WO2010050333A1

WO2010050333A1 - Information processing device

Info

Publication number: WO2010050333A1
Application number: PCT/JP2009/067205
Authority: WO
Inventors: 大介梶
Original assignee: コニカミノルタエムジー株式会社
Priority date: 2008-10-30
Filing date: 2009-10-02
Publication date: 2010-05-06
Also published as: JPWO2010050333A1

Abstract

It is possible to calculate the certainty factor of an identification result obtained by an identification device. An information processing device includes an identification device and a control means. The identification device outputs as an estimated value, a probability that an image belongs to each of a plurality of classes in accordance with a characteristic amount obtained by analyzing the image and identifies the class to which the image belongs according to the estimated value. The control means uses the estimated value of each of the classes outputted from the identification device so as to calculate the certainty factor of the identification result obtained by the identification device.

Description

Information processing device

The present invention relates to an information processing apparatus.

Generally, machine learning techniques are used to support user decision making. For example, when image processing is performed on an X-ray image, since there are appropriate types and parameters of image processing depending on the imaging region and imaging direction in which X-ray imaging has been performed, the imaging region and imaging direction of the X-ray image are first specified. There is a need. Accordingly, it is possible to create a discriminator by a machine learning method, discriminate the radiographed region and radiographing direction of the X-ray image by the discriminator, and select image processing according to the discriminated radiographing region and / or imaging direction. (For example, refer to Patent Document 1).

Machine learning methods include support vector machines and neural networks, as well as boosting methods such as AdaBoost and AdaBoostMlt.

JP 2008-11900 A

判別 Discriminating with a classifier using machine learning does not always give a correct answer. For example, an X-ray image in which the chest is imaged and an X-ray image in which the thoracic vertebra is imaged are similar in image characteristics, and may be misidentified. Even in such a case, if the imaging region is forcibly discriminated and image processing specialized for the imaging region is selected, improper image processing will be performed in the case of erroneous determination. In a scene where discrimination is difficult, it is preferable to select general-purpose image processing that does not rely on forcible discrimination and does not depend on the imaging region. However, since only the discrimination result by the discriminator is provided to the user, the user cannot judge whether or not the discrimination result can be trusted. *

An object of the present invention is to calculate a certainty factor for the discrimination result of the discriminator.

According to the invention of claim 1,
From the feature amount obtained by analyzing the image, a probability that the image belongs to each class among a plurality of classes is output as an estimated value, and a discriminator that determines the class to which the image belongs based on the estimated value;
Using the estimated value of each class output by the discriminator, calculating a certainty factor for the discrimination result by the discriminator, and displaying the calculated confidence factor on a display unit,
An information processing apparatus is provided.

According to invention of Claim 2,
The control means obtains a difference between the maximum estimated value and the estimated value other than the estimated value of each class output by the discriminator, and there are a predetermined number or more classes having the difference within a threshold value. The information processing apparatus according to claim 1, which outputs a binary certainty factor depending on whether or not.

According to invention of Claim 3,
The said control means outputs the absolute value of the difference of the largest estimated value of the estimated value of each class output by the said discriminator, and the 2nd largest estimated value as a certainty factor. An information processing apparatus is provided.

According to invention of Claim 4,
The control means obtains an absolute value of a difference between a maximum estimated value of the estimated values of each class output from the classifier and a second largest estimated value, and the absolute value of the difference is smaller than a threshold value. The information processing apparatus according to claim 1, which outputs a binary certainty factor depending on whether or not.

According to the invention of claim 5,
2. The information processing apparatus according to claim 1, wherein the control unit outputs a binary certainty factor depending on whether or not a maximum estimated value is equal to or less than a threshold value among the estimated values of each class output by the classifier. Is provided.

According to the invention of claim 6,
The control unit according to any one of claims 1 to 5, wherein the control unit normalizes the estimated value of each class output by the classifier and calculates the certainty factor using the normalized estimated value. An information processing apparatus is provided.

According to the invention of claim 7,
The control means determines whether the certainty factor is high or low based on the calculated certainty factor value, and when it is determined that the certainty factor is low, warns that the certainty factor is low together with the determination result of the class. The information processing apparatus according to any one of claims 1 to 6, wherein warning information to be displayed is displayed on the display means.

According to the invention described in claim 8,
The image is a medical image in which a patient is photographed by an image generation device,
The control means determines whether the calculated certainty factor is high or low,
When it is determined by the control means that the certainty level is high, image processing specialized for the class having the maximum estimated value is performed on the medical image, and when the certainty level is determined to be low, the medical image is The information processing apparatus according to any one of claims 1 to 7, further comprising image processing means for performing general-purpose image processing.

