CN115019940A

CN115019940A - Prediction method and device of digestive tract diseases based on eye images and electronic equipment

Info

Publication number: CN115019940A
Application number: CN202210784197.7A
Authority: CN
Inventors: 孙旭芳; 谭宇和
Original assignee: Tongji Hospital Affiliated To Tongji Medical College Of Huazhong University Of Science & Technology
Current assignee: Tongji Hospital Affiliated To Tongji Medical College Of Huazhong University Of Science & Technology
Priority date: 2022-06-28
Filing date: 2022-06-28
Publication date: 2022-09-06

Abstract

The invention discloses a method for predicting digestive tract diseases based on eye images, which comprises the steps of obtaining eye images of target persons and target inspection data of the target persons; performing feature extraction on the eye image to obtain target eye features; performing feature extraction on the target inspection data to obtain target inspection features; and inputting the target eye features and the target examination features into a trained digestive tract recognition model to obtain a target disease probability of the digestive tract disease of the target person, wherein the digestive tract recognition model is obtained by training according to a training sample set, and each training sample in the training sample set comprises a historical eye image, historical examination data and historical digestive tract diagnosis data. The method, the device and the electronic equipment for predicting the digestive tract diseases based on the eye images can effectively reduce the detection cost of the digestive tract diseases and effectively improve the detection efficiency of the digestive tract diseases.

Description

Prediction method and device of digestive tract diseases based on eye images and electronic equipment

Technical Field

The invention relates to the technical field of medical treatment, in particular to a method and a device for predicting digestive tract diseases based on eye images and electronic equipment.

Background

In the prior art, because digestive tract diseases are high, the incidence rate and the mortality rate of cancers of stomach cancer and colorectal cancer in China are always high, and digestive tract diseases are screened by using a digestive endoscope, but the conventional mode of using the digestive endoscope for screening the digestive tract diseases is influenced by the problems of high cost, long detection time, scarce talents of endoscopists, insufficient supply of endoscope equipment and the like of using the digestive endoscope for detecting the digestive tract diseases. Therefore, a method for predicting digestive tract diseases with low detection cost and high detection speed is needed.

Disclosure of Invention

The embodiment of the invention provides a method and a device for predicting a digestive tract disease based on an eye image and an electronic device, which can effectively reduce the detection cost of the digestive tract disease and effectively improve the detection efficiency of the digestive tract disease.

A first aspect of an embodiment of the present invention provides a method for predicting a disease of an alimentary canal based on an eye image, the method including:

acquiring an eye image of a target person and target inspection data of the target person;

performing feature extraction on the eye image to obtain target eye features;

performing feature extraction on the target inspection data to obtain target inspection features;

and inputting the target eye features and the target examination features into a trained digestive tract recognition model to obtain a target disease probability of the digestive tract disease of the target person, wherein the digestive tract recognition model is obtained by training according to a training sample set, and each training sample in the training sample set comprises a historical eye image, historical examination data and historical digestive tract diagnosis data.

Optionally, the inputting the target eye feature and the target examination feature into a trained digestive tract recognition model to obtain a target prevalence probability of a digestive tract disease of the target person includes:

performing feature fusion on the target eye feature and the target inspection feature to obtain a target fusion feature;

and inputting the target fusion characteristics into the digestive tract recognition model to obtain the target disease probability.

Optionally, if the target inspection data includes target inspection item data and target description item data of the target person, performing feature extraction on the target inspection data to obtain a target inspection feature includes:

respectively extracting the features of the target examination item data and the target description item data to obtain target examination item features and target description item features;

and obtaining the target inspection feature according to the target inspection item feature and the target description item feature.

Optionally, the training step of the digestive tract recognition model includes:

acquiring the training sample set;

acquiring training eye features, training examination features and digestive tract diagnosis features of each training sample in the training sample set, wherein the digestive tract diagnosis features are obtained by performing feature extraction on historical digestive tract diagnosis data in the historical digestive tract diagnosis data set;

and taking the training eye characteristics and the training inspection characteristics of each training sample as input data of the model, and taking the digestive tract diagnosis characteristics of each training sample as output data of the model to carry out model training, so as to obtain a trained model as the digestive tract recognition model.

