determined that the inference engine is unqualified, it may choose to continue training until the inference engine is qualified. This facilitates training to obtain a more efficient inference engine, thereby obtaining a more accurate and efficient auxiliary diagnostic system for transcranial Doppler examination.

DRAWINGS

The above as well as other objects, features and advantages of the present invention will become more apparent from the embodiments of the invention. The drawings are intended to provide a further understanding of the embodiments of the invention, In the figures, the same reference numerals generally refer to the same parts or steps.

1 shows a schematic block diagram of an auxiliary diagnostic system and related elements for transcranial Doppler examination, in accordance with one embodiment of the present invention;

2 is a schematic flow chart showing an inference engine training method according to an embodiment of the present invention;

Figure 3 shows a schematic view of the cerebral arterial ring;

4 is a schematic flow chart showing an upgrade method of an auxiliary diagnosis system according to an embodiment of the present invention;

Figure 5 shows a schematic block diagram of an inference engine training device in accordance with one embodiment of the present invention;

6 shows a schematic block diagram of an upgrade apparatus of an auxiliary diagnostic system in accordance with one embodiment of the present invention;

Figure 7 shows a schematic block diagram of an inference engine training device in accordance with one embodiment of the present invention;

Figure 8 shows a schematic block diagram of an upgrade apparatus of an auxiliary diagnostic system in accordance with one embodiment of the present invention.

detailed description

In order to make the objects, the technical solutions and the advantages of the present invention more apparent, the exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. It is apparent that the described embodiments are only a part of the embodiments of the present invention, and are not all embodiments of the present invention. It should be understood that the present invention is not limited by the examples described herein. The limitations of the example. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention described herein without departing from the scope of the invention are intended to fall within the scope of the invention.

Computer-aided diagnostic techniques have not yet been applied to the diagnosis of transcranial Doppler. Because medical diagnosis has a very high degree of professionalism, the general auxiliary diagnostic method does not solve the practical problems encountered in the transcranial Doppler field. According to an embodiment of the present invention, the results obtained by the auxiliary diagnostic technique can be combined with the specific characteristics of the transcranial Doppler to effectively solve the clinical problem of transcranial Doppler. The auxiliary diagnostic system for transcranial Doppler examination according to an embodiment of the present invention may include a knowledge base, a database, an inference engine, an interpreter, and a user interface.

1 shows a schematic block diagram of an auxiliary diagnostic system and related elements for transcranial Doppler examination, in accordance with one embodiment of the present invention. As shown in Figure 1, the auxiliary diagnostic system includes a knowledge base, a database, an inference engine, an interpreter, and a user interface. Experts and literature are elements related to assisted diagnostic systems. The literature mainly refers to the research conclusions that can be consulted. After being transformed by experts, it can be directly put into the knowledge base. Experts mainly refer to human experts, who can be those who have many years of work experience in the transcranial Doppler field, have a large amount of professional knowledge, and have the ability to make correct diagnosis conclusions for clinical manifestations. The foundation of the knowledge base is laid by experts, and new knowledge imports can also be reviewed by experts. The database is used to store the raw data, intermediate results and final conclusions required in the reasoning process of the inference engine, and is often used as a temporary storage area. The interpreter can convert the input transcranial Doppler test results into auxiliary diagnostic conclusions based on knowledge in the knowledge base, such as known rules stored in the knowledge base. The user interface can provide an interface for the operator (which can be a normal user or an expert) to interact with the diagnostic system. For example, a transcranial Doppler examination typically involves examining about 10 blood vessels, and during the examination and at the end of the total examination, the operator (eg, an expert) and the secondary diagnostic system can interact via the user interface. The role of the knowledge base and inference engine is described below.

As described above, there is currently a lack of training methods for inference engines for assisted diagnostic systems for transcranial Doppler examination. In order to solve the above problems, an embodiment of the present invention provides an inference engine training method and apparatus. According to the inference engine training method of the embodiment of the present invention, in the training process of the inference engine, the hemodynamic check is used to verify whether the inference engine after the training is qualified. If it fails, you can choose to retrain until the inference engine is qualified.

Next, an inference engine training method 2000 according to an embodiment of the present invention will be described with reference to FIG. 2 shows a schematic flow diagram of an inference engine training method 2000 in accordance with one embodiment of the present invention. As shown in FIG. 2, the inference engine training method 2000 includes the following steps.

In step S2010, using the plurality of sets of blood flow data stored in the knowledge base as a sample training inference engine, Wherein, each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions.

A knowledge base can store data calibrated by an expert and rules refined by an inference engine. Data that is usually calibrated by experts is a very scarce resource. Some of the knowledge in the knowledge base is output by the inference engine and can be used as an important supplement to expert knowledge. The knowledge base is the key to the superior quality of the auxiliary diagnostic system, that is, the quality and quantity of knowledge in the knowledge base determines the quality level of the auxiliary diagnostic system. Knowledge in the knowledge base can include a priori facts, a priori conclusions, rules, and so on.

The data calibrated by the expert above may include several sets of blood flow data, each set of blood flow data including blood flow characteristic parameters (ie, a priori facts). In addition, each set of blood flow data can also include known conclusions (i.e., a priori conclusions) corresponding to blood flow characteristic parameters in the set of blood flow data. A set of blood flow data can be considered as one case. Blood flow parameters may include systolic maximum blood flow velocity, end-diastolic blood flow velocity, mean blood flow velocity, pulsation index, resistance index, blood flow reversal, blood stealing, eddy current, turbulence, and dash. These blood flow characteristic parameters can be extracted from the transcranial Doppler spectrum, and the transcranial Doppler spectrum can be obtained by an existing or future ultrasound transcranial Doppler flow analyzer. The main blood vessels in the brain include: middle cerebral artery, anterior cerebral artery, posterior cerebral artery, vertebral artery and basilar artery. Since the blood vessels of the neck and the blood vessels of the brain are directly connected, the relevant common carotid artery, internal carotid artery and external carotid artery are also blood vessels that can be examined by an ultrasound transcranial Doppler blood flow analyzer. For each blood vessel, a transcranial Doppler spectrum can be obtained and the blood flow characteristic parameters can be identified from the transcranial Doppler spectrum.

Assuming that each set of blood flow data includes a total of ten blood flow characteristic parameters, these blood flow characteristic parameters can be represented by a vector indicating V=(x0, x1...x9). Among them, the speed can be represented by actual values, and the state quantity can be represented by 0 and 1, for example, there is a eddy current of 1, and no existence is 0. For multiple blood vessels, multiple vectors can be obtained, such as VMCA, VACA, VBA, and the like. The combination of blood flow characteristic parameters of all blood vessels is T = (VMCA, VACA, VBA...). T can represent all blood flow characteristic parameters in a set of blood flow data. For a subject, all blood flow characteristic parameters in the initial state can be set to 0, and each sub-vector can be filled after each blood vessel is examined.

The inference engine is the core of the auxiliary diagnostic system and can perfect the detailed rules that human experts cannot generalize. The knowledge summarized by human experts is usually rough and limited by human resources, and it is not very comprehensive. Usually hundreds of rules are already very large knowledge resources. However, this is far from sufficient for assisted diagnosis. Auxiliary diagnosis requires more complete and rich diagnostic criteria to cope with various complex diseases in practical applications. The thresholds involved in the rules given by human experts (ie known rules stored in the knowledge base) are often not precise enough, which also requires an inference engine to perfect. When the rules provided by the inference engine are inconsistent with human experts, the conclusions of human experts are the main ones. Although the work of human experts The use is decisive, but the result of the inference engine is the subject of the knowledge base.

The inference engine can be trained by the knowledge base. Alternatively, the machine learning algorithm is an effective implementation of the inference engine. For example, the inference engine can be implemented using a Bayesian classifier, a support vector machine, or a deep neural network.

At step S2020, at least one set of blood flow data is selected from the knowledge base.

At least one set of blood flow data can be selected from the knowledge base as a test set to test whether the trained inference engine meets the requirements. Illustratively, at least one set of blood flow data can be randomly selected.

In step S2030, for each set of blood flow data in the at least one set of blood flow data, the inference engine is used to infer the blood flow characteristic parameters in the blood flow data to obtain a corresponding test conclusion.

The inference engine can reason the blood flow characteristic parameters based on the trained rules. The blood flow characteristic parameters in a set of blood flow data are input into the trained inference engine, and the corresponding test conclusions (ie, the cause corresponding to the blood flow data) can be obtained. Illustratively, after the inference engine is trained, the following rule is obtained: an average blood flow velocity greater than 65 cm/s represents an abnormality in the blood vessel (eg, a stenosis lesion). Assuming that the average blood flow velocity of a blood vessel in a set of blood flow data input to the inference engine is 80 cm/s, the test conclusion may be that there is a vascular stenosis lesion.

In step S2040, hemodynamic verification is performed based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data.

Illustratively, a predefined hemodynamic model can be established to verify the input and output of the inference engine. That is, step S2040 may include: establishing a predefined hemodynamic model; and determining whether blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data conform to a predefined one The physical laws specified by the hemodynamic model to obtain hemodynamic test results.

The hemodynamic model has its existence value as a verification method of the inference engine, which can effectively avoid the situation that violates the physical law due to pure machine learning. Illustratively, the knowledge in the knowledge base generated by machine learning and the rules provided by the inference engine may be verified by the hemodynamic model, and only if it conforms to the basic physical laws, it can be considered as an effective knowledge or rule.