According to the invention of claim 9,
The information processing apparatus according to claim 8, wherein the class of the medical image is a class for each type of the image generation apparatus in which the medical image is captured.

According to the invention of claim 10,
The information processing apparatus according to claim 8, wherein the class of the medical image is any one of a class for each imaging region where the medical image is captured, a class for each imaging direction, or a class for each combination of the imaging region and the imaging direction. Provided.

According to the invention of claim 11,
The discriminator is configured by the AdaBoostMlt method, performs discrimination of at least three classes,
The information processing apparatus according to any one of claims 1 to 10, wherein the estimated value output by the discriminator is given together with a value corresponding to the certainty factor.

According to the invention of claim 12,
The information processing apparatus according to any one of claims 1 to 11, wherein the control unit displays the calculated certainty factor on a display unit.

According to the present invention, the certainty factor for the discrimination result of the discriminator can be calculated, and information on the certainty factor can be provided together with the discrimination result.

It is a figure which shows the medical image system containing the information processing apparatus which concerns on this Embodiment. It is a figure which shows the example of arrangement | positioning of each apparatus when a medical image system is used for a medical facility. It is a figure which shows the functional structure of information processing apparatus. It is a flowchart which shows the process performed by information processing apparatus when calculating a certainty factor. It is a figure which shows an example of a viewer screen. It is a figure which shows an example of a viewer screen.

Embodiments of the present invention will be described below with reference to the drawings.
In this embodiment, in a medical image system that processes medical images, an example of an information processing apparatus that determines a class to which a medical image belongs (a class for each imaging region) by a classifier and provides information on certainty for the determination result. Will be explained.

FIG. 1 is a diagram showing a system configuration of a medical image system 1 including an information processing apparatus 3 according to the present embodiment. FIG. 2 shows an arrangement example of each device when the medical image system 1 is used in a small-scale medical facility.

The medical image system 1 is constructed for a relatively small medical facility such as a medical practitioner or a clinic, and generates and manages a medical image by examining and photographing a patient. The medical image is managed by patient information. The correspondence between the medical image and the patient information is determined by a doctor, and the information on the determined correspondence is input to the medical image system 1 by the operation of the doctor. The medical image system 1 stores a medical image and patient information in association with each other according to the input. In a medical image system constructed for a large-scale medical facility such as a general hospital, order information including patient information and examination information is issued, and medical images are managed based on the order information. This is different from the medical imaging system 1 for a small medical facility. In a medical image system for a large-scale medical facility, the imaging part and the imaging direction of the medical image can be easily determined from the order information. However, order information can be obtained by the medical image system 1 for a small-scale medical facility. If not, the doctor himself / herself needs to specify the imaging part and imaging direction of the medical image. In the medical image system 1, in order to reduce such a burden on the doctor, the imaging region and the imaging direction are discriminated by a discriminator.

As shown in FIG. 1, the medical imaging system 1 includes an ultrasonic diagnostic apparatus 2a, an endoscope apparatus 2b, and a CR (Computed Radiography) apparatus 2c. The ultrasonic diagnostic apparatus 2a, the endoscope apparatus 2b, and the CR apparatus 2c are one type of image generation apparatus that generates a medical image.
The medical image system 1 includes an information processing device 3 and a reception device 4 as shown in FIG. The information processing apparatus 3 is preferably a WS (workstation) provided in an examination room where a doctor is resident.

The ultrasonic diagnostic apparatus 2a, the endoscope apparatus 2b, the CR apparatus 2c, the information processing apparatus 3, and the receiving apparatus 4 are connected to the network 5 via a switching hub (not shown). The network 5 is, for example, a LAN (Local Area Network). As a communication standard, DICOM (Digital-Imaging-and-Communication-in-Medicine), which is a communication standard that handles medical-related data, is generally used.

Next, each device constituting the medical image system 1 will be described.
The ultrasonic diagnostic apparatus 2a emits ultrasonic waves and generates a medical image based on the reflected waves.
A conversion device 21 is connected to the ultrasonic diagnostic apparatus 2 a and is connected to the network 5 via the conversion device 21. The conversion device 21 performs conversion from an analog signal to a digital signal, and converts the format into a format compliant with DICOM when the medical image is in a format not compliant with DICOM. Further, the conversion device 21 adds a UID (unique ID) for individually specifying a medical image in the medical image system 1 to the medical image. The UID is created by combining, for example, a device ID unique to the ultrasound diagnostic apparatus 2a and an imaging date / time.