Optionally, if the historical inspection data in the historical inspection data set includes historical inspection item data and historical description item data, the obtaining the training inspection feature of each training sample in the training sample set includes:

for each training sample, obtaining historical examination item data and historical description item data of the training sample from the historical examination data set; carrying out normalization and whitening processing on historical examination item data of the training sample to obtain training examination item characteristics of the training sample; carrying out classification marking on the historical description item data of the training sample to obtain the training description item characteristics of the training sample; and obtaining the training examination characteristics of the training sample according to the training examination item characteristics and the training description item characteristics of the training sample.

Optionally, the using the training eye features and the training examination features of each training sample as input data of the model includes:

performing feature fusion on training inspection features and training eye features of the training samples to obtain training fusion features of the training samples;

and taking the training fusion characteristics of each training sample as input data of the model.

Optionally, after obtaining the target prevalence probability of the digestive tract disease of the target human, the method further comprises:

and acquiring and outputting the digestive tract diagnosis data of the target person according to the target disease probability.

The second aspect of the embodiments of the present invention also provides an apparatus for predicting a disease of an alimentary tract based on an eye image, the apparatus including:

a data acquisition unit configured to acquire an eye image of a target person and target inspection data of the target person;

the feature extraction unit is used for extracting features of the eye image to obtain target eye features; performing feature extraction on the target inspection data to obtain target inspection features;

and the model identification unit is used for inputting the target eye characteristics and the target examination characteristics into a trained digestive tract identification model to obtain the target disease probability of the digestive tract disease of the target person, wherein the digestive tract identification model is obtained by training according to a training sample set, and each training sample in the training sample set comprises a historical eye image, historical examination data and historical digestive tract diagnosis data.

Optionally, the model identifying unit is configured to perform feature fusion on the target eye feature and the target inspection feature to obtain a target fusion feature; and inputting the target fusion characteristics into the digestive tract recognition model to obtain the target disease probability. .

A third aspect of the embodiments of the present invention provides an electronic device, including a memory and one or more programs, where the one or more programs are stored in the memory and configured to be executed by one or more processors to execute operating instructions included in the one or more programs for performing the method for predicting an eye-based digestive tract disease according to the first aspect.

A fourth aspect of the embodiments of the present invention provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps corresponding to the method for predicting a gastrointestinal disease based on an eye image as provided in the first aspect.

The above one or at least one technical solution in the embodiments of the present application has at least the following technical effects:

based on the technical scheme, the eye image of the target person and the inspection data of the target person are obtained; carrying out feature extraction on the eye image and the inspection data to obtain target eye features and target inspection features; and inputting the target eye characteristics and the target inspection characteristics into a trained digestive tract recognition model to obtain the target illness probability of the digestive tract diseases of the target person, so that after eye images and inspection data are obtained, the target illness probability of the target person can be recognized through the trained digestive tract recognition model without using a digestive endoscope for detecting the digestive tract diseases, thereby effectively reducing the detection cost of the digestive tract diseases and effectively improving the detection efficiency of the digestive tract diseases.

And because the digestive tract identification model is obtained by training according to the training sample set, and each training sample in the training sample set comprises the historical eye image, the historical examination data and the historical digestive tract diagnosis data, the training sample is consistent with the actually predicted data, so that the accuracy of the digestive tract identification model is higher, and the accuracy of the acquired target disease probability can be effectively improved.

Drawings

Fig. 1 is a schematic flowchart of a method for predicting a gastrointestinal disease based on an eye image according to an embodiment of the present disclosure;

fig. 2 is a block diagram of an apparatus for predicting a disease of an alimentary tract based on an eye image according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The main implementation principle, the specific implementation mode and the corresponding beneficial effects of the technical solutions of the embodiments of the present application are described in detail below with reference to the accompanying drawings.