The principle of hemodynamic verification will be described below in conjunction with FIG. The most important blood vessels inside the brain form a ring structure that links the hemispheres to the anterior and posterior circulation, called the cerebral artery ring (Willis ring), as shown in Figure 3. The arteries in the brain are bilaterally symmetric. The main arteries include the anterior, middle, posterior P1, posterior P2, basilar, vertebral, common carotid, internal carotid and external carotid arteries. Since the brain is protected by the skull, and the skull is not conducive to the transmission of ultrasound signals, it is necessary to find the ratio on the skull. Thinner places ensure that the signal is properly acquired. Typical checkpoints are the sash window (checkpoint 1 in Figure 3) and the pillow window (checkpoint 2 in Figure 3). For the cervical artery (the common carotid artery, the internal carotid artery, and the external carotid artery), since the outside is only soft tissue protection, the examination is relatively easy (checkpoint 3 in Fig. 3). From the sacral window, the anterior, middle, and posterior arteries can usually be examined. Corresponding blood flow direction is the anterior artery away from the probe, the middle artery is toward the probe, the P1 segment of the posterior artery is facing the probe, and the P2 segment is facing away from the probe. From the occipital window examination, the basilar artery and vertebral artery can usually be examined. Corresponding blood flow direction is the basilar artery away from the probe, and the vertebral artery is away from the probe. The direction of blood flow is an important feature in transcranial Doppler examination. If the direction of blood flow is different from the normal direction (ie, the direction of blood flow does not conform to the physical law), it can be used as an important basis for disease diagnosis. Although the intracranial blood vessels are described as an example, in the actual use, the neck blood vessels (the common carotid artery, the internal carotid artery, the external carotid artery, etc.) can also follow this rule.

For example, if one side of the internal carotid artery is blocked, the blood supply to the ipsilateral middle and anterior arteries may be insufficient. The slowing of the blood flow velocity of the middle and anterior arteries is a hemodynamic rule. The increase in blood flow velocity in the middle and anterior arteries violates the hemodynamic model and is a false criterion. For another example, if one side of the internal carotid artery is blocked, the resistance index of the common carotid artery will increase, and the resistance index of the middle artery will decrease. Therefore, if the resistance index of the common carotid artery is decreased, it is related to hemodynamics. Inconsistent, it is also a criterion for further evaluation.

In step S2050, it is judged based on the hemodynamic check result whether the inference engine is qualified.

The conditions under which the inference engine passes the hemodynamic check can be set as needed. Illustratively, if there is only one group of blood flow data participating in the hemodynamic check, it may be specified that the blood flow characteristic parameter and the corresponding test conclusion satisfy the physical law in the blood flow data of the group, the blood flow force school The result of the test is that the inference engine passes the hemodynamic check, and the hemodynamic check result is that the inference engine fails the hemodynamic check. Illustratively, if more than one group of blood flow data is involved in the hemodynamic check, blood bleeds in blood flow data for a particular number (eg, 10 groups) or a particular ratio (eg, 80%) may be specified. When the flow characteristic parameters and the corresponding test conclusions satisfy the physical laws, the hemodynamic check result is the hemodynamic check by the inference engine, and the hemodynamic check result is that the inference engine fails the hemodynamic check.

If the inference engine passes the hemodynamic check, it can be determined that the inference engine is qualified, in which case the rules provided by the inference engine and/or certain diagnostic conclusions inferred by the inference engine and the corresponding blood can be exemplarily Flow feature parameters are stored in the knowledge base as a complement to expert knowledge. If the inference engine fails the hemodynamic check, it can be determined that the inference engine is unqualified, in which case the inference engine can be exemplarily retrained.

According to the inference engine training method of the embodiment of the present invention, the hemodynamic check is used to verify whether the inference engine after the training is qualified, thereby accurately knowing the training level of the current inference engine. If it is determined that the inference engine is unqualified, it may choose to continue training until the inference engine is qualified. This facilitates training to obtain a more efficient inference engine, thereby obtaining a more accurate and efficient auxiliary diagnostic system for transcranial Doppler examination.

According to an embodiment of the present invention, step S2050 may include: if it is determined that the inference engine fails, then go to step S2060; the inference engine training method 2000 may further include step S2060: outputting an indication for indicating that the inference engine fails the hemodynamic check information.

Illustratively, if it is determined that the inference engine has not passed the hemodynamic check, that is, the inference engine is determined to be unsatisfactory, the indication information may be output, for example, the text information "hemodynamic check failed" is displayed on the display screen. This facilitates notification to the operator (eg, expert) of the auxiliary diagnostic system whether the inference engine meets the requirements, and facilitates the operator (eg, an expert) to take timely measures to deal with the inference engine that meets the requirements as soon as possible.

According to an embodiment of the present invention, after the step S2060, the inference engine training method 2000 may further include the following steps.

At step S2070, receiving one or more of modifying a model for implementing the inference engine, parameters of the inference engine, blood flow data stored in the knowledge base, and a hemodynamic model employed in the hemodynamic check The first modification instruction.

In step S2080, the corresponding modification action is executed according to the first modification instruction and returns to step S2010.

The operator (such as an expert) knows that the inference engine does not meet the requirements when seeing the text information similar to the above "hemodynamic check fails". In this case, the operator (for example, an expert) can choose to modify the knowledge base. Raw blood flow data, adding new blood flow data to the knowledge base, modifying models used to implement the inference engine (eg, changing from a Bayesian classifier to a deep neural network), modifying parameters of the inference engine, or modifying blood flow Learning models, etc. An operator (for example, an expert) may input a modification instruction (referred to as a first modification instruction in this embodiment) that is desired to be executed into the auxiliary diagnosis system, and after receiving the instruction, the auxiliary diagnosis system will perform a corresponding modification action. After the modification action is performed, the method returns to step S2010, the training inference engine is restarted, and the above steps S2010 to S2050 are cycled, and if the inference engine is still unsatisfactory, steps S2060 to S2080 may be performed again. That is to say, steps S2010 to S2080 can be repeated until the inference engine is qualified. By timely adjusting the knowledge base, inference engine or hemodynamic model, it is possible to train a more robust inference engine to obtain a more accurate and reliable auxiliary diagnosis system.

According to an embodiment of the present invention, step S2050 may include: if it is determined that the inference engine is qualified, then go to Step S2090; the inference engine training method 2000 may further include step S2090: updating the known rules stored in the knowledge base with rules provided by the inference engine.

The inference engine can provide rules, and there are known rules that are initially stored in the knowledge base. After the inference engine is trained, the rules provided by the trained inference engine can be updated into the knowledge base. If the inference engine passes the hemodynamic check, then the inference engine can be considered to have been trained. Illustratively, known rules stored in the knowledge base may include rules pre-defined by the expert. Illustratively, the known rules may also include rules that the inference engine previously provided. It can be understood that the inference engine can be continuously trained and updated. When the number of knowledge (ie, blood flow data) in the knowledge base is small, the inference engine obtained by the training can provide some rules. As the knowledge in the knowledge base grows, the inference engine can be retrained, the inference engine obtained after retraining can provide some new rules, and so on. The rules provided by each trained inference engine can update the known rules currently stored in the knowledge base.

According to an embodiment of the present invention, step S2050 may include: if it is determined that the inference engine is qualified, go to step S2100; the inference engine training method 2000 may further include: step S2100: determining a rule provided by the inference engine and a known rule stored in the knowledge base Whether it is a conflict, if yes, go to step S2110; and step S2110: output indication information indicating that the rule conflicts.

After the inference engine passes the hemodynamic check, it can be judged by the conflict handling mechanism whether the rules provided by the inference engine conflict with known rules stored in the knowledge base (mainly rules stored by experts). For example, suppose that the known rules stored in the knowledge base specify that the mean blood flow velocity is greater than 70 cm/s and less than 30 cm/s indicates an abnormality, while the ruler provides rules that specify that the mean blood flow velocity is greater than 65 cm/s indicating an abnormality, then the reasoning The rules provided by the machine are not conflicting with the known rules stored in the knowledge base. If the rules provided by the inference engine stipulate that the average blood flow velocity is greater than 75 cm/s to indicate an abnormality, the rules provided by the inference engine and the stored in the knowledge base Knowing the rules is conflicting. That is, by way of example, the rules provided by the inference engine may be considered to be more stringent than the known rules stored in the knowledge base, and the rules provided by the inference engine may be considered to conflict with known rules stored in the knowledge base.

If the rules provided by the inference engine conflict with known rules stored in the knowledge base, the indication information may be output, for example, text information such as "rule conflict" is displayed on the display screen. This facilitates notification of the conflict between the inference rules of the operator (eg, expert) of the auxiliary diagnostic system and the known rules, so that the operator (such as an expert) can take timely measures to deal with the inference engine that meets the requirements as soon as possible.

According to an embodiment of the present invention, after step S2110, the inference engine training method 2000 may further include: step S2120: receiving one of modifying a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base. a second modification instruction of the item or a plurality of items; and step S2130: according to The second modification instruction executes the corresponding modification action and returns to step S2010.

For the conflict between the rules provided by the inference engine and the known rules stored in the knowledge base, the auxiliary diagnostic system may be re-optimized, such as modifying machine learning parameters or hemodynamic models, and retraining the inference engine. Conflict handling can be supplemented by experts when the operator's capabilities are insufficient.