The endoscope apparatus 2b includes a small imaging device provided at the distal end portion of the tube, and performs imaging using the imaging device to generate a medical image.
The CR device 2c includes an X-ray source that emits X-rays, an X-ray detector that detects X-rays, and the like. The X-ray detector may be a cassette with a built-in volatile phosphor plate that stores X-ray energy, or may be an FPD (Flat Panel Detector). The FPD includes an X-ray detection element and a photoelectric conversion element arranged in a matrix. The photoelectric conversion element photoelectrically converts the X-ray detected by the X-ray detection element to generate a medical image. When a cassette is used, a reading unit is provided in the CR device 2c. The reading unit irradiates the stimulable phosphor plate with excitation light, and photoelectrically converts the stimulated light emitted from the stimulable phosphor plate to generate a medical image.

The endoscope apparatus 2b and the CR apparatus 2c give a UID to the medical image generated by each. The UID is created by combining the device ID of the endoscope device 2b or the device ID of the ultrasonic diagnostic device 2a that respectively generated the medical image with the photographing date and time.

The information processing device 3 performs, for example, information processing of medical images, patient information, and electronic medical record information, displays information necessary for a doctor's examination, and performs image processing on the medical images.
FIG. 3 is a diagram illustrating a functional configuration of the information processing apparatus 3.
As illustrated in FIG. 3, the information processing apparatus 3 includes a control unit 31, an operation unit 32, a display unit 33, a communication unit 34, a storage unit 35, and an image processing unit 36.

The control unit 31 includes a CPU (Central Processing Unit) and a RAM (Random Access Memory). The control unit 31 reads out various programs stored in the storage unit 35 and expands them in the RAM, performs various calculations according to the expanded programs, and centrally controls each component.

Also, the control unit 31 constitutes a discriminator 3a that discriminates the class to which the input data belongs. The discriminator 3a uses the medical image as input data, discriminates the class of the imaging region of the medical image, and outputs the discrimination result. The control unit 31 is a control unit that calculates a certainty factor for the discrimination result by the discriminator 3 a and causes the display unit 33 to display the certainty factor.

The operation unit 32 includes, for example, a keyboard and a mouse, generates an operation signal corresponding to a user operation, and outputs the operation signal to the control unit 31. In addition, you may use the touchscreen with which the display of the operation part 32 and the display part 33 was integrated.
The display unit 33 includes a display, and is a display unit that displays a medical image and various operation screens on the display according to display control of the control unit 31.
The communication unit 34 includes a communication interface and communicates with an external device on the network 5. For example, a medical image is received from the CR device 2c.

The storage unit 35 stores programs used in the control unit 31, parameters necessary for executing the programs, and files. As the storage unit 35, a hard disk or a semiconductor nonvolatile memory can be used.
The storage unit 35 stores a medical image database. This database includes, for example, a medical image, a UID assigned to the medical image, and patient information associated with the medical image.

The storage unit 35 stores information on the types and parameters of image processing specialized for each imaging region for each imaging region (for example, chest, abdomen, and head) of the medical image. The image processing parameter specialized for the imaging region refers to a parameter designed to obtain an appropriate processing result according to the imaging region. For example, the storage unit 35 stores gradation conversion parameter 1 and frequency enhancement parameter 2 information for the chest, and gradation conversion parameter 1 for the abdomen. The storage unit 35 stores general-purpose image processing types and parameter information. General-purpose image processing parameters are parameters designed to be applicable to medical images of any imaging region regardless of the imaging region.

The image processing unit 36 is an image processing unit that performs various types of image processing on medical images. Image processing may be realized as software processing by storing a program for image processing in the storage unit 35 and cooperation between the program and the CPU. Various image processing may be realized by hardware like an image processing circuit.

Examples of the image processing include gradation conversion processing, frequency enhancement processing, dynamic range compression processing, granularity suppression processing, enlargement / reduction processing, display position adjustment processing, and black and white inversion processing.
Such image processing parameters include a gradation conversion curve used in the gradation conversion process (a curve that defines the relationship between the input pixel value and the output pixel value so that the output pixel value becomes the target gradation), Normalization curve slope (G value; value indicating contrast) used for key conversion processing, y-intercept of normalization curve (S value; density correction value), enhancement degree during frequency enhancement processing, enlargement during enlargement / reduction processing Rate (reduction rate).

The reception device 4 performs, for example, reception registration, accounting calculation, and insurance score calculation for a patient who has visited the hospital.

Next, the flow of processing by the medical image system 1 will be described together with the workflows of doctors and patients.
As shown in FIG. 2, there are a patient reception 11 and a waiting room 12 near the entrance 10. The reception device 4 is arranged at the reception 11. On the other hand, the information processing apparatus 3 and the ultrasonic diagnostic apparatus 2a are arranged in the examination room 13, and the CR apparatus 2c is arranged in the X-ray imaging room 15. In the examination room 16, an endoscope apparatus 2b is arranged.