The digestive tract diseases in the embodiments of the present specification may refer to eight types of digestive tract diseases, i.e., gastritis, gastric polyp, gastric cancer, colorectal polyp, colorectal inflammation, ulcerative colitis, and colorectal cancer, and of course, the digestive tract diseases may include digestive tract diseases other than the eight types of diseases described above, and the present specification is not particularly limited.

In the embodiment of the present specification, the eye images include fundus photographic images and slit-lamp photographic images, and of course, only fundus photographic images or slit-lamp photographic images may be included, and the present specification is not particularly limited.

Examples

Referring to fig. 1, an embodiment of the present application provides a method for predicting a disease of a digestive tract based on an eye image, the method including:

s101, acquiring an eye image of a target person and inspection data of the target person;

s102, extracting the features of the eye image to obtain the features of the target eye;

s103, extracting the features of the target inspection data to obtain target inspection features;

s104, inputting the target eye feature and the target inspection feature into a trained digestive tract recognition model to obtain a target illness probability of the digestive tract disease of the target person, wherein the digestive tract recognition model is obtained by training according to a training sample set, and each training sample in the training sample set comprises a historical eye image, historical inspection data and historical digestive tract diagnosis data.

The method for predicting a gastrointestinal disease based on an eye image in an embodiment of the present disclosure is generally applied to a server, where the server may be a cloud server and a local server, and the server may be an electronic device such as a notebook computer, a desktop computer, a tablet computer, and an all-in-one machine.

In step S101, after the eye image of the target person is checked by the eye image checking device, the eye image checking device transmits the eye image to the server, so that the server can receive the eye image of the target person, and the server can acquire the eye image; accordingly, after the target examination data of the target person is examined by the corresponding medical examination apparatus, the medical examination apparatus may transmit the examined target examination data of the target person to the server, thereby enabling the server to acquire the target examination data of the target person.

In the embodiment of the present specification, both the eye portion image examination apparatus and the medical examination apparatus are in communication connection with the server, and the eye portion image examination apparatus may be, for example, a digital fundus oculi angiography examination apparatus, a direct ophthalmoscope, an indirect ophthalmoscope, or the like.

Specifically, the target examination data may include target examination item data of the target person, and may further include target description item data of the target person, and the target examination item data may include eye examination data of the target person, including eye-near, glaucoma, refractive index, and the like, and general examination data including, for example, blood sugar, blood pressure, and the like, and target description item data including data of sex, age, height, weight, and presence or absence of complications, and types of complications, such as keratopathy, cataract, glaucoma, diabetic retinopathy, age-related macular degeneration, high-myopia retinopathy, and the like, and extraocular complications (diabetes, hypertension, coronary heart disease, chronic kidney disease, chronic liver disease, and the like). In the following, it is specifically exemplified that the object check data includes object check item data and object description item data.

In the embodiment of the present specification, the eye images include a fundus image and a slit lamp image, wherein the fundus image may be divided into regions such as an optic disc, a macula lutea, and upper and lower retinal vascular arches, and the slit lamp image may be divided into regions such as upper and lower eyelids, sclera, cornea, iris, and pupil.

After the target examination data and the target fundus data are acquired, step S102 is executed.

In step S102, Feature extraction may be performed on the target eye image through a convolutional neural network, or Feature extraction may be performed using a Scale-Invariant Feature Transform (SIFT) algorithm, an orb Feature extraction algorithm, and an accelerated Up Robust Features (SURF) algorithm, so as to obtain a target eye Feature.

After the target eye feature is acquired, step S103 is performed. Step S103 may be executed before or simultaneously with step S102, and this specification is not particularly limited.

Specifically, taking the example that the target inspection data includes target inspection item data and target description item data, after the target inspection item data and the target description item data are acquired, feature extraction may be performed on the target inspection item data and the target description item data, respectively, to obtain a target inspection item feature and a target description item feature; and obtaining the target inspection characteristic according to the target inspection item characteristic and the target description item characteristic. In this case, feature extraction may be performed using a convolutional neural network.