According to an embodiment of the present invention, step S2100 may include: if the rule provided by the inference engine does not conflict with the known rule stored in the knowledge base, then go to step S2140; the inference engine training method 2000 may further include: step S2140: using the inference engine The provided rules update the known rules stored in the knowledge base.

In the case where it is determined that the provided rules do not conflict with the known rules stored in the knowledge base, the inference engine can be considered to be trained, and the rules provided by the trained inference engine can be updated into the knowledge base.

According to an embodiment of the present invention, after step S2010, the inference engine training method 2000 may further include step S2012: cross-validating the inference engine with the plurality of sets of blood flow data to evaluate the correct rate of the inference engine.

For training sets, K-Fold Cross Validation can be used to assess the correct rate. The general idea of K-time cross-validation is to roughly divide the data into K sub-samples, take one sub-sample each time as verification data, and take the remaining K-1 sub-samples as training data. The model used to implement the inference engine is constructed to act on the verification data to calculate the current correct rate (or error rate). Repeat K times and average the K correct rate (or error rate) to get an overall correct rate (or error rate). The correct rate (or error rate) of the current inference engine can be estimated by the overall correct rate (or error rate).

According to an embodiment of the present invention, step S2012 may include: if the correct rate is less than the preset threshold, proceeding to step S2014; the inference engine training method 2000 may further include: step S2014: outputting indication information indicating that the inference engine fails the cross-validation .

Illustratively, the preset threshold may be any suitable value, which is not limited by the invention. If the correct rate is greater than or equal to a predetermined threshold, the inference engine can be considered to be successful. The tolerance for errors in general clinical use is very low, so the preset threshold can be set very high, for example 95%. If the correct rate cannot meet the requirements, the target correct rate can be achieved by modifying the parameters of the inference engine, checking whether the knowledge base has errors, increasing the types of blood flow characteristic parameters in the blood flow data, and the like. When the correct rate is close to 100%, the conclusion equivalent to the inference engine is consistent with the conclusions given by the experts. The rules provided by the inference engine actually contain the knowledge given by the experts.

If the inference engine does not pass the cross-validation, the indication information can be output, for example, text information such as "cross-validation failed" is displayed on the display screen. This facilitates notification to the operator of the auxiliary diagnostic system (eg If the inference engine meets the requirements, it is convenient for the operator (such as an expert) to take timely measures to deal with the inference engine that meets the requirements as soon as possible.

According to an embodiment of the present invention, after the step S2014, the inference engine training method 2000 may further include the following steps.

At step S2016, a third modification instruction is received regarding modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base.

In step S2018, the corresponding modification action is executed according to the third modification instruction and returns to step S2010.

Similar to the case where the inference engine does not pass the hemodynamic check, the operator (for example, an expert) knows that the inference engine does not satisfy the requirement when seeing the text information similar to the above-mentioned "cross-validation failed", in which case, The operator (eg, an expert) may choose to modify the original blood flow data in the knowledge base, add new blood flow data to the knowledge base, modify the model used to implement the inference engine, modify the parameters of the inference engine, and the like. An operator (for example, an expert) may input a modification instruction (referred to as a third modification instruction in this embodiment) that is desired to be executed into the auxiliary diagnosis system, and after receiving the instruction, the auxiliary diagnosis system will perform a corresponding modification action. After performing the modification action, the method returns to step S2010 to restart the training of the inference engine. After the training, step S2012 can be performed again, that is, the trained inference engine is cross-validated again. If the inference engine still fails the cross-validation, steps S2010, S2012, and S2014 may be repeatedly performed until the inference engine passes the cross-validation. Cross-validation can improve the accuracy of the inference engine.

The training method of the inference engine in the auxiliary diagnosis system has been described above. After the inference engine is trained, it can be applied to the auxiliary diagnosis system. It can be understood that the training process of the inference engine and the application process of the auxiliary diagnosis system can be complementary, and the two can be implemented crosswise. For example, in the practical application of the auxiliary diagnostic system, the inference engine can be retrained as needed (regularly or irregularly) using the inference engine training method described above to achieve continuous improvement of the inference engine. In the future work of the auxiliary diagnostic system, the newly trained inference engine can be used for reasoning.

In addition, in the practical application of the auxiliary diagnosis system, other parts in the auxiliary diagnosis system, such as a knowledge base, an interpreter, and the like, can be continuously updated to implement an upgrade of the auxiliary diagnosis system. The upgrade of the auxiliary diagnostic system helps to provide a more accurate diagnosis of the diagnosis. An upgrade method of an auxiliary diagnosis system according to an embodiment of the present invention is described below.

4 shows a schematic flow diagram of an upgrade method 400 of an auxiliary diagnostic system in accordance with one embodiment of the present invention. As shown in FIG. 4, the upgrade method 400 includes the following steps.

In step S410, at least one or more sets of actual blood flow characteristic parameters are inferred by the inference engine obtained by the above-described inference engine training method 200 using the auxiliary diagnosis system to obtain one or more sets. One or more actual conclusions corresponding to the actual blood flow characteristic parameters of the group.

As described above, for the auxiliary diagnostic system, the inference engine can be retrained once when needed (regularly or irregularly) using the inference engine training method 200 described above. After the end of a certain inference engine training, the inference engine obtained by the training can be used to start the inference of the actual blood flow characteristic parameters. Preferably, the actual blood flow characteristic parameter may be a blood flow characteristic parameter in various real cases that need to be processed during the application of the auxiliary diagnostic system. Of course, the actual blood flow characteristic parameters may have other sources, such as simulation parameters in a simulation experiment, etc., which are not limited by the present invention.

Optionally, each set of actual blood flow characteristic parameters may also include the maximum systolic blood flow velocity, end-diastolic blood flow velocity, mean blood flow velocity, pulsation index, resistance index, blood flow reversal, and blood stealing described above. One or more of the parameters of eddy current, turbulence, and dash.

Illustratively, the one or more sets of actual blood flow characteristic parameters may be actual blood flow characteristic parameters that have not been processed by the auxiliary diagnostic system, or may be actual blood flow characteristic parameters previously processed by the auxiliary diagnostic system, which will be Described below.

As described above, the inference engine can provide rules that can be inferred for each set of actual blood flow characteristic parameters based on at least the rules provided by the inference engine to obtain a corresponding conclusion (this is referred to as the actual conclusion).

At step S420, one or more actual conclusions are output.

One or more actual conclusions can be output through the user interface described above for viewing by an operator (eg, an expert).

At step S430, indication information regarding at least a portion of the actual conclusions in the one or more actual conclusions is received.

After the operator (such as an expert) looks at the actual conclusions, it can analyze which actual conclusions belong to the common, more common cases, and which actual conclusions belong to the less common atypical cases. Subsequently, the operator (such as an expert) can choose to The actual conclusions corresponding to which type of case are stored. For example, an operator (eg, an expert) can interact with the auxiliary diagnostic system through the user interface, instructing the auxiliary diagnostic system to store the actual conclusions corresponding to the atypical case and the actual blood flow characteristic parameters corresponding to the actual conclusion in the knowledge base as new knowledge. Illustratively, after the new knowledge is stored in the knowledge base (ie, the knowledge base is updated), the inference engine can be retrained based on the updated knowledge base. In addition, after the new knowledge is stored in the knowledge base (that is, the knowledge base is updated), it can also be generated based on the updated knowledge base and the current inference engine (which may be an inference engine obtained after retraining based on the updated knowledge base). New interpreter.

At step S440, based on the indication information, at least part of the actual conclusion and at least part of the actual conclusion The one-to-one correspondence of at least one set of actual blood flow characteristic parameters is stored in a knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

In one example, one or more sets of actual blood flow characteristic parameters may be actual blood flow characteristic parameters that the auxiliary diagnostic system currently needs to process, ie, new actual blood flow characteristic parameters that the auxiliary diagnostic system has not previously processed. That is to say, after receiving the new actual blood flow characteristic parameter in the application process, the auxiliary diagnosis system infers the new actual blood flow characteristic parameter and obtains the corresponding actual conclusion. At this time, if the operator (for example, an expert) thinks that an actual conclusion belongs to a more valuable case (for example, an atypical case), the actual conclusion and a set of actual blood flow characteristic parameters corresponding to the actual conclusion can be stored in the knowledge. In the library.

In another example, the database of the auxiliary diagnostic system is used to store actual blood flow characteristic parameters previously processed by the auxiliary diagnostic system and previous conclusions obtained by reasoning for previously processed actual blood flow characteristic parameters, one or more sets of actual blood flow The characteristic parameters are the previously processed actual blood flow characteristic parameters from the database of the auxiliary diagnostic system.