When the patient comes to the hospital, at the reception 11, the reception person in charge gives a reception number tag printed with a reception number for identifying each patient. The receptionist inputs patient information such as the patient's name, age, and address, and the reception number assigned to the patient into the reception device 4. In the reception device 4, patient information is stored in a database, and a patient list in which patient names are arranged in the order of reception numbers is created. The patient list is transmitted to the information processing device 3.

Patients who are given a receipt number move to the examination room 13 in the order of receipt and receive a doctor's examination. When the doctor determines that examination imaging is necessary by medical examination, the information processing apparatus 3 operates to display the patient list. Since the information processing apparatus 3 displays a patient list according to the operation, the doctor performs an operation of selecting a patient to be imaged from the patient list. Thereafter, the doctor guides the patient to the CR device 2c, for example, and sets an imaging region and an imaging direction, and performs an imaging instruction operation on the CR device 2c. In the CR device 2c, after photographing, a medical image is generated, and a UID is attached to the medical image and transmitted to the information processing device 3. After the imaging is finished, the doctor and the patient return to the examination room 13.

In the information processing apparatus 3, the discriminating unit 3a determines the imaging region of the medical image. For example, the feature amount of the medical image is calculated by the control unit 31, and the feature amount is input to the discriminator 3a as input data. Examples of the feature amount of the medical image include an average value of pixel values, a standard deviation, a variance, edge information, a high-order autocorrelation function, and density gradient information. The discriminator 3a discriminates an imaging region from these feature values such as a chest, an abdomen, and a foot. Since the image processing parameters corresponding to the imaging region are stored in the storage unit 35, the image processing unit 36 performs image processing using the parameters.

Thus, the discriminator 3a is used to discriminate the imaging part of the medical image and assists the doctor in identifying the imaging part. The discriminator 3a is configured by a machine learning technique such as AdaBoost, AdaBoostMlt, artificial neural network, and support vector machine. In the present embodiment, an example of a classifier 3a configured using learning data by the AdaBoostMlt method is shown. As learning data, a feature amount of a medical image whose imaging region is already known is used. AdaBoostMlt is an algorithm for creating a final classifier by linear combination of a plurality of weak classifiers. A weak classifier refers to a classifier whose discrimination accuracy may be low. Even if the discrimination accuracy of each weak classifier is low, it is possible to approximate the empirical distribution of learning data because the weak classifiers have sufficient diversity.

A configuration method of the discriminator 3a by the AdaBoostMlt method will be described.
In general, AdaBoost is a technique for configuring classifiers that classify into two classes, while AdaBoostMlt is a technique for configuring classifiers for multi-class classification of two or more classes.

A set of output values Y = {1,... from an input value x that is x∈X⊂R ⁿ (R ⁿ represents an n-dimensional space. X is a subset of the n-dimensional space and is a set of x). , K}, the set of mappings to the power set 2 ^Y is expressed as follows.
W ^mlt = {h ^mlt : X → 2 ^Y }
At this time, the set W of weak classifiers h is expressed as follows.
W = {h: X × Y → R | h (x, y) = [y∈h ^mlt (x), h ^mlt ∈W ^mlt ]}
[] Indicates a function that outputs 1 when the condition in [] is satisfied, and outputs 0 otherwise.

Further, when the learning data is represented by (x ₁ , y ₁ ),..., (X _N , y _N ), the experience distribution p ₀ (x, y) given by the learning data is defined as follows.

x represents vector data defined by the feature amount group of the medical image adopted as the learning data, and y represents a class known for the medical image. N indicates the number of learning data.

When the occurrence probability of the output data is formally expanded to the class y ′ other than the class y of the learning data x with respect to the empirical distribution p ₀ (x, y) of the learning data, the expanded distribution p ₀ (x , Y, y ′) are defined as follows:

Furthermore, each weak classifier h about discrimination error function _e t (h), the total evaluation points _p + a correct determination result (h, p), the total evaluation points of the erroneous discrimination result _p - (h, p), the loss function L (F) is defined as follows.

[] Indicates a function that outputs 1 when the condition in [] is satisfied, and outputs 0 otherwise.

Under the above definition, the weak classifiers h ₁ ,..., H _T obtained by T times of learning are combined, and the function f _λ (x, y) of the classifier 3a finally obtained is expressed as follows. .
f _λ (x, y) = λ ₁ h ₁ (x, y) +,..., λ _T h _T (x, y)
Note that λ ₁ to λ _T are weighting coefficients of the weak classifiers h ₁ to h _T.

In order to obtain the function f _λ (x, y) of the discriminator 3a, the control unit 31 first inputs learning data (x ₁ , y ₁ ),..., (X _N , y _N ), and weights coefficients λ _t , The learning count t is initialized. When λ _t is represented by vector data λ _t = (λ ₁ ,..., λ _T ) having λ ₁ to λ _T as elements in the function f _λ (x, y), the weighting coefficient is λ ₀ = ( 0,..., 0), and the learning count is set to t = 0. The control unit 31 sets the distribution of the expanded training data as p _t. In the case of the initial value t = 0, p ₀ is represented by the above formula p ₀ (x, y, y ′).