Specifically, when feature extraction is performed on target inspection item data, an inspection vector obtained after normalization and whitening processing of the target inspection item data can be used as a target inspection item feature; when the feature extraction is performed on the target description item data, for each description item data in the target description item data, if the description item data is binary variable data, obtaining a mark for describing the item data from a preset binary mark, where the binary mark may be divided into 1 and 0, or 2 or 1, or 2 and 0, and for example, taking the binary variable data that the description data is gender as an example, a male mark is 1 and a female mark is 0 in the binary mark for gender; while other binary variables may be labeled, for example, with a presence flag of 1 and an absence flag of 0, etc.; if the description item data is level variable data, the flag corresponding to the description item data is acquired according to the set level variable flag value, and the level variable flag value may be determined according to the level of the variable data, for example, the flag may be set from a low level to a high level, and the interval may be 1 or 2, and the like.

In this way, after the above steps are performed for each description item data in the target description item data, the label of each description item data is acquired, and vectors of multiple dimensions are formed as target description item features according to the label of each description item data.

And after the target inspection item feature and the target description item feature are obtained, performing feature fusion on the target inspection item feature and the target description item feature, and taking the obtained fusion feature as the target inspection feature. Of course, the target inspection item feature and the target description item feature may also be directly used as the target inspection feature, and this specification is not particularly limited.

After the target eye feature and the target examination feature are acquired through steps S102 and S103, step S104 is executed.

Before step S104, a digestive tract recognition model is obtained by pre-training, wherein the training step of the digestive tract recognition model includes:

a1, obtaining a training sample set;

when the training sample set is obtained, clinical data of patients who are subjected to digestive endoscopy examination can be obtained from a corresponding hospital system, patients meeting an admission standard are screened from the obtained clinical data of all the patients, and after informed consent of the patients is obtained, a historical eye image of each informed patient, historical examination data of each informed patient and historical digestive endoscopy diagnosis data are obtained; after the eye image of each informed patient is acquired, images with unqualified image quality are deleted from the eye images of all the informed patients, and at the moment, the remaining data such as the eye image of each informed patient, digestive endoscopy diagnosis data and historical examination data are used as training samples and are added into a training sample set.

The historical examination data may include historical examination item data of an informed patient, and may also include historical description item data of an informed patient, where the historical examination data is consistent with the target examination data, for example, if the historical examination data includes the historical description item data, the target examination data also includes the target description item data, and if the historical examination data does not include the historical description item data, the target examination data also does not include the target description item data.

In the embodiment of the present specification, the selection criterion may be set according to actual requirements, or may be set manually or by a device, and the selection criterion may be, for example, the age is greater than or equal to 18 years; the eye position is normal, the central fixation point is good, and fundus photography examination can be performed; and more than 90% of the eye image area includes four main areas (optic disc, macula lutea, upper and lower retinal vascular arches), and the like. Further, the unqualified image quality can be set according to actual requirements, or can be set by people or equipment, and the unqualified image quality can be eyelid shielding, light leakage (> 10% area), lens flare or stain, overexposure and the like.

And after the training sample set is obtained, marking the eye images of the training samples by using digestive endoscopy diagnostic data corresponding to the training samples according to each training sample, and marking the eye images of the training samples as digestive tract disease marks or digestive tract disease marks. Furthermore, the eye images marked with the digestive tract diseases can be marked, eight types of digestive tract diseases in the digestive tract diseases can be used for marking the eye images, and one eye image marked with the digestive tract diseases can be marked with a plurality of digestive tract diseases. For example, if a patient has gastritis, colorectal polyps, colorectal inflammation, and ulcerative colitis corresponding to an eye image marked with a gastrointestinal disease, the patient has 4 gastrointestinal diseases, namely gastritis, colorectal polyps, colorectal inflammation, and ulcerative colitis.

After the training sample set is obtained, the training sample set may be further grouped into a training set and an internal test set and an external test set, each of the 3 groups includes a part of the training samples in the training sample set, of course, the number of the training samples in the training set and the internal test set and the number of the training samples in the external test set may be selected according to a set proportion, and the set proportion corresponding to the training set and the internal test set and the external test set may be, for example, 8: 1: 1,7: 2: 1 and 8: 2: 2, etc. Thus, model training can be performed through the training set, and the model is verified in the internal test set and the external test set.

A2, obtaining training eye characteristics, training inspection characteristics and digestive tract diagnosis characteristics of each training sample in a training sample set, wherein the digestive tract diagnosis characteristics are obtained by performing characteristic extraction on historical digestive tract diagnosis data in a historical digestive tract diagnosis data set;

specifically, for each training sample in a training sample set, feature extraction is carried out on historical eye images of the training samples to obtain training eye features of the training samples; extracting the characteristics of the historical inspection data of the training sample to obtain the training inspection characteristics of the training sample; and extracting the characteristics of the historical digestive tract diagnostic data of the training sample to obtain the digestive tract diagnostic characteristics of the training sample. The above features may be extracted by a convolutional neural network, or may be extracted by other feature extraction methods, and this specification is not limited specifically.

Specifically, when the historical examination data of the training sample comprises the historical examination item data and the historical description item data of the training sample, the training examination feature is made to comprise a training examination item feature and a training description item feature. Of course, the historical examination data of the training sample may also contain only the historical examination item data. The following takes as an example that the history check data includes history check item data and history description item data.

In this way, if the historical check data includes the historical check item data and the historical description item data, when the training check feature of each training sample is obtained, the historical check item data and the historical description item data of the training sample are obtained from the historical check data set for each training sample; carrying out normalization and whitening processing on historical examination item data of the training sample to obtain training examination item characteristics of the training sample; and carrying out classified marking on the historical description item data of the training sample to obtain the training description item characteristics of the training sample. And after the training description item features and the training check item features of the training samples are obtained, the training check features of the training samples are obtained according to the training description item features and the training check item features of the training samples.

And when the training check features of the training samples are obtained according to the training description item features and the training check item features of the training samples, performing feature fusion on the training check item features and the training description item features of the training samples, and taking the obtained check fusion features as the training check features of the training samples. Of course, the training examination item features and the training description item features of the training sample may also be directly used as the training examination features of the training sample, and this specification is not limited in particular.

In this embodiment of the present specification, reference may be made to the descriptions of steps S102 and S103 for specific obtaining manners of training examination item features, training description item features, and training eye features of a training sample, and for brevity of the specification, details are not repeated here.

And A3, taking the training eye characteristics and the training inspection characteristics of each training sample as input data of the model, and taking the digestive tract diagnosis characteristics of each training sample as output data of the model to carry out model training, so as to obtain a trained model as a digestive tract recognition model.

Specifically, for each training sample, the training inspection features and the training eye features of the training sample may be subjected to feature fusion to obtain training fusion features of the training sample, and the training fusion features of each training sample may be used as input data of the model. And taking the digestive tract diagnosis characteristics of each training sample as output data of the model, and performing model training by inputting the input data and the output data of the model to obtain a trained model as a digestive tract recognition model.

In the practical application process, the digestive tract identification model is specifically realized as follows: aiming at each training sample, firstly, training a pre-constructed stacked self-encoder by using a layer-by-layer greedy training method by using training fusion characteristics of the training samples, taking the output of a hidden layer of the last layer of self-encoder of the trained stacked self-encoder as the input of a pre-selected classifier, taking digestive tract diagnosis characteristics of the training samples as the output of the classifier, training the classifier, and obtaining the trained classifier as a digestive tract recognition model.

Specifically, the trained classifier may be evaluated by using an ROC curve, for example, the performance of the classifier on the test data set may be evaluated by using an area under an operating characteristic curve of a subject (AUROC), sensitivity, specificity, and F1 score (harmonic mean of sensitivity and specificity), and expressed by 95% CI, and when the performance of the classifier on the test data set is evaluated to reach a set performance, the classifier corresponding to the set performance is used as the digestive tract recognition model.

In practical applications, each training sample contains data specifically as described in table 1 below, such that each training sample contains historical eye images, historical examination data, and historical digestive tract diagnostic data.

Training sample ID:

TABLE 1

In this way, for each training sample, the data described in table 1 corresponding to the training sample can be acquired, and thereby the historical eye image, the historical examination data, and the historical gastrointestinal diagnosis data of the training sample can be acquired.

After the digestive tract recognition model is obtained through the training of the steps A1-A3 and the target eye feature and the target examination feature are obtained through the steps S102 and S103, feature fusion can be firstly carried out on the target eye feature and the target examination feature to obtain a target fusion feature; and inputting the target fusion characteristics into a digestive tract recognition model to obtain the target disease probability.

For example, taking user B as an example, feature extraction is performed on an eye image of B, and a target eye feature is obtained and represented by B1; b, extracting the characteristics of the inspection data of B to obtain target inspection characteristics which are represented by B2; then performing feature fusion on B1 and B2 to obtain a target fusion feature represented by B12; b12 is input into the pre-trained digestive tract recognition model, and the target prevalence probability of the model output data as B is, for example, 80%.

In another embodiment, after obtaining the target prevalence probability of the gastrointestinal disease of the target person, the gastrointestinal diagnostic data of the target person may be obtained and output according to the target prevalence probability.

Specifically, a set value may be preset, and the set value may be set according to actual requirements, or may be set manually or by the device, and the set value is usually any value between 50% and 100%, for example, 55% and 60%; thus, after the target disease probability is obtained, the target disease probability is compared with a set value, and digestive tract diagnosis data of the target person is obtained and output according to the comparison result.

Specifically, if the comparison result represents that the target disease probability is greater than a set value, determining that the target person has the digestive tract disease, recording data with the digestive tract disease into digestive tract diagnosis data of the target person and outputting the data; and if the comparison result represents that the target disease probability is not larger than the set value, determining that the target person does not suffer from the digestive tract disease, recording the data without the digestive tract disease into the digestive tract diagnosis data of the target person and outputting the data, so that the target person can intuitively know whether the target person suffers from the digestive tract diagnosis data.

Because the digestive tract recognition model is trained in advance, after the eye image and the target inspection data of the target person are obtained, the feature extraction can be automatically carried out on the eye image and the target inspection data through a machine, and the feature extracted by the feature is input into the digestive tract recognition model, so that the target illness probability of the target person is obtained, and the digestive tract disease detection is carried out without manually using a digestive endoscope, so that the detection cost of the digestive tract disease is effectively reduced, and compared with manual detection, the automatic detection of the machine can effectively improve the detection efficiency of the digestive tract disease.

In another embodiment, a stomach disease identification model and a bowel disease identification model may be trained in advance, wherein the training step of the stomach disease identification model may refer to the training steps a1-A3, which are different from the digestive tract diagnosis features in the step a2, specifically, when the feature extraction is performed on the historical digestive tract diagnosis data of the training sample, the feature extraction is performed on the stomach disease diagnosis data in the historical digestive tract diagnosis data to obtain stomach disease diagnosis features; in this way, in the training step of the stomach illness recognition model, the training eye feature, the training examination feature and the stomach illness diagnosis feature of each training sample in the training sample set are obtained in step a2, and then the training eye feature and the training examination feature of each training sample are used as the input data of the model, and the stomach illness diagnosis feature of each training sample is used as the output data of the model to perform model training, so that the trained model is obtained as the stomach illness recognition model.