The database accumulates the raw data, intermediate results and final conclusions required for reasoning in the application process of the auxiliary diagnosis system. The raw data in the database may include the blood flow characteristic parameters described above (referred to as "actual blood flow characteristic parameters" in the description of the upgrade method 400). That is to say, the actual blood flow characteristic parameters that the auxiliary diagnostic system has previously processed and the corresponding previous conclusions can be stored in the database. With the application time of the auxiliary diagnosis system, the more data accumulated in the database, and the accumulated in the database are mainly the various cases encountered in reality, so it has very important practical significance and reference value. These cases can be used to augment the knowledge base for some of the less common atypical cases in the database. For example, after a training acquires a new inference engine, the inference engine obtained by the training can be used to re-infer the previously processed actual blood flow characteristic parameters stored in the database to obtain corresponding actual conclusions. This processing can correct some of the previous inference errors and help to discover important information that was previously ignored. For example, from the actual conclusions obtained from re-reasoning, it is possible to find some cases that have not been discovered or are not valued. If some valuable cases are found, the actual conclusions corresponding to such cases and the relevant actual blood flow characteristics can be found. The parameters are stored in the database as new knowledge. This also enables the expansion of the knowledge base, that is, the upgrade of the auxiliary diagnostic system.

The auxiliary diagnostic system needs to be continuously upgraded during the application process. For example, when the auxiliary diagnostic system is initially set up, the data in the knowledge base may be insufficient. Through the upgrade method 400, the database can be continuously expanded, thereby achieving continuous upgrading of the auxiliary diagnosis system.

According to an embodiment of the invention, at least part of the actual conclusion is that it is compared with a manually selected atypical case The actual conclusions should be. At least some of the actual conclusions can be manually selected, and exemplarily, can be selected by an expert. With the application of the auxiliary diagnostic system, a large amount of new data will be added to the database, and there may be some unusual atypical cases. These atypical conditions can be referred to human experts for further analysis. That is to say, the operator (such as an expert) can independently select the actual conclusion corresponding to the atypical case. Such data has very important practical significance and reference value, which is beneficial to improve the effectiveness of the auxiliary diagnosis system.

According to an embodiment of the present invention, step S410 may include: synthesizing rules provided by the inference engine and known rules stored in the knowledge base of the auxiliary diagnosis system to obtain the integrated rules; and grouping one or more groups based on the integrated rules The actual blood flow characteristic parameters are reasoned.

Illustratively, an interpreter can be generated by synthesizing the rules provided by the inference engine with known rules stored in the knowledge base. For example, the known rules in the knowledge base and the rules provided by the inference engine can be converted into a set of functions F = (f1, f2 ... fn), where n can be the number of conclusions. Each function in f1, f2...fn outputs a conclusion. Assuming that the actual blood flow characteristic parameters are represented by the vector T described above, the vector T can be placed in the interpreter F for calculation, and then the output with the largest value among the obtained n outputs is taken as the final auxiliary diagnosis conclusion. (ie the above actual conclusions).

According to an embodiment of the present invention, the database of the auxiliary diagnosis system is used to store feedback information about whether the previous conclusion is adopted, and the upgrade method 400 may further include: determining the number of times the previous conclusion of the specific type is not adopted according to the feedback information stored in the database. Exceeding the number of thresholds, outputting prompt information about a particular type of previous conclusion that may be problematic; receiving one or more of modifying the model used to implement the inference engine, the parameters of the inference engine, and the known rules stored in the knowledge base The modification instruction; and the corresponding modification action is performed according to the modification instruction.

Illustratively, the database may also store some information entered by the operator (eg, an expert) of the auxiliary diagnostic system during the human-computer interaction process, such as its opinion on the auxiliary diagnostic conclusion provided by the auxiliary diagnostic system (eg, whether the auxiliary diagnosis conclusion can be Adoption, some additional comments on the conclusions of the auxiliary diagnosis, etc.).

Although human experts have a very high professional quality, they may also make mistakes. It is also impossible for the system design to be entirely expert-oriented, so it is possible to design a correction mechanism to minimize the errors of human experts. When the auxiliary diagnosis system is actually applied, the operator's adoption of the auxiliary diagnosis conclusion is an important indicator. This data can be analyzed periodically. If a conclusion is found to be frequently changed or frequently adopted by the operator, the evaluation can be re-evaluated. Conclusion and the rules on which the conclusion is based. The assessment can be carried out by the operator (mainly an expert). The operator (mainly an expert) can judge the reason for the inaccuracy of the conclusion, for example, it may be that the known rules stored in the knowledge base are unreasonable, the rules provided by the inference engine are unreasonable, etc. And develop a corresponding revision plan. The operator (mainly an expert) can interact with the auxiliary diagnostic system through the user interface to instruct the auxiliary diagnostic system to modify certain places, such as modifying the model used to implement the inference engine and/or modifying the parameters of the inference engine to modify the inference engine The rules provided. As another example, an operator (primarily an expert) can instruct the secondary diagnostic system to modify known rules and the like stored in the knowledge base.

Illustratively, the upgrade method 400 can further include: for each of the plurality of blood vessels, at least using an inference engine to infer a set of actual blood flow characteristic parameters associated with the blood vessel to obtain a blood vessel correlation corresponding to the blood vessel Conclusion; blood vessel related conclusions corresponding to the blood vessel are output in real time; and feedback information on whether or not the blood vessel related conclusion corresponding to the blood vessel is accepted is received and stored in real time.

For example, transcranial Doppler examination usually involves examining about 10 blood vessels, and the operator and the auxiliary diagnostic system can interact during the examination and at the end of all examinations. The multiple blood vessels that need to be examined can be examined one by one. Once a blood vessel is examined, a set of actual blood flow characteristic parameters associated with the blood vessel can be obtained. During the examination, the auxiliary diagnosis system can output the auxiliary diagnosis conclusion (ie, the blood vessel-related conclusion corresponding to the blood vessel) in real time based on the obtained information. The auxiliary diagnosis conclusion of the real-time output, the operator can determine whether to adopt, and can determine whether to change the default inspection order according to the current auxiliary diagnosis conclusion. For example, if it is determined that there is a problem with a certain part based on the current auxiliary diagnosis result, and the blood vessel of the part is not the next inspection target, the operator can input an instruction to adjust the inspection order. In addition, the operator can also choose to maintain the default inspection order, and wait until the inspection is completed. The feedback information of the operator can be obtained by the above method, and the feedback information is feedback information for each blood vessel. Subsequent corrections to the knowledge base or inference engine based on the feedback information can be used to further upgrade the auxiliary diagnostic system. The present invention is not limited to the above-described implementation, and for example, it is possible to output only the final general conclusion without outputting a blood vessel related conclusion associated with each blood vessel to the operator.

Illustratively, the upgrade method 400 may further include: inferring, by using at least an inference engine, a plurality of sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence to obtain a plurality of blood vessel related conclusions corresponding to the plurality of blood vessels one-to-one Determining a general conclusion based on a plurality of blood vessel related conclusions; outputting a general conclusion; and receiving and storing feedback information as to whether the overall conclusion is adopted.

After all vascular examinations are completed, the auxiliary diagnostic system can give an overall test report recommendation (which can be called a “total conclusion”). If the operator accepts it, it can be signed and confirmed; if the operator feels incomplete or not correct, you can choose Not adopted. The feedback information of the operator can be obtained by the above method, and the feedback information is overall feedback information for a plurality of blood vessels. Subsequent corrections to the knowledge base or inference engine based on the feedback information can be used to further upgrade the auxiliary diagnostic system.

Illustratively, each time the knowledge base and the inference engine are updated, they can be regenerated Interpreter. That is, a new interpreter in the auxiliary diagnostic system can be generated based on the knowledge base and/or the inference engine. The update of the interpreter is also an upgrade to the secondary diagnostic system. The interpreter can simply classify actual conclusions, such as "common", "uncommon", etc., and human experts can choose more valuable practical conclusions. The classification of interpreters can greatly reduce the work of human experts, giving human experts the opportunity to focus on more characteristic cases, thus greatly improving the iterative upgrade efficiency of the knowledge base.

The upgrade method of the auxiliary diagnosis system described above is an autonomous upgrade method. With the method, the auxiliary diagnosis system can be automatically upgraded in the actual application process, and the user is continuously improved, so that the correct rate and performance of the system can be continuously improved. .

According to another aspect of the present invention, an inference engine training device is provided. FIG. 5 shows a schematic block diagram of an inference engine training device 500 in accordance with one embodiment of the present invention.

As shown in FIG. 5, the inference engine training device 500 according to an embodiment of the present invention includes a training module 510, a selection module 520, an inference module 530, a hemodynamic verification module 540, and an eligibility determination module 550. The various modules may perform the various steps/functions of the inference engine training method described above in connection with Figures 1-3, respectively. Only the main functions of the components of the inference engine training device 500 will be described below, and the details already described above are omitted.

The training module 510 is configured to utilize the plurality of sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions.

The selection module 520 is configured to select at least one set of blood flow data from the knowledge base.

The inference module 530 is configured to infer the blood flow characteristic parameters in the blood flow data of the set of blood flow data for each set of blood flow data in the at least one set of blood flow data to obtain a corresponding test conclusion.

The hemodynamic check module 540 is configured to perform a hemodynamic check based on blood flow characteristic parameters and corresponding test conclusions in each of the at least one set of blood flow data.

The qualification determination module 550 is configured to determine whether the inference engine is qualified based on the hemodynamic verification result.

According to an embodiment of the present invention, the inference engine training device 500 further includes a first output module (not shown), and the pass judgment module 550 includes: a first boot submodule, configured to start the first output module if it is determined that the inference engine is unqualified The first output module is configured to output indication information for indicating that the inference engine has not passed the hemodynamic check.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes: a first receiving module (not shown) for receiving data about modifying a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base. And a first modification instruction of one or more of the hemodynamic models employed by the hemodynamic check; and a first execution module (not shown) for performing a corresponding modification based on the first modification instruction Make and start the training module.