Next, the control unit 31 repeats the following processes (1) to (4) until t = 1,.
(1) Select the weak classifier h that minimizes the discrimination error function e (p _t , h _t ). That is, the discrimination error function e ( _pt , _ht ) is calculated while changing the combination of a plurality of weak discriminators h, and the weak discriminator h having the smallest discriminating error function is selected.
(2) λ _t is set by replacing λ _t−1 + (0,..., Α _t ,..., 0). Here, α _t is a parameter for adjusting the weighting coefficient λ _t of the weak classifier h _t of interest, and is expressed as follows.

(3) updates the extended distribution _{p t,} the _{p t + 1} by the following equation.

(4) Set by replacing t with t + 1.
As a result of repeating the above processes (1) to (4), the function f _λ (x, y) of the discriminator 3a is obtained.

Next, processing executed by the information processing apparatus 3 when image processing is performed on a medical image will be described with reference to FIG. In this process, the class of the imaging region of the medical image is determined by the classifier 3a, and the certainty factor for the determination result is calculated and displayed by the control unit 31. The certainty factor refers to the degree of reliability with respect to the discrimination result of the discriminator 3a.

As shown in FIG. 4, first, the image processing unit 36 performs image analysis of a medical image and calculates a feature amount (step S1). The feature amount obtained by the image analysis is input to the discriminator 3a, and the discriminator 3a outputs the degree of possibility that the medical image belongs to the class of each imaging region (this is referred to as an estimated value). The estimated value is given with a value corresponding to the certainty factor for the discrimination result of the discriminator 3a. Specifically, the estimated value y indicating the degree of possibility of being a class k is obtained for each class k when the feature quantity to which the medical image is input is x by the function f _λ (x, y) of the classifier 3a. Is output. Then, the classifier 3a determines that the imaging part of the class for which the maximum estimated value y is output among the estimated values y of each class is the imaging part of the medical image (step S2). For example, in the case of discriminating three classes of imaging regions of the chest, abdomen, and foot, an estimated value y of the chest, an estimated value y of the abdomen, and an estimated value y of the foot are obtained. Among these, when the estimated value y of the chest is the maximum, it is determined that the imaging part of the medical image is the chest.

The certainty factor may be calculated using the estimated value y output from the discriminator 3a by the function f _λ (x, y) as it is, but the estimated value y by the function f _λ (x, y) is linearly combined. It is the sum of the outputs of multiple weak classifiers. Since the value range varies depending on the number of weak classifiers, the control unit 31 normalizes the estimated value y (step S3).
When normalization is performed so that the range of the estimated value y is 0 ≦ y ≦ 1, the probability function f ′ _λ (x | y) of the discriminator 3a that outputs the normalized estimated value y is expressed by the following equation. .
f ′ _λ (x | y) = 1 / Z (y) · (λ ₁ h ₁ (x, y) +... + λ _T h _T (x, y))
Here, Z (y) is a normalization constant determined so that the total sum of Z at each estimated value y is 1.

Note that the exponent of the estimated value y may be normalized, and the estimated value y obtained by normalizing the exponent may be used to calculate the certainty factor. The normalized value obtained by normalizing the exponent of the estimated value y is the estimated value y under the conditional probability distribution. The probability function p (x | y) of the discriminator 3a under the conditional probability distribution is expressed by the following equation.
p (y | x) = 1 / Z · e ^{fλ (x, y)}
Z represents a normalization constant.

Next, the control unit 31 calculates a certainty factor for the discrimination result by the discriminator 3a using the normalized estimated value y. When calculating the certainty factor, the control unit 31 obtains a difference between the maximum estimated value y of the normalized estimated values y of each class and the other estimated values y (step S4). Next, when there are a predetermined number or more of classes in which the obtained difference is equal to or less than the threshold (step S5; Y), the control unit 31 sets and outputs a certainty factor of 0 (step S6). (Step S5; N), the reliability is set to a value of 1 and output (Step S7).

Here, a value of 1 indicates that the certainty factor is high, and a value of 0 indicates that the certainty factor is low. The class of the maximum estimated value y is output as the discrimination result of the discriminator 3a. However, if the difference between the estimated value y of another class and the maximum estimated value y is small, it is considered that the discrimination is difficult. In such a case, since the reliability of the discrimination result is considered to be low, a binary certainty factor is output depending on how many classes have a small difference between the maximum estimated value y and other estimated values.