Accordingly, the training step of the bowel disease recognition model may refer to the above-mentioned training steps a1-A3, which are different from the above-mentioned digestive tract diagnostic features in step a2, specifically, when feature extraction is performed on the historical digestive tract diagnostic data of the training sample, feature extraction is performed on the intestinal disease diagnostic data in the historical digestive tract diagnostic data to obtain an intestinal disease diagnostic feature; in this way, in the training step of the intestinal disease recognition model, the training eye feature, the training examination feature and the intestinal disease diagnosis feature of each training sample in the training sample set are obtained in step a2, and then the training eye feature and the training examination feature of each training sample are used as the input data of the model, and the intestinal disease diagnosis feature of each training sample is used as the output data of the model to perform model training, so that the trained model is obtained as the intestinal disease recognition model.

In the practical application process, taking a digestive tract recognition model as a model1, a stomach disease recognition model as a model2 and a bowel disease recognition model as a model3 as examples, firstly, an eye image and target examination data of a target person are obtained, after feature extraction is carried out on the eye image and the target examination data, target eye features and target examination features obtained by feature extraction are input into the model1, and if the target person is recognized to have a digestive tract disease; inputting the target eye characteristic and the target examination characteristic into the model2 and the model3 respectively to obtain the specific stomach illness prevalence probability of the target person output by the model2 and the specific intestinal illness prevalence probability of the target person output by the model 3; according to the prevalence probability of the stomach illness and the prevalence probability of the intestinal diseases, the target person is determined to be suffered from certain intestinal diseases and/or certain stomach illness.

In this way, after the target person is identified to have the gastrointestinal disease by the gastrointestinal disease identification model, whether the target person has the gastrointestinal disease or the gastrointestinal disease can be further identified by the gastrointestinal disease identification model and the intestinal disease identification model, and a specific disease diagnosis report can be given.

In view of the foregoing, an embodiment of the present invention provides a method for predicting a gastrointestinal disease based on an eye image, and an embodiment of the present invention further provides a device for predicting a gastrointestinal disease based on an eye image, please refer to fig. 2, where the device includes:

a data acquisition unit 201 for acquiring an eye image of a target person and target examination data of the target person;

a feature extraction unit 202, configured to perform feature extraction on the eye image to obtain a target eye feature; performing feature extraction on the target inspection data to obtain target inspection features;

a model identification unit 203, configured to input the target eye feature and the target examination feature into a trained digestive tract identification model, so as to obtain a target prevalence probability of a digestive tract disease of the target person, where the digestive tract identification model is obtained by training according to a training sample set, and each training sample in the training sample set includes a historical eye image, historical examination data, and historical digestive tract diagnosis data.

In an optional implementation manner, the model identifying unit 203 is configured to perform feature fusion on the target eye feature and the target examination feature to obtain a target fusion feature; and inputting the target fusion characteristics into the digestive tract recognition model to obtain the target disease probability.

In an optional implementation manner, the feature extraction unit 202 is configured to, if the target inspection data includes target inspection item data and target description item data of the target person, perform feature extraction on the target inspection item data and the target description item data respectively to obtain a target inspection item feature and a target description item feature; and obtaining the target inspection feature according to the target inspection item feature and the target description item feature.

In an alternative embodiment, the method further comprises:

the model training unit is used for acquiring the training sample set; acquiring training eye features, training examination features and digestive tract diagnosis features of each training sample in the training sample set, wherein the digestive tract diagnosis features are obtained by performing feature extraction on historical digestive tract diagnosis data in the historical digestive tract diagnosis data set; and taking the training eye characteristics and the training inspection characteristics of each training sample as input data of the model, and taking the digestive tract diagnosis characteristics of each training sample as output data of the model to carry out model training, so as to obtain a trained model as the digestive tract recognition model.

In an optional implementation manner, the model training unit is configured to, for each training sample, obtain historical examination item data and historical description item data of the training sample from the historical examination data set if the historical examination data in the historical examination data set includes the historical examination item data and the historical description item data; carrying out normalization and whitening processing on historical examination item data of the training sample to obtain training examination item characteristics of the training sample; carrying out classification marking on the historical description item data of the training sample to obtain the training description item characteristics of the training sample; and obtaining the training examination characteristics of the training sample according to the training examination item characteristics and the training description item characteristics of the training sample.