According to an embodiment of the present invention, the inference engine training device 500 further includes a first update module (not shown), and the qualification determination module 550 includes: a second promoter module, configured to start the first update module if it is determined that the inference engine is qualified; The first update module is for updating the known rules stored in the knowledge base with rules provided by the inference engine.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes a conflict determination module and a second output module (not shown), and the qualification determination module 550 includes: a third promoter module for initiating a conflict if it is determined that the inference engine is qualified a judging module; the conflict judging module is configured to judge whether the rule provided by the inference engine conflicts with a known rule stored in the knowledge base, and if so, start a second output module; and the second output module is configured to output a rule for indicating that the rule conflicts Instructions.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes: a second receiving module (not shown) for receiving data about modifying the model for implementing the inference engine, the parameters of the inference engine, and the blood flow data stored in the knowledge base. And a second execution instruction (not shown) for performing a corresponding modification action and starting the training module according to the second modification instruction.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes a second update module (not shown), and the conflict determination module includes: a fourth promoter module for using the rules provided in the knowledge base and the known knowledge stored in the knowledge base If the rules do not conflict, the second update module is started; the second update module is used to update the known rules stored in the knowledge base with the rules provided by the inference engine.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes: a cross-validation module (not shown) for utilizing the plurality of sets of blood after the training module utilizes the plurality of sets of blood flow data stored in the knowledge base as the sample training inference engine The stream data cross-validates the inference engine to evaluate the correctness of the inference engine.

According to an embodiment of the present invention, the inference engine training device 500 further includes a third output module (not shown), and the cross-validation module includes: a third promoter module, configured to start the third output module if the correct rate is less than the preset threshold The third output module is configured to output indication information indicating that the inference engine has not passed the cross-validation.

According to an embodiment of the present invention, the inference engine training apparatus 500 further includes: a third receiving module (not shown) for receiving data about modifying the model for implementing the inference engine, the parameters of the inference engine, and the blood flow data stored in the knowledge base. And a third execution module (not shown) for performing a corresponding modification action and starting the training module according to the third modification instruction.

According to an embodiment of the invention, the inference engine is implemented using a Bayesian classifier, a support vector machine or a deep neural network.

According to an embodiment of the invention, the hemodynamic verification module 540 includes: a setup sub-module for establishing a predefined hemodynamic model; and a determination sub-module for determining each set of blood in at least one set of blood flow data The blood flow characteristic parameters and corresponding test conclusions in the flow data conform to the physical laws prescribed by the predefined hemodynamic model to obtain hemodynamic test results.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.

According to another aspect of the present invention, an upgrade apparatus for an auxiliary diagnostic system is provided. FIG. 6 shows a schematic block diagram of an upgrade apparatus 600 of an auxiliary diagnostic system in accordance with one embodiment of the present invention.

As shown in FIG. 6, the upgrading apparatus 600 of the auxiliary diagnosis system according to the embodiment of the present invention includes a first inference module 610, a first output module 620, a first receiving module 630, and a first storage module 640. The various modules may perform the various steps/functions of the upgrade method of the auxiliary diagnostic system described above in connection with FIG. 4, respectively. The main functions of the components of the upgrading apparatus 600 of the auxiliary diagnostic system will be described below, and the details already described above are omitted.

The first inference module 610 is configured to infer one or more sets of actual blood flow characteristic parameters by using an inference engine trained by the above inference engine training method to at least utilize an auxiliary inference system to obtain one or more sets of actual blood flow. One or more actual conclusions in which the feature parameters correspond one-to-one.

The first output module 620 is configured to output one or more actual conclusions.

The first receiving module 630 is configured to receive indication information about at least part of the actual conclusions of the one or more actual conclusions.

The first storage module 640 is configured to store at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with at least part of the actual conclusions into the knowledge base of the auxiliary diagnosis system based on the indication information for implementing the upgrade of the auxiliary diagnosis system. .

According to an embodiment of the invention, the database of the auxiliary diagnostic system is used to store the actual blood flow characteristic parameters previously processed by the auxiliary diagnostic system and previous conclusions obtained from the inference of the previously processed actual blood flow characteristic parameters, one or more sets of actual blood flow The characteristic parameters are the previously processed actual blood flow characteristic parameters from the database of the auxiliary diagnostic system.

According to an embodiment of the invention, at least part of the actual conclusion is the actual conclusion corresponding to the manually selected atypical case.

According to an embodiment of the present invention, the inference module 610 includes: an integration sub-module for synthesizing the rules provided by the inference engine and the known rules stored in the knowledge base of the auxiliary diagnosis system to obtain the integrated rules; and the inference sub-module One or more sets of actual blood flow characteristic parameters are reasoned based on the integrated rules.

According to an embodiment of the present invention, the database of the auxiliary diagnosis system is configured to store feedback information about whether the previous conclusion is adopted, and the upgrading apparatus further includes: a second output module (not shown) for determining according to the feedback information stored in the database If the number of times the previous conclusion of the particular type is not adopted exceeds the threshold of the number of times, then the prompt information may be output regarding the previous conclusion of the particular type; the second receiving module (not shown) is configured to receive the modification for implementing the inference engine. A modification instruction of one or more of a model, a parameter of the inference engine, and a known rule stored in the knowledge base; and a modification module (not shown) for performing a corresponding modification action according to the modification instruction.

According to an embodiment of the invention, the upgrading apparatus further comprises: a second inference module (not shown) for using, for each of the plurality of blood vessels, at least one set of actual blood flow characteristic parameters associated with the blood vessel using the inference engine Reasoning is performed to obtain a blood vessel related conclusion corresponding to the blood vessel; a third output module (not shown) for outputting a blood vessel related conclusion corresponding to the blood vessel in real time; and a second storage module (not shown) for Feedback information on whether or not the blood vessel related conclusion corresponding to the blood vessel is accepted is received and stored in real time.

According to an embodiment of the present invention, the upgrading apparatus further includes: a third inference module (not shown), configured to infer, by using at least an inference engine, a plurality of sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence to obtain Multiple vessel-corresponding multiple blood vessel-related conclusions; a general conclusion determination module (not shown) for determining a total conclusion based on a plurality of blood vessel-related conclusions; a fourth output module (not shown) for outputting a general conclusion And a third storage module (not shown) for receiving and storing feedback information as to whether the overall conclusion is accepted.

According to an embodiment of the invention, one or more sets of actual blood flow characteristic parameters are actual blood flow characteristic parameters that the auxiliary diagnostic system currently needs to process, and the upgrading apparatus further includes: a generating module (not shown) for the knowledge base and/or Or the inference engine generates a new interpreter in the auxiliary diagnostic system.

FIG. 7 shows a schematic block diagram of an inference engine training device 700 in accordance with one embodiment of the present invention. The inference engine training device 700 includes a memory 710 and a processor 720.

The memory 710 stores program code (i.e., program) for implementing respective steps in the inference engine training method according to an embodiment of the present invention.

The processor 720 is configured to execute program code stored in the memory 710 to perform respective steps of an inference engine training method according to an embodiment of the present invention, and to implement an embodiment according to the present invention. The inference engine training device 500 has a training module 510, a selection module 520, an inference module 530, a hemodynamic verification module 540, and an eligibility determination module 550.

In one embodiment, the program code, when running in the processor 720, is configured to perform the following steps: Step S2010: using a plurality of sets of blood flow data stored in a knowledge base as a sample training inference engine, wherein the knowledge base Each set of blood flow data stored therein includes blood flow characteristic parameters and corresponding known conclusions; step S2020: selecting at least one set of blood flow data from the knowledge base; step S2030: for each set of blood in at least one set of blood flow data Flow data, using an inference engine to infer blood flow characteristic parameters in the blood flow data to obtain corresponding test conclusions; Step S2040: based on blood flow characteristics in each set of blood flow data in at least one set of blood flow data The parameters and corresponding test conclusions are subjected to hemodynamic verification; and step S2050: determining whether the inference engine is qualified based on the hemodynamic verification result.

In one embodiment, when the program code is running by the processor 720, the step S2050 for performing includes: if it is determined that the inference engine fails, then the process goes to step S2060; the program code is at the processor 720. When running in the middle, it is also used to perform the following steps: Step S2060: Output indication information indicating that the inference engine has not passed the hemodynamic check.

In one embodiment, after the step S2060 for execution of the program code in the processor 720, the program code is further configured to perform the following steps when the processor 720 is run: step S2070 Receiving a first modification of one or more of the hemodynamic models employed in modifying the model used to implement the inference engine, the parameters of the inference engine, the blood flow data stored in the knowledge base, and the hemodynamics check And the step S2080: executing the corresponding modification action according to the first modification instruction and returning to step S2010.

In one embodiment, when the program code is run in the processor 720, the step S2050 for performing includes: if it is determined that the inference engine is qualified, then the process goes to step S2090; the program code is at the processor 720. In the middle run, it is also used to perform the following steps: Step S2090: Update the known rules stored in the knowledge base with the rules provided by the inference engine.

In one embodiment, when the program code is run in the processor 720, the step S2050 for performing includes: if it is determined that the inference engine is qualified, then the process goes to step S2100; the program code is at the processor 720. The middle runtime is further configured to perform the following steps: Step S2100: judge whether the rule provided by the inference engine conflicts with a known rule stored in the knowledge base, if yes, go to step S2110; and step S2110: output is used to indicate that the rule occurs Instructions for conflicts.

In one embodiment, after step S2110 for execution of the program code in the processor 720, the program code is also used to execute when executed in the processor 720. The following steps: Step S2120: receiving a second modification instruction for modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base; and step S2130: according to the second The modification instruction executes the corresponding modification action and returns to step S2010.

In one embodiment, when the program code is run in the processor 720, the step S2100 for performing includes: if the rule provided by the inference engine does not conflict with a known rule stored in the knowledge base, then go to the step S2140; when the program code is run in the processor 720, is further configured to perform the following steps: Step S2140: Update the known rules stored in the knowledge base with rules provided by the inference engine.

In one embodiment, after the step S2010 for execution of the program code in the processor 720, the program code is further configured to perform the following steps when the processor 720 is run: step S2012 : Cross-validation of the inference engine using multiple sets of blood flow data to evaluate the correct rate of the inference engine.

In one embodiment, when the program code is run in the processor 720, the step S2012 for performing includes: if the correct rate is less than the preset threshold, then going to step S2014; the program code is in the process The runtime in the 720 is also used for execution: Step S2014: Outputting indication information indicating that the inference engine has not passed the cross-validation.

In one embodiment, after the step S2014 for execution of the program code in the processor 720, the program code is further configured to perform the following steps when the processor 720 is run: step S2016 Receiving a third modification instruction for modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base; and step S2018: performing the corresponding according to the third modification instruction Modify the action and return to step S2010.

In one embodiment, the inference engine is implemented using a Bayesian classifier, a support vector machine, or a deep neural network.

In one embodiment, when the program code is run in the processor 720, the step S2040 used to perform includes: establishing a predefined hemodynamic model; and determining each of the at least one set of blood flow data The blood flow characteristic parameters and corresponding test conclusions in the blood flow data conform to the physical laws prescribed by the predefined hemodynamic model to obtain hemodynamic verification results.

FIG. 8 shows a schematic block diagram of an upgrade apparatus 800 of an auxiliary diagnostic system in accordance with one embodiment of the present invention. The upgrade device 800 of the auxiliary diagnostic system includes a memory 810 and a processor 820.

The memory 810 stores program code (ie, a program) for implementing respective steps in an upgrade method of the auxiliary diagnostic system according to an embodiment of the present invention.

The processor 820 is configured to execute program code stored in the memory 810 to perform respective steps of an upgrade method of the auxiliary diagnostic system according to an embodiment of the present invention, and to implement an auxiliary diagnostic system according to an embodiment of the present invention. The first inference module 610, the first output module 620, the first receiving module 630, and the first storage module 640 in the device 600 are upgraded.

In one embodiment, the program code, when executed in the processor 820, is configured to perform at least one or more groups of inference engines trained by the above-described inference engine training method using at least the auxiliary diagnostic system. The actual blood flow characteristic parameters are inferred to obtain one or more actual conclusions that correspond one-to-one with one or more sets of actual blood flow characteristic parameters; one or more actual conclusions are output; and received in one or more actual conclusions At least part of the actual conclusion indication information; and storing at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with at least part of the actual conclusion based on the indication information into the knowledge base of the auxiliary diagnosis system for implementing the auxiliary diagnosis System upgrade.

In one embodiment, the database of the auxiliary diagnostic system is used to store actual blood flow characteristic parameters previously processed by the auxiliary diagnostic system and previous conclusions obtained by reasoning for previously processed actual blood flow characteristic parameters, one or more sets of actual blood flow The characteristic parameters are the previously processed actual blood flow characteristic parameters from the database of the auxiliary diagnostic system.

In one embodiment, at least some of the actual conclusions are actual conclusions corresponding to manually selected atypical cases.

In one embodiment, the program code, when executed in the processor 820, is used to perform at least one or more groups of inference engines trained by the above-described inference engine training method using the auxiliary diagnostic system. The step of inferring the actual blood flow characteristic parameter comprises: synthesizing the rules provided by the inference engine and the known rules stored in the knowledge base of the auxiliary diagnosis system to obtain the integrated rule; and grouping one or more groups based on the integrated rule The actual blood flow characteristic parameters are reasoned.

According to an embodiment of the present invention, a database of the auxiliary diagnostic system is used to store feedback information as to whether the previous conclusion was adopted, and when the program code is run in the processor 820, the program code is further configured to perform the following steps: if stored according to the database The feedback information determines that the number of times the previous conclusion of the particular type has not been adopted exceeds the threshold of the number of times, and outputs prompt information that may be problematic with respect to a particular type of previous conclusion; receiving parameters and knowledge about modifying the model used to implement the inference engine, the inference engine A modification instruction of one or more of the known rules stored in the library; and performing a corresponding modification action according to the modification instruction.

In one embodiment, the program code, when run in the processor 820, is further configured to perform the step of, for each of the plurality of blood vessels, using at least an inference engine to associate a set of actuals associated with the blood vessel Blood flow characteristic parameters are reasoned to obtain blood vessel related conclusions corresponding to the blood vessel; real time A blood vessel related conclusion corresponding to the blood vessel is output; and feedback information as to whether or not the blood vessel related conclusion corresponding to the blood vessel is accepted is received and stored in real time.

In one embodiment, the program code, when run in the processor 820, is further configured to perform the following steps: using at least an inference engine to infer multiple sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence Obtaining a plurality of blood vessel related conclusions corresponding to the plurality of blood vessels one by one; determining a total conclusion based on the plurality of blood vessel related conclusions; outputting the total conclusion; and receiving and storing feedback information as to whether the total conclusion is adopted.

In one embodiment, one or more sets of actual blood flow characteristic parameters are actual blood flow characteristic parameters that the diagnostic diagnostic system currently needs to process, and the program code, when run in the processor 820, is also used to perform the following steps : Generating a new interpreter in the auxiliary diagnostic system based on the knowledge base and/or inference engine.

Moreover, according to an embodiment of the present invention, there is also provided a computer readable storage medium on which program instructions (ie, programs) are stored, which are used to execute the program when the program instructions are executed by a computer or a processor Corresponding steps of the inference engine training method or the upgrade method of the auxiliary diagnostic system of the embodiment of the invention, and for implementing respective modules in the inference engine training device or the upgrade device of the auxiliary diagnostic system according to an embodiment of the present invention. The storage medium may include, for example, a memory card of a smart phone, a storage unit of a tablet, a hard disk of a personal computer, a read only memory (ROM), an erasable programmable read only memory (EPROM), a portable compact disk read only memory. (CD-ROM), USB memory, or any combination of the above storage media.

In one embodiment, the computer program instructions, when executed by a computer or processor, may cause a computer or processor to implement various functional modules of an inference engine training device or an upgrade device of an auxiliary diagnostic system in accordance with an embodiment of the present invention, and/ Alternatively, an inference engine training method or an upgrade method of the auxiliary diagnosis system according to an embodiment of the present invention may be performed.

In one embodiment, the computer program instructions are used at runtime to perform the following steps: Step S2010: using a plurality of sets of blood flow data stored in a knowledge base as a sample training inference engine, wherein each set of blood stored in the knowledge base The flow data includes blood flow characteristic parameters and corresponding known conclusions; step S2020: selecting at least one set of blood flow data from the knowledge base; step S2030: utilizing the inference engine for each set of blood flow data in the at least one set of blood flow data The blood flow characteristic parameters in the blood flow data are reasoned to obtain corresponding test conclusions; Step S2040: based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in at least one set of blood flow data Performing a hemodynamic check; and step S2050: determining whether the inference engine is qualified based on the hemodynamic check result.

In one embodiment, the computer program instructions are used to perform step S2050 at runtime. Including: if it is determined that the inference engine fails, then go to step S2060; the computer program instructions are also used to perform the following steps during operation: step S2060: output indication information indicating that the inference engine has not passed the hemodynamic check.

In one embodiment, after the step S2060 for execution of the computer program instructions at runtime, the computer program instructions are further configured to perform the following steps at runtime: step S2070: receiving modifications regarding implementation of the inference engine a first modification instruction of the model, the parameters of the inference engine, the blood flow data stored in the knowledge base, and one or more of the hemodynamic models employed in the hemodynamic check; and step S2080: according to the first modification The instruction executes the corresponding modification action and returns to step S2010.

In one embodiment, the step S2050 of the computer program instructions for execution at runtime comprises: if it is determined that the inference engine is qualified, then proceeds to step S2090; the computer program instructions are further configured to perform the following steps at runtime: S2090: Update the known rules stored in the knowledge base with the rules provided by the inference engine.

In one embodiment, the step S2050 of the computer program instructions for execution at runtime comprises: if it is determined that the inference engine is qualified, then proceeds to step S2100; the computer program instructions are further configured to perform the following steps at runtime: S2100: judge whether the rule provided by the inference engine conflicts with the known rule stored in the knowledge base, if yes, go to step S2110; and step S2110: output indication information indicating that the rule conflicts.

In one embodiment, after the step S2110 used by the computer program instructions to be executed at runtime, the computer program instructions are further configured to perform the following steps at runtime: step S2120: receiving a modification for implementing the inference engine a second modification instruction of one or more of the model, the parameters of the inference engine, and the blood flow data stored in the knowledge base; and step S2130: performing a corresponding modification action according to the second modification instruction and returning to step S2010.