For example, when the estimated value y (value normalized to a range of 0.00 to 1.00) shown in Table 1 below is obtained for the three classes of the chest, abdomen, and feet, the largest of the three classes The estimated value y is an estimated value y = 0.80 of the chest class. The difference between the estimated value y = 0.80 and the estimated values y of the other abdominal and foot classes is 0.60 and 0.50 as shown in Table 1. If the threshold is 0.20 and the predetermined number of classes below the threshold is set to 2, the estimated values y of the abdominal and foot classes both exceed the threshold, so the number of classes below the threshold is 0. Since the number of classes is not a predetermined number of 2 or more, the certainty factor is set to 1.

If the obtained estimated value y is the estimated value y shown in Table 2 below under the same conditions, the maximum estimated value y is the estimated value y = 0.40 of the chest. Then, the difference between the estimated value y and the maximum estimated value y of the other abdominal and foot classes is less than the threshold value, and the number of classes less than or equal to the threshold value is 2. Since the number 2 of this class is a predetermined number 2 or more, the certainty factor is set to zero.

The certainty factor may be an index that can indicate the degree of reliability with respect to the determination result of the discriminator 3a, and the certainty factor may be calculated by another method. For example, there is a method of obtaining a binary certainty factor based on whether or not the difference between the maximum estimated value y and the second largest estimated value y (hereinafter referred to as quasi-estimated value y) is equal to or less than a threshold value. In this method, the certainty factor is determined by focusing on the relative positioning of the maximum estimated value y, how far the class of the maximum estimated value is from the semi-estimated value y.

Specifically, the control unit 31 obtains the absolute value of the difference between the maximum estimated value y and the semi-estimated value y, and if the obtained absolute value is less than or equal to the threshold value, sets the confidence value to 0 and outputs it. If it is not less than the threshold, the confidence value is set to 1 and output. A value of 1 indicates that the certainty factor is high, and a value of 0 indicates that the certainty factor is low. If the difference from the quasi-estimated value is less than or equal to the threshold value, it was difficult to discriminate from the quasi-estimated class. This is because the possibility is high and the reliability is high.

For example, when the estimated value y shown in Table 3 below is obtained for three classes of the chest, abdomen, and feet, the maximum estimated value y is 0.30 for the chest and the quasi-estimated value y is 0.20 for the abdomen. . Since the absolute value of the difference between the maximum estimated value y and the quasi-estimated value y is 0.10, if the threshold value is set to 0.15, the absolute value of the difference is less than or equal to the threshold value, and the certainty factor is Set to 0 and output.

Note that the absolute value of the difference between the maximum estimated value y and the semi-estimated value y may be calculated as the certainty factor. The absolute value of the difference is the difference in possibility between the class of the maximum estimated value y that is most likely to belong to the class and the class of the semi-estimated value y that is the second most likely. It becomes an index to show and shows the reliability for the discrimination result.

As another method for calculating the certainty factor, the certainty factor is set to a value of 0 if the maximum estimated value y is not more than the threshold value, and the certainty factor is set to a value of 1 if it is not less than the threshold value. The method of doing is also mentioned. This is because the maximum estimated value y is compared with other classes, but if the maximum estimated value y itself is so small that it does not reach a certain value, it is considered that the reliability of the discrimination result is low.

For example, when the estimated value y shown in Table 4 below is obtained for three classes of the chest, abdomen, and feet, the maximum estimated value y is 0.30 for the chest. Here, if the threshold value is 0.5, the maximum estimated value y is equal to or less than the threshold value, and thus a certainty value of 0 is output.

Furthermore, as another method for calculating the certainty factor, a combination of classes that can be easily discriminated is set in advance, and the maximum estimated value y and the class of the maximum estimated value y are combined. There is a method of determining the certainty factor based on whether or not the difference from the estimated value y of the class is equal to or less than a threshold value. If the threshold value is equal to or less than the threshold value, the value of 1 indicating that the certainty factor is high is set as the certainty factor. This is because, when the difference in the estimated value y is small between classes that are easy to discriminate, it is considered that the discrimination is easy but the discrimination is not performed well and the reliability of the discrimination result is low.

For example, when the estimated value y shown in Table 5 below is obtained for the three classes of the chest, abdomen, and feet, the maximum estimated value y is 0.50 for the chest. Here, the chest and the foot are set as a combination of classes that can be easily discriminated, and if the threshold is 0.20, the difference between the estimated value y of the chest and the foot is 0.03. is there. The image characteristics of the chest and feet are greatly different and the difference between the estimated values y is less than the threshold value despite the fact that the discrimination is easy, and the reliability is considered to be low.

When the certainty factor is calculated in this way, the control unit 31 determines whether the certainty factor is high or low based on the calculated certainty factor value (step S8). As described above, depending on the certainty factor calculation method, the certainty factor is output as a continuous value such as the absolute value of the difference between the maximum estimated value and the semi-estimated value when the certainty factor is output as a binary value of 1 or 0. There are cases. When the certainty factor is output as a binary value, a value of 0 indicates that the certainty factor is low, and a value of 1 indicates that the certainty factor is high. Therefore, the certainty factor depends on whether the value is 0 or 1. What is necessary is just to determine whether the degree is high or low. On the other hand, when the certainty factor is output as a continuous value, it may be determined whether it is high or low by comparison with a threshold value. For example, the control unit 31 determines that the certainty factor is low if the absolute value of the difference between the maximum estimated value and the quasi-estimated value is equal to or less than the threshold value, and otherwise, the certainty factor is high.

Here, a case where the certainty factor is output in binary will be described as an example. If the certainty factor is a value of 1, the control unit 31 determines that the certainty factor is high (step S8; Y), and the image processing specialized for the class output as a discrimination result by the discriminator 3a is performed by the image processing unit 36. Is applied to the medical image (step S9). For example, when the classifying unit of the medical image is determined to be the chest by the discriminator 3a, the control unit 31 acquires the image processing type and parameter information determined for the chest from the storage unit 35, and performs image processing. To the unit 36. The image processing unit 36 executes the type of image processing determined for the chest using the determined parameters.

On the other hand, if the certainty factor is 0, the control unit 31 determines that the certainty factor is low (step S8; N), and the image processing unit 36 performs general-purpose image processing on the medical image (step S10). For example, even when the class of the imaging region of the medical image is determined to be the chest by the discriminator 3a, the control unit 31 acquires general-purpose image processing type and parameter information from the storage unit 35, and the image The data is output to the processing unit 36. The image processing unit 36 executes a type of image processing defined as general-purpose using a predetermined parameter.

Thus, when the certainty factor for the determination result of the imaging region of the medical image is calculated and image processing is performed, a viewer screen D1 as shown in FIG. 5 is displayed on the display unit 13 by display control of the control unit 31. Is done.
As shown in FIG. 5, a patient information field d1, an image display field d2, and an image adjustment field d3 are displayed on the viewer screen D1. The patient information column d1 is a display column for patient information of a patient on which a medical image is displayed. The image display field d2 is a display field for medical images obtained by examination imaging. The image adjustment column d3 is a display column for operation buttons used for operating parameters of image processing performed on the displayed medical image.

The image display field d2 includes imaging part information d21 of the displayed medical image. This imaging region is a result determined by the discriminator 3a. If the patient information displayed in the patient information column d1 corresponds to the medical image displayed in the image display column d2, and the imaging region displayed in the image display column d2 is correct, the doctor may operate the OK button d22. . In addition, when the correspondence or the imaging region is incorrect, the doctor may operate the NG button d23.

When the OK button d22 is operated, in the information processing apparatus 3, the medical image, patient information, UID assigned to the medical image, and the like are made into a database by the control unit 31 and stored in the storage unit 35. When the NG button d23 is operated, a correction screen for correcting the imaging region is displayed by the display control of the control unit 31, and thus the doctor inputs the imaging region determined by visual observation on the correction screen. In the information processing apparatus 3, image processing is performed again with parameters according to the imaging region input by the image processing unit 36, and the viewer screen is displayed with the re-image-processed medical image.

When it is determined that the certainty factor for the determination result of the imaging region is low, the control unit 31 shades or adds a color to the character of the imaging region of the medical image determined that the certainty factor is low on the viewer screen D1. To warn that the certainty level is low. In addition, the control unit 31 displays a message d4 such as “Please check the imaging region” to warn that the certainty level is low.

Since the doctor can easily determine the imaging part of the medical image by visual observation, it can be confirmed immediately by the message d4 whether or not the determination result of the classifier 3a is erroneously determined. Such a viewer screen D2 may be further displayed. The viewer screen D2 is a screen that displays detailed information for each medical image. In this viewer screen D2, as shown in FIG. 6, as the discrimination result of the discriminator 3a, information d5 indicating that the discriminated imaging region is the chest, its estimated value is 0.4, and the certainty factor is 0. It is displayed. Further, since the degree of certainty is low, a message d6 is displayed notifying the user that general-purpose image processing has been performed on the medical image, not image processing specialized for the chest.

As described above, according to the present embodiment, the discriminator 3a determines the probability that the medical image belongs to the class of each imaging region from the feature amounts obtained by analyzing the medical image. Output as an estimated value, discriminate the imaging region class of the medical image based on the estimated value, and the control unit 31 uses the estimated value of each imaging region class output by the discriminator 3a to The certainty factor for the determination result by 3a is calculated, and the display unit 33 displays the calculated certainty factor. Thereby, the information on the certainty factor for the discrimination result can be provided together with the discrimination result of the imaging part by the discriminator 3a.

Further, the control unit 31 determines whether the calculated certainty factor is high or low, and when it is determined that the certainty factor is high, the image processing unit 36 selects an imaging region having the maximum estimated value for the medical image. When class-specific image processing is performed and it is determined that the certainty level is low, general-purpose image processing is performed on the medical image. As a result, general-purpose image processing can be selected when the possibility of erroneous determination is high, and inappropriate image processing can be prevented from being performed.

In addition, since the control unit 31 displays warning information such as a message, a color, and shading that warns that the certainty factor is low on the display unit 33, the user can easily grasp that the certainty factor is low.

In addition, the above-mentioned embodiment is a suitable example of this invention, and is not limited to this.
For example, the characteristics of a medical image differ depending on the photographing direction and the image generation apparatus in which the medical image has been photographed. Therefore, the present invention may be applied in the case of class discrimination for each imaging direction instead of the class for each imaging region, or may be applied in the case of class determination for each combination of imaging region and imaging direction. . Alternatively, the present invention can also be applied to class determination for each image generation apparatus.

In addition, the present invention can be applied not only to medical images but also to general image discrimination by a classifier.

Further, as a computer-readable medium for the program used in the information processing apparatus 3 described above, a non-volatile memory such as a ROM or a flash memory, or a portable recording medium such as a CD-ROM can be applied. Further, a carrier wave (carrier wave) can also be applied as a medium for providing the program data via a communication line.

It can be used for an information processing apparatus that performs discrimination using a discriminator.

DESCRIPTION OF SYMBOLS 1 Medical image system 2a Ultrasound diagnostic apparatus

2b Endoscope apparatus

2c CR apparatus 3 Information processing apparatus 31 Control part 3a Classifier 32 Operation part 33 Display part 35 Storage part 36 Image processing part 4 Reception apparatus

Claims

From the feature amount obtained by analyzing the image, a probability that the image belongs to each class among a plurality of classes is output as an estimated value, and a discriminator that determines the class to which the image belongs based on the estimated value;
Control means for calculating a certainty factor for the discrimination result by the classifier using the estimated value of each class output by the classifier;
An information processing apparatus comprising:
The control means obtains a difference between the maximum estimated value and the estimated value other than the estimated value of each class output by the discriminator, and there are a predetermined number or more classes having the difference within a threshold value. The information processing apparatus according to claim 1, wherein a binary certainty factor is output depending on whether or not.
The said control means outputs the absolute value of the difference of the largest estimated value of the estimated value of each class output by the said discriminator, and the 2nd largest estimated value as a certainty factor. Information processing device.
The control means obtains an absolute value of a difference between a maximum estimated value of the estimated values of each class output from the classifier and a second largest estimated value, and the absolute value of the difference is smaller than a threshold value. The information processing apparatus according to claim 1, wherein a binary certainty factor is output depending on whether or not.
2. The information processing apparatus according to claim 1, wherein the control unit outputs a binary certainty factor depending on whether or not a maximum estimated value is equal to or less than a threshold value among the estimated values of each class output by the classifier. .
The control means normalizes the estimated value of each class output by the discriminator and calculates the certainty factor using the normalized estimated value, or claims of the exponent of the estimated value of each class 6. The information processing apparatus according to any one of 1 to 5.
The control means determines whether the certainty factor is high or low based on the calculated certainty factor value, and when it is determined that the certainty factor is low, warns that the certainty factor is low together with the determination result of the class. The information processing apparatus according to any one of claims 1 to 6, wherein warning information to be displayed is displayed on the display means.
The image is a medical image in which a patient is photographed by an image generation device,
The control means determines whether the calculated certainty factor is high or low,
When it is determined by the control means that the certainty level is high, image processing specialized for the class having the maximum estimated value is performed on the medical image, and when the certainty level is determined to be low, the medical image is The information processing apparatus according to any one of claims 1 to 7, further comprising image processing means for performing general-purpose image processing.
The information processing apparatus according to claim 8, wherein the class of the medical image is a class for each type of the image generation apparatus in which the medical image is captured.
The information processing apparatus according to claim 8, wherein the class of the medical image is any one of a class for each imaging region where the medical image is captured, a class for each imaging direction, or a class for each combination of the imaging region and the imaging direction.
The discriminator is configured by the AdaBoostMlt method, performs discrimination of at least three classes,
The information processing apparatus according to any one of claims 1 to 10, wherein the estimated value output by the discriminator is given together with a value corresponding to the certainty factor.
12. The information processing apparatus according to claim 1, wherein the control unit displays the calculated certainty factor on a display unit.