In an optional implementation manner, the model training unit is configured to perform feature fusion on training inspection features and training eye features of training samples to obtain training fusion features of the training samples, for each training sample; and taking the training fusion characteristics of each training sample as input data of the model.

In an alternative embodiment, the method further comprises:

and the diagnosis output unit is used for acquiring and outputting the digestive tract diagnosis data of the target person according to the target affection probability after obtaining the target affection probability of the digestive tract diseases of the target person.

With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.

Fig. 3 is a block diagram illustrating an electronic device 800 for a method for predicting a disease of an alimentary tract based on an eye image according to an exemplary embodiment. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, an exercise device, a personal digital assistant, and the like.

Referring to fig. 3, electronic device 800 may include one or more of the following components: processing component 802, memory 804, power component 806, multimedia component 808, audio component 810, input/presentation (I/O) interface 812, sensor component 814, and communication component 816.

The processing component 802 generally controls overall operation of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing elements 802 may include one or more processors 820 to execute instructions to perform all or a portion of the steps of the methods described above. Further, the processing component 802 can include one or more modules that facilitate interaction between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

The memory 804 is configured to store various types of data to support operation at the device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 804 may be implemented by any type or combination of volatile or non-volatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

The power supply component 806 provides power to the various components of the electronic device 800. The power components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the electronic device 800.

The multimedia component 808 includes a screen that provides a presentation interface between the electronic device 800 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front facing camera and/or a rear facing camera. The front-facing camera and/or the rear-facing camera may receive external multimedia data when the device 800 is in an operating mode, such as a shooting mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

The audio component 810 is configured to present and/or input audio signals. For example, the audio component 810 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may further be stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for presenting audio signals.

The I/O interface 812 provides an interface between the processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.

The sensor assembly 814 includes one or more sensors for providing various aspects of state assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open/closed state of the device 800, the relative positioning of components, such as a display and keypad of the electronic device 800, the sensor assembly 814 can also detect a change in position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, orientation or acceleration/deceleration of the electronic device 800, and a change in temperature of the electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast associated information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communications component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the electronic device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described methods.

In an exemplary embodiment, a non-transitory computer-readable storage medium comprising instructions, such as the memory 804 comprising instructions, executable by the processor 820 of the electronic device 800 to perform the above-described method is also provided. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.

It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is only limited by the appended claims

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A method for predicting a disease of a digestive tract based on an eye image, the method comprising:

performing feature extraction on the eye image to obtain target eye features;

2. The prediction method of claim 1, wherein the inputting the target eye feature and the target examination feature into a trained digestive tract recognition model to obtain the target prevalence probability of the digestive tract disease of the target human comprises:

3. The prediction method of claim 2, wherein if the object inspection data includes object inspection item data and object description item data of the object person, the performing feature extraction on the object inspection data to obtain an object inspection feature comprises:

4. The prediction method of claim 1, wherein the step of training the digestive tract recognition model comprises:

acquiring the training sample set;

5. The prediction method of claim 4, wherein if the historical inspection data in the historical inspection data set includes historical inspection item data and historical description item data, the obtaining the training inspection feature of each training sample in the training sample set comprises:

6. The prediction method of claim 5, wherein the using the training eye features and the training examination features of each training sample as input data for the model comprises:

7. The prediction method of claim 1, wherein after obtaining the target prevalence probability of the target human for the digestive tract disease, the method further comprises:

8. An apparatus for predicting a disease of a digestive tract based on an eye image, the apparatus comprising:

9. An electronic device comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors to execute operating instructions included in the one or more programs for performing the corresponding method according to any one of claims 1 to 6.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps corresponding to the method according to any one of claims 1 to 6.