Jaundice analysis system and method

Abstract

The jaundice analysis system includes a database and a processing device that accesses the database. The processing device includes a data processing module that generates training data in response to first image data, correlates the training data with first category data, and stores the training data in the database, and a deep learning module that acquires a trained convolutional neural network module by training a target convolutional neural network module using the training data correlated with the category data. The image data includes a first sclera image. The trained convolutional neural network module of the processing device generates test data in response to input image data. The input image data includes a second sclera image of a target subject. The test data indicates a bilirubin concentration range of the target subject.

Description

[Technical field]

[0001] The present disclosure relates to a system and method for analyzing jaundice, and more particularly to a system and method for analyzing jaundice that determines whether a subject has jaundice symptoms based on a sclera image of the subject. [Background technology]

[0002] Compared with other metabolic diseases, jaundice is accompanied by specific symptoms, such as yellowish skin and sclera. However, if a patient with jaundice symptoms is born with yellowish skin or mistakes the skin color caused by jaundice as a sunburned skin color, they are likely to overlook their own disease or organ failure. As a result, their jaundice or organ degeneration is less likely to be diagnosed early. Therefore, it is essential to provide a jaundice analysis system and method that can determine whether a target subject has jaundice symptoms based on the sclera image of the target subject. Summary of the Invention [Problem to be solved by the invention]

[0003] In view of the above-mentioned shortcomings of the prior art, an object of the present disclosure is to provide a jaundice analysis system and method for determining whether a target subject has jaundice symptoms based on the target subject's sclera image. [Means for solving the problem]

[0004] In order to achieve the above and other objects, the present disclosure provides a jaundice analysis system, comprising a database and a processing device accessing the database. The processing device includes a data processing module for generating first training data in response to first image data, correlating the first training data with first category data, and storing the first training data in a database, and a deep learning module for training a target convolutional neural network module using the first training data correlated with the first category data to obtain a trained convolutional neural network module. The first image data includes a first sclera image. The database is communicatively connected to the data processing module and / or the deep learning module. The trained convolutional neural network module of the processing device generates test data in response to input image data. The input image data includes a second sclera image of the target subject. The test data indicates a bilirubin concentration range of the target subject.

[0005] In a preferred embodiment of the present disclosure, the deep learning module obtains the trained convolutional neural network module by transfer learning, and thus the target convolutional neural network module is a trained convolutional neural network module (the training is not limited to detecting jaundice symptoms or measuring bilirubin concentration).

[0006] In a preferred embodiment of the present disclosure, the data processing module performs a first cropping process on the first image data to generate first cropped image data, and generates first learning data according to the first cropped image data.

[0007] In a preferred embodiment of the present disclosure, the data processing module performs a mirroring process on the first image data to generate mirrored image data, and generates first learning data according to the mirrored image data.

[0008] In a preferred embodiment of the present disclosure, the data processing module performs a second cropping process on the mirroring image data to generate second cropped image data, and generates first learning data according to the second cropped image data, where the second cropped image data has a specific image shape.

[0009] In a preferred embodiment of the present disclosure, the data processing module performs a third cropping process on the second cropped image data to generate third cropped image data, and generates first learning data according to the third cropped image data.

[0010] In a preferred embodiment of the present disclosure, the data processing module performs a glare removal process on the first image data to generate glare removed image data, and generates first learning data according to the glare removed image data.

[0011] In a preferred embodiment of the present disclosure, the data processing module generates second image data according to the first image data, generates second training data according to the second image data, correlates the second training data with the second category data, and stores the second training data in a database, and the deep learning module acquires a trained convolutional neural network module by training a target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data.

[0012] In a preferred embodiment of the present disclosure, the data processing module performs one of an image conversion process, an image rotation process, and an image flip process on the first image data to generate second image data.

[0013] In a preferred embodiment of the present disclosure, the jaundice analysis system further comprises a mobile terminal storing the input image data, and the processing device further includes a communication module communicatively connected to the mobile terminal and the trained convolutional neural network module of the processing device, the communication module receiving the input image data from the mobile terminal and transmitting the test data to the mobile terminal.