In one embodiment, the step S2100 for executing the computer program instructions at runtime includes: if the rules provided by the inference engine do not conflict with the known rules stored in the knowledge base, then proceeding to step S2140; the computer program The instructions are also used at runtime to perform the following steps: Step S2140: Update the known rules stored in the knowledge base with rules provided by the inference engine.

In one embodiment, after the step S2010 for execution of the computer program instructions at runtime, the computer program instructions are further configured to perform the following steps at runtime: step S2012: using multiple sets of blood flow data pair inference engines Cross-validation is performed to assess the correct rate of the inference engine.

In one embodiment, the step S2012 used by the computer program instructions to be executed at runtime includes: if the correct rate is less than the preset threshold, then proceeding to step S2014; the computer program instructions are The runtime is also used to perform the following steps: Step S2014: Output indication information indicating that the inference engine has not passed the cross-validation.

In one embodiment, after the step S2014 for execution of the computer program instructions at runtime, the computer program instructions are further used at runtime to perform the following steps: Step S2016: receiving a modification regarding the implementation of the inference engine a third modification instruction of one or more of the model, the parameters of the inference engine, and the blood flow data stored in the knowledge base; and step S2018: performing a corresponding modification action according to the third modification instruction and returning to step S2010.

In one embodiment, the step S2040 of the computer program instructions for execution at runtime includes: establishing a predefined hemodynamic model; and determining blood in each of the at least one set of blood flow data The flow characteristic parameters and corresponding test conclusions conform to the physical laws specified by the predefined hemodynamic model to obtain hemodynamic test results.

In one embodiment, the computer program instructions are operative at runtime to perform at least one or more sets of actual blood flow characteristic parameters using at least an inference engine trained by the above-described inference engine training method using an auxiliary diagnostic system. Reasoning to obtain one or more actual conclusions that correspond one-to-one with one or more sets of actual blood flow characteristic parameters; output one or more actual conclusions; receive at least partial actual conclusions about one or more actual conclusions Instructing information; and storing, based on the indication information, at least a portion of the actual conclusions and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with at least a portion of the actual conclusions into a knowledge base of the auxiliary diagnostic system for implementing an upgrade of the auxiliary diagnostic system.

In one embodiment, the computer program instructions, at runtime, perform at least one of the set of actual blood flow characteristic parameters by using an inference engine trained by the above-described inference engine training method using the auxiliary diagnostic system. The steps of reasoning include: synthesizing the rules provided by the inference engine and the known rules stored in the knowledge base of the auxiliary diagnosis system to obtain the integrated rules; and performing one or more sets of actual blood flow characteristic parameters based on the integrated rules reasoning.

According to an embodiment of the invention, the database of the auxiliary diagnostic system is used to store feedback information as to whether the previous conclusions were adopted, the computer program instructions being also used at runtime to perform the step of determining a particular type based on feedback information stored in the database If the number of previous conclusions that have not been adopted exceeds the threshold of the number of times, then the prompt information about possible problems with the previous conclusion of the particular type is output; the model for modifying the model used to implement the inference engine, the parameters of the inference engine, and the known knowledge stored in the knowledge base are received. A modification instruction of one or more of the rules; and performing a corresponding modification action according to the modification instruction.

In one embodiment, the computer program instructions are further operative to perform the step of: for each of the plurality of blood vessels, at least using an inference engine to reason a set of actual blood flow characteristic parameters associated with the blood vessel Obtaining a blood vessel-related conclusion corresponding to the blood vessel; real-time outputting a blood vessel-related conclusion corresponding to the blood vessel; and receiving and storing, in real time, feedback information regarding whether or not the blood vessel-related conclusion corresponding to the blood vessel is adopted.

In one embodiment, the computer program instructions are further, at runtime, performing the step of: inferring, by using at least an inference engine, a plurality of sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence to obtain a plurality of roots A plurality of blood vessel-related conclusions corresponding to the blood vessels one by one; determining a total conclusion based on a plurality of blood vessel related conclusions; outputting a total conclusion; and receiving and storing feedback information as to whether the total conclusion is adopted.

In one embodiment, one or more sets of actual blood flow characteristic parameters are actual blood flow characteristic parameters that the diagnostic diagnostic system currently needs to process, and the computer program instructions are also used at runtime to perform the following steps: based on the knowledge base and/or Or the inference engine generates a new interpreter in the auxiliary diagnostic system.

Modules in the inference engine training apparatus according to embodiments of the present invention may be implemented by a processor executing an inference engine trained electronic device according to an embodiment of the present invention running computer program instructions stored in a memory, or may be in accordance with the present invention The computer instructions stored in the computer readable storage medium of the computer program product of the embodiments are implemented by the computer when executed.

Similarly, each module in the upgrade apparatus of the auxiliary diagnostic system according to an embodiment of the present invention may be implemented by a computer program instruction stored in a memory of a processor of an electronic device implementing an upgrade of the auxiliary diagnostic system according to an embodiment of the present invention. Or may be implemented when computer instructions stored in a computer readable storage medium of a computer program product according to an embodiment of the invention are executed by a computer.

Although the example embodiments have been described herein with reference to the drawings, it is understood that the foregoing exemplary embodiments are only illustrative, and are not intended to limit the scope of the invention. A person skilled in the art can make various changes and modifications without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the present invention as claimed.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another device, or some features can be ignored or not executed.

In the description provided herein, numerous specific details are set forth. However, it is understood that the embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of the description.

Similarly, the various features of the present invention are sometimes grouped together into a single embodiment, figure, in the description of exemplary embodiments of the invention, in the description of the exemplary embodiments of the invention. Or in the description of it. However, the method of the present invention should not be construed as reflecting the intention that the claimed invention requires more features than those specifically recited in the appended claims. Rather, as the invention is reflected by the appended claims, it is claimed that the technical problems can be solved with fewer features than all of the features of a single disclosed embodiment. Therefore, the claims following the specific embodiments are hereby explicitly incorporated into the embodiments, and each of the claims as a separate embodiment of the invention.

It will be understood by those skilled in the art that all features disclosed in the specification, including the accompanying claims, the abstract and the drawings, and all methods or devices so disclosed, may be employed in any combination, unless the features are mutually exclusive. Process or unit combination. Each feature disclosed in this specification (including the accompanying claims, the abstract and the drawings) may be replaced by alternative features that provide the same, equivalent or similar purpose.

In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features that are included in other embodiments and not in other features, combinations of features of different embodiments are intended to be within the scope of the present invention. Different embodiments are formed and formed. For example, in the claims, any one of the claimed embodiments can be used in any combination.

Various component embodiments of the present invention may be implemented in hardware or in one or more processes Implemented by software modules running on the device, or in combinations of them. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used in practice to implement some of some of the inference engine training devices or upgrade devices of the auxiliary diagnostic system in accordance with embodiments of the present invention or All features. The invention can also be implemented as a device program (e.g., a computer program and a computer program product) for performing some or all of the methods described herein. Such a program implementing the invention may be stored on a computer readable medium or may be in the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

It is to be noted that the above-described embodiments are illustrative of the invention and are not intended to be limiting, and that the invention may be devised without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as a limitation. The word "comprising" does not exclude the presence of the elements or steps that are not recited in the claims. The word "a" or "an" The invention can be implemented by means of hardware comprising several distinct elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means can be embodied by the same hardware item. The use of the words first, second, and third does not indicate any order. These words can be interpreted as names.

The above is only the specific embodiment of the present invention or the description of the specific embodiments, and the scope of the present invention is not limited thereto, and any person skilled in the art can easily within the technical scope disclosed by the present invention. Any changes or substitutions are contemplated as being within the scope of the invention. The scope of the invention should be determined by the scope of the claims.

Claims

An inference engine training method, comprising:

Step S2010: using multiple sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions;

Step S2020: selecting at least one set of blood flow data from the knowledge base;

Step S2030: For each set of blood flow data in the at least one set of blood flow data, using the inference engine to infer blood flow characteristic parameters in the blood flow data of the group to obtain a corresponding test conclusion;

Step S2040: performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in the at least one set of blood flow data;

Step S2050: Determine whether the inference engine is qualified based on the hemodynamic verification result.
The inference engine training method according to claim 1, wherein

The step S2050 includes:

If it is determined that the inference engine fails, then go to step S2060;

The inference engine training method further includes:

Step S2060: Output indication information indicating that the inference engine has not passed the hemodynamic check.
The inference engine training method according to claim 2, wherein after the step S2060, the inference engine training method further comprises:

Step S2070: Receive one of a hemodynamic model used to modify a model for implementing the inference engine, parameters of the inference engine, blood flow data stored in the knowledge base, and hemodynamic check The first modification instruction of the item or items;

Step S2080: Perform a corresponding modification action according to the first modification instruction and return to the step S2010.
The inference engine training method according to claim 1, wherein

The step S2050 includes:

If it is determined that the inference engine is qualified, then go to step S2090;

The inference engine training method further includes:

Step S2090: Update the known rules stored in the knowledge base with the rules provided by the inference engine.
The inference engine training method according to claim 1, wherein

The step S2050 includes:

If it is determined that the inference engine is qualified, then go to step S2100;

The inference engine training method further includes:

Step S2100: determining whether the rule provided by the inference engine conflicts with a known rule stored in the knowledge base, and if yes, proceeding to step S2110;

Step S2110: Output indication information indicating that the rule conflicts.
The inference engine training method according to claim 5, wherein after the step S2110, the inference engine training method further comprises:

Step S2120: receiving a second modification instruction for modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base;

Step S2130: Perform a corresponding modification action according to the second modification instruction and return to the step S2010.
The inference engine training method according to claim 5, wherein

The step S2100 includes:

If the rule provided by the inference engine does not conflict with the known rules stored in the knowledge base, then go to step S2140;

The inference engine training method further includes:

Step S2140: Update the known rules stored in the knowledge base with rules provided by the inference engine.
The inference engine training method according to claim 1, wherein after the step S2010, the inference engine training method further comprises:

Step S2012: cross-validating the inference engine with the plurality of sets of blood flow data to evaluate the correct rate of the inference engine.
The inference engine training method according to claim 8, wherein

The step S2012 includes:

If the correct rate is less than the preset threshold, then go to step S2014;

The inference engine training method further includes:

Step S2014: Output indication information indicating that the inference engine has not passed the cross-validation.
The inference engine training method according to claim 9, wherein after the step S2014, the inference engine training method further comprises:

Step S2016: receiving a third modification instruction for modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and blood flow data stored in the knowledge base;

Step S2018: Perform a corresponding modification action according to the third modification instruction and return to the step S2010.
The inference engine training method according to claim 1, wherein said inference engine is implemented using a Bayesian classifier, a support vector machine, or a deep neural network.
The inference engine training method according to claim 1, wherein said step S2040 comprises:

Establish a predefined hemodynamic model;

Determining whether blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in the at least one set of blood flow data meet physical laws prescribed by the predefined hemodynamic model to obtain the blood Flow dynamics verification results.
An upgrade method for an auxiliary diagnostic system, comprising:

At least an inference engine trained by the inference engine training method according to any one of claims 1 to 12 using the auxiliary diagnostic system to infer one or more sets of actual blood flow characteristic parameters to obtain One or more actual conclusions in which one or more sets of actual blood flow characteristic parameters correspond one-to-one;

Outputting the one or more actual conclusions;

Receiving indication information regarding at least a portion of the actual conclusions of the one or more actual conclusions;

And storing at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with the at least part of the actual conclusion based on the indication information into a knowledge base of the auxiliary diagnostic system for implementing the auxiliary diagnosis System upgrade.
The upgrade method according to claim 13, wherein the database of the auxiliary diagnostic system is configured to store an actual blood flow characteristic parameter previously processed by the auxiliary diagnostic system and to obtain an inference for the previously processed actual blood flow characteristic parameter As a previous conclusion, the one or more sets of actual blood flow characteristic parameters are previously processed actual blood flow characteristic parameters from a database of the auxiliary diagnostic system.
The upgrade method of claim 13 wherein said at least partial actual conclusion is an actual conclusion corresponding to a manually selected atypical case.
The upgrading method according to claim 13, wherein said inference engine obtained by training at least said auxiliary diagnostic system using the inference engine training method according to any one of claims 1 to 12 is one or more The reasoning of the actual blood flow characteristic parameters of the group includes:

Synthesizing the rules provided by the inference engine with known knowledge stored in the knowledge base of the auxiliary diagnostic system Rules to obtain a consolidated rule;

The one or more sets of actual blood flow characteristic parameters are reasoned based on the integrated rules.
The upgrade method of claim 13, wherein the database of the auxiliary diagnostic system is configured to store feedback information as to whether the previous conclusion was adopted, the upgrade method further comprising:

If it is determined according to the feedback information stored in the database that the number of previous conclusions of a particular type has not been adopted exceeds the threshold of times, then

Outputting prompt information regarding possible problems with the previous conclusions of the particular type;

Receiving a modification instruction for modifying one or more of a model for implementing the inference engine, parameters of the inference engine, and known rules stored in the knowledge base;

Performing a corresponding modification action according to the modification instruction.
The upgrade method of claim 17, wherein the upgrading method further comprises:

For each blood vessel in a plurality of blood vessels,

At least utilizing the inference engine to infer a set of actual blood flow characteristic parameters associated with the blood vessel to obtain a blood vessel related conclusion corresponding to the blood vessel;

Real-time output of blood vessel-related conclusions corresponding to the blood vessel;

Feedback information on whether or not the blood vessel related conclusion corresponding to the blood vessel is accepted is received and stored in real time.
The upgrade method of claim 17, wherein the upgrading method further comprises:

At least using the inference engine to infer a plurality of sets of actual blood flow characteristic parameters associated with the plurality of blood vessels in a one-to-one correspondence to obtain a plurality of blood vessel related conclusions corresponding to the plurality of blood vessels one-to-one;

Determining a general conclusion based on the plurality of blood vessel related conclusions;

Output the general conclusion; and

Feedback information is received and stored as to whether the overall conclusion is accepted.
The upgrading method according to claim 13, wherein said one or more sets of actual blood flow characteristic parameters are actual blood flow characteristic parameters currently required to be processed by said auxiliary diagnostic system,

The upgrade method further includes generating a new interpreter in the auxiliary diagnostic system based on the knowledge base and/or the inference engine.
An inference engine training device comprising:

a training module, configured to use the plurality of sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes a blood flow characteristic parameter and a corresponding known conclusion;

a selection module, configured to select at least one set of blood flow data from the knowledge base;

And a reasoning module, configured to infer, for each set of blood flow data in the at least one set of blood flow data, the blood flow characteristic parameter in the blood flow data by using the inference engine to obtain a corresponding test conclusion;

a hemodynamic verification module for performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each of the at least one set of blood flow data;

The qualification judging module is configured to judge whether the inference engine is qualified based on the hemodynamic verification result.
An upgrade device for an auxiliary diagnostic system, comprising:

a first inference module, configured to perform at least one or more sets of actual blood flow characteristic parameters by using an inference engine trained by the inference engine training method according to any one of claims 1 to 12; Inferring to obtain one or more actual conclusions that correspond one-to-one with the one or more sets of actual blood flow characteristic parameters;

a first output module, configured to output the one or more actual conclusions;

a first receiving module, configured to receive indication information about at least part of the actual conclusions of the one or more actual conclusions;

a first storage module, configured to store, according to the indication information, the at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters that are in one-to-one correspondence with the at least part of the actual conclusion into the knowledge base of the auxiliary diagnosis system Used to implement an upgrade of the auxiliary diagnostic system.
An inference engine training device comprising:

Memory for storing programs;

a processor for running the program;

Wherein, when the program runs in the processor, it is used to perform the following steps:

Step S2010: using multiple sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions;

Step S2020: selecting at least one set of blood flow data from the knowledge base;

Step S2030: For each set of blood flow data in the at least one set of blood flow data, using the inference engine to infer blood flow characteristic parameters in the blood flow data of the group to obtain a corresponding test conclusion;

Step S2040: performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in the at least one set of blood flow data;

Step S2050: Determine whether the inference engine is qualified based on the hemodynamic verification result.
An upgrade device for an auxiliary diagnostic system, comprising:

Memory for storing programs;

a processor for running the program;

Wherein, when the program runs in the processor, it is used to perform the following steps:

At least an inference engine trained by the inference engine training method according to any one of claims 1 to 12 using the auxiliary diagnostic system to infer one or more sets of actual blood flow characteristic parameters to obtain One or more actual conclusions in which one or more sets of actual blood flow characteristic parameters correspond one-to-one;

Outputting the one or more actual conclusions;

Receiving indication information regarding at least a portion of the actual conclusions of the one or more actual conclusions;

And storing at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with the at least part of the actual conclusion based on the indication information into a knowledge base of the auxiliary diagnostic system for implementing the auxiliary diagnosis System upgrade.
A computer readable storage medium having stored thereon a program for performing the following steps at runtime:

Step S2010: using multiple sets of blood flow data stored in the knowledge base as a sample training inference engine, wherein each set of blood flow data stored in the knowledge base includes blood flow characteristic parameters and corresponding known conclusions;

Step S2020: selecting at least one set of blood flow data from the knowledge base;

Step S2030: For each set of blood flow data in the at least one set of blood flow data, using the inference engine to infer blood flow characteristic parameters in the blood flow data of the group to obtain a corresponding test conclusion;

Step S2040: performing hemodynamic verification based on blood flow characteristic parameters and corresponding test conclusions in each set of blood flow data in the at least one set of blood flow data;

Step S2050: Determine whether the inference engine is qualified based on the hemodynamic verification result.
A computer readable storage medium having stored thereon a program for performing the following steps at runtime:

At least an inference engine trained by the inference engine training method according to any one of claims 1 to 12 using the auxiliary diagnostic system to infer one or more sets of actual blood flow characteristic parameters to obtain One or more actual conclusions in which one or more sets of actual blood flow characteristic parameters correspond one-to-one;

Outputting the one or more actual conclusions;

Receiving indication information regarding at least a portion of the actual conclusions of the one or more actual conclusions;

And storing at least part of the actual conclusion and at least one set of actual blood flow characteristic parameters in one-to-one correspondence with the at least part of the actual conclusion based on the indication information into a knowledge base of the auxiliary diagnostic system for implementing the auxiliary diagnosis System upgrade.