[0014] In order to achieve the above and other objects, the present disclosure provides a jaundice analysis method applicable to a jaundice analysis system, the jaundice analysis method including: a data processing module of the jaundice analysis system generates first learning data according to first image data, and the data processing module correlates the first learning data with first category data; a deep learning module of the jaundice analysis system learns a target convolutional neural network module using the first learning data correlated with the first category data to obtain a trained convolutional neural network module; and a trained convolutional neural network module of the jaundice analysis system generates test data according to input image data. The input image data includes a second sclera image of the target subject. The first image data includes a first sclera image. The test data indicates a bilirubin concentration range of the target subject.

[0015] In a preferred embodiment of the present disclosure, the deep learning module obtains the trained convolutional neural network module by transfer learning, and thus the target convolutional neural network module is a trained convolutional neural network module (the training is not limited to detecting jaundice symptoms or measuring bilirubin concentration, but is adapted to include, for example, detecting other matters according to related images, but the present disclosure is not limited thereto).

[0016] In a preferred embodiment of the present disclosure, the step of generating first learning data in response to the first image data further includes a step of a data processing module performing a first cropping process on the first image data to generate first cropped image data, and the data processing module generates the first learning data in response to the first cropped image data.

[0017] In a preferred embodiment of the present disclosure, the step of generating first learning data in response to the first image data further includes a step of the data processing module performing a mirroring process on the first image data to generate mirrored image data, and the data processing module generates the first learning data in response to the mirrored image data.

[0018] In a preferred embodiment of the present disclosure, the step of generating first learning data in response to the first image data further includes a step of the data processing module performing a second cropping process on the mirrored image data to generate second cropped image data, wherein the data processing module generates the first learning data in response to the second cropped image data, and the second cropped image data has a specific image shape.

[0019] In a preferred embodiment of the present disclosure, the step of generating first learning data in response to the first image data further includes a step of the data processing module performing a third cropping process on the second cropped image data to generate third cropped image data, and the data processing module generates the first learning data in response to the third cropped image data.

[0020] In a preferred embodiment of the present disclosure, the step of generating first learning data in response to the first image data further includes a step of the data processing module performing a glare removal process on the first image data to generate glare removed image data, and the data processing module generates the first learning data in response to the glare removed image data.

[0021] In a preferred embodiment of the present disclosure, the jaundice analysis method further includes a step of: a data processing module generating second image data according to the first image data; a step of the data processing module generating second training data according to the second image data, and a step of the data processing module correlating the second training data with second category data; and a step of the deep learning module acquiring a trained convolutional neural network module by training a target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data.

[0022] In a preferred embodiment of the present disclosure, the data processing module performs one of an image conversion process, an image rotation process, and an image flip process on the first image data to generate second image data.

[0023] In a preferred embodiment of the present disclosure, the jaundice analysis method further includes the steps of: a communication module of the jaundice analysis system receiving input image data from a mobile terminal; and the communication module transmitting test data to the mobile terminal.

[0024] The foregoing and other aspects of the disclosure are illustrated in the following non-limiting specific embodiments and are illustrated in the accompanying drawings. [Brief description of the drawings]

[0025] [Figure 1] FIG. 1 is a block diagram of a jaundice analysis system in accordance with certain embodiments of the present disclosure.

[0026] [Figure 2A] FIG. 2 is a schematic diagram of first image data in accordance with a specific embodiment;

[0027] [Figure 2B] FIG. 2 is a schematic diagram of first image data in accordance with a specific embodiment;

[0028] [Figure 2C] FIG. 2 is a schematic diagram of first image data in accordance with a specific embodiment;

[0029] [Diagram 3] FIG. 2 is a schematic diagram of mirroring image data in accordance with a specific embodiment.

[0030] [Figure 4] FIG. 11 is a schematic diagram of second cropped image data in a specific embodiment.

[0031] [Diagram 5] 11 is a schematic diagram of a method for generating third cropped image data in a specific embodiment. FIG.

[0032] [Figure 6] 1 is a schematic diagram of a method for generating glare-removed image data in a specific embodiment.

[0033] [Figure 7] FIG. 1 is a schematic diagram of a process flow for a jaundice analysis method in accordance with certain embodiments of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0034] Referring to FIG. 1, a block diagram of a jaundice analysis system 100 in a specific embodiment of the present disclosure is shown. In the embodiment shown in FIG. 1, the jaundice analysis system 100 includes a database 110 and a processing device 120. The processing device 120 can access the database 110 and is therefore communicatively connected to the database 110. The processing device 120 includes a data processing module 122 and a deep learning module 124. Preferably, the data processing module 122 is communicatively connected to the database 110, and the deep learning module 124 is also communicatively connected to the database 110. In the embodiment shown in FIG. 1, the data processing module 122 generates first learning data according to first image data, associates the first learning data with first category data, and stores the first learning data in the database 110. The first image data includes a first sclera image. The first category data indicates a bilirubin concentration value or a bilirubin concentration range. Jaundice is caused by an excess of bilirubin circulating in the blood, which causes the skin, sclera, and mucous membranes to have a yellowish appearance. Therefore, the bilirubin concentration value or bilirubin concentration range indicates the degree of jaundice symptoms, or in other words, the first category data indicates the degree of jaundice symptoms. In a specific embodiment, the first image data is stored in the database 110.

[0035] It should be noted that the data processing module 122 generates different first learning data according to different first image data, and the different first learning data are correlated with different first category data. In some cases, the different first learning data are correlated with the same first category data. For example, the data processing module 122 generates a first group of first learning data according to a first group of first image data, and each first learning data in the first group of first learning data is correlated with first category data indicating a first bilirubin concentration range. The data processing module 122 generates a second group of first learning data according to a second group of first image data, and each first learning data in the second group of first learning data is correlated with another first category data indicating a second bilirubin concentration range. The first bilirubin concentration range is different from the second bilirubin concentration range. In certain embodiments, the first image data is already correlated with the first category data before the processing device 120 or data processing module 122 receives the first image data (e.g., from the database 110), and the data processing module 122 correlates the first learning data generated in response to the first image data with the first category data.

[0036] In the embodiment shown in FIG. 1, the deep learning module 124 obtains (or generates) the trained convolutional neural network module 128 by training the target convolutional neural network module using the first training data correlated with the first category data. Preferably, the deep learning module 124 obtains the trained convolutional neural network module 128 by transfer learning. Thus, the target convolutional neural network module becomes the trained convolutional neural network module. In a specific embodiment, the deep learning module 124 undergoes transfer learning according to EfficientNetB5. Note that the trained convolutional neural network module 128 generated by the deep learning module 124 is also included in the processing device 120. Thus, the trained convolutional neural network module 128 of the processing device 120 generates test data in response to input image data. The input image data includes a second sclera image of the target subject. The test data indicates a bilirubin concentration range of the target subject.

[0037] In a particular embodiment, the deep learning module 124 generates various filters by itself to obtain different eigenvalues ​​in the process of training the target convolutional neural network module using the first training data correlated with the first category data. These filters are, for example, a histogram filter, a Clahe (adaptive histogram equalization) filter, and a Gaussian filter, but the present disclosure is not limited thereto. In a particular embodiment, the deep learning module 124 is communicatively connected to the target convolutional neural network module and the trained convolutional neural network module 128. In a particular embodiment, the deep learning module 124 includes the target convolutional neural network module and the trained convolutional neural network module 128. In a particular embodiment, the data processing module 122 is communicatively connected to the deep learning module 124 and / or the trained convolutional neural network module 128.

[0038] In a particular embodiment, the data processing module 122 obtains more training data by generating second image data according to the first image data and generating second training data according to the second image data, thereby improving the accuracy of the analysis performed by the trained convolutional neural network module 128 for the bilirubin concentration range or jaundice degree. The data processing module 122 then correlates the second training data with the second category data and stores the second training data in the database 110. Preferably, the second category data is the first category data correlated with the first image data. In a particular embodiment, the deep learning module 124 obtains the trained convolutional neural network module 128 by training the target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data.

[0039] In a particular modified embodiment, the data processing module 122 generates the second image data by performing an image conversion process on the first image data (e.g., various conversion processes such as horizontal conversion and vertical conversion are performed on the first image data, but the present disclosure is not limited thereto), an image rotation process on the first image data (e.g., a rotation process of 0 to 180 degrees is performed on the first image data, but the present disclosure is not limited thereto), an image inversion process on the first image data (e.g., various inversion processes such as horizontal inversion and vertical inversion are performed on the first image data, but the present disclosure is not limited thereto), or a gap correction constant process on the first image data. Note that the present disclosure does not limit the method in which the data processing module 122 generates the second image data according to the first image data.

[0040] In a particular embodiment, the processing device 120 further includes a communication module 126. The processing module 122 receives data (e.g., input image data) from the terminal 900 via the communication module 126 or transmits data (e.g., test data) to the terminal 900 via the communication module 126. The communication module 126 is communicatively connected to the terminal 900 and the trained convolutional neural network module 128 of the processing device 120. The terminal 900 may be, for example, a computer, a mobile terminal (or may be in the form of a computer), or a remote server, but the present disclosure is not limited thereto. In a particular embodiment, the terminal 900 is considered as a part of the jaundice analysis system 100, and the input image data is stored in the terminal 900. In a particular embodiment, the terminal 900 includes an image capture device, whereby the terminal 900 captures an image and generates the input image data. Preferably, the input image data includes first input image data and second input image data. The first input image data includes a left sclera image of the subject. The second input image data includes a right sclera image of the subject. In particular embodiments, the communications module 126 is communicatively coupled to the data processing module 122 and / or the deep learning module 124.

[0041] In a particular embodiment, the jaundice analysis system 100 of the present disclosure includes one or more processors and implements the database 110 and the processing unit 120 through a combination of hardware and software. In a particular embodiment, the processing unit 120 includes one or more processors and implements the data processing module 122, the deep learning module 124, the communication module 126, and the trained convolutional neural network module 128 through a combination of hardware and software. In a particular embodiment, the terminal 900 includes one or more processors and implements the image capture device through a combination of hardware and software.

[0042] 2A to 2C, different first image data are shown. The first category data correlated with the first image data 201, 202 in FIG. 2A indicates that the bilirubin concentration range is 0 to 1.2 mg / dL. The first category data correlated with the first image data 203, 204 in FIG. 2B indicates that the bilirubin concentration range is 1.3 to 3.5 mg / dL. The first category data correlated with the first image data 205, 206 in FIG. 2C indicates that the bilirubin concentration range is greater than 3.6 mg / dL. In a specific embodiment, the data processing module of the jaundice analysis system 100 of the present disclosure performs a first cropping process on the first image data to generate first cropped image data, thereby cropping or adjusting the first image data so that the first image data have the same image size. The data processing module generates first learning data according to the first cropped image data.

[0043] Referring to FIG. 3, different mirrored image data 302, 304 are shown. The mirrored image data 302, 304 are generated when the data processing module performs different mirroring processes on the first image data. The short side of the image data 304 is complemented by a mirroring process (where the short side serves as an axis of symmetry) that makes the length and width of the image data 304 equal. In a particular embodiment, the first cropped image data is considered as the first image data. Thus, the data processing module performs a mirroring process on the first image data on which the first cropping process has been performed. In a particular embodiment, the data processing module generates first learning data according to the mirrored image data.

[0044] Referring to FIG. 4, a schematic diagram of the second cut-out image data in a specific embodiment is shown. In the embodiment shown in FIG. 4, the data processing module performs a second cut-out process on the mirrored image data 304 of FIG. 3 to generate the second cut-out image data 402. In a specific embodiment, the cut-out of the second cut-out process includes using one side of the mirrored image data 304 as the diameter of a tangent circle (a circle inscribed in a square), cut out the image content outside the tangent circle, and keep the image content inside the tangent circle. Thus, the data processing module cuts out the different mirrored image data so that the different mirrored image data have the same image size. Note that FIG. 4 merely serves the purpose of illustration, and thus the purpose of the data processing module performing the second cut-out process is not limited to cutting out an image so that the image is circular. In one variant, the image is cut out so that the image has a specific image shape as required. In a specific embodiment, the data processing module generates the first learning data according to the second cut-out image data.

[0045] Referring to FIG. 5, a schematic diagram of a method for generating third cropped image data in a specific embodiment is shown. In the embodiment shown in FIG. 5, the data processing module performs a third cropping process on the second cropped image data 502 to generate third cropped image data 504. In a specific embodiment, the cropping of the third cropping process includes using the center of the second cropped image data 502 to take a square of a certain size, and keeping the image content inside the square to reduce unnecessary features. In a specific embodiment, the data processing module generates first training data according to the third cropped image data.

[0046] Referring to FIG. 6, a schematic diagram of a method for generating glare-removed image data in a specific embodiment is shown. In the embodiment shown in FIG. 6, the data processing module performs a glare removal process on the first image data 602 to generate glare-removed image data 604. Thus, according to this embodiment, the deep learning module is able to train the target convolutional neural network module by suppressing the white portion of the eyeball reflection light in the image from interfering with the deep learning. In a specific embodiment, the glare removal process is performed by removing the eyeball reflection light portion using the mimicking function of PhotoImpact. In a specific embodiment, the data processing module generates the first learning data according to the glare-removed image data.

[0047] In addition, the first cropped image data, mirrored image data, second cropped image data, third cropped image data and reflection removal image data are regarded as first image data and a data processing module, and the first cropping process, mirroring process, second cropping process, third cropping process and / or reflection removal process can be performed on the image data.

[0048] Referring to FIG. 7, a schematic diagram of a process flow of a jaundice analysis method 700 in a specific embodiment of the present disclosure is shown. In the embodiment shown in FIG. 7, the jaundice analysis method 700 is applicable to the jaundice analysis system 100. The jaundice analysis method 700 starts at step 710, where the data processing module of the jaundice analysis system 100 generates first training data according to first image data, and correlates the first training data with first category data. The first image data includes a first sclera image. Then, at step 720, the deep learning module of the jaundice analysis system 100 obtains a trained convolutional neural network module by training a target convolutional neural network module using the first training data that is correlated with the first category data. In a specific embodiment, the deep learning module obtains the trained convolutional neural network module by transfer learning. Thus, the target convolutional neural network module is a trained convolutional neural network module that has been pre-trained and is not limited to detecting jaundice symptoms or measuring bilirubin concentrations, but is adapted to include, for example, detecting other matters responsive to associated images, although the present disclosure is not limited thereto.

[0049] In step 730, the communication module of the jaundice analysis system 100 receives input image data from a mobile terminal (e.g., a mobile phone or a tablet, but the present disclosure is not limited thereto). The input image data includes a second sclera image of the target subject. Preferably, the input image data includes a first input image data and a second input image data. The first input image data includes a left sclera image of the target subject. The second input image data includes a right sclera image of the target subject. Note that step 730 may precede step 710 or step 720 as necessary.

[0050] Steps 710 to 730 are followed by step 740. In step 740, the trained convolutional neural network module of the jaundice analysis system 100 generates test data according to the input image data. The test data indicates a bilirubin concentration range of the target subject. The bilirubin concentration range indicates the degree of jaundice. Then, in step 750, the communication module of the jaundice analysis system 100 transmits the test data to the mobile terminal.

[0051] In a specific embodiment, the step of generating the first learning data according to the first image data further includes a step of the data processing module performing a first cropping process on the first image data to generate first cropped image data. The data processing module generates the first learning data according to the first cropped image data. In a specific embodiment, the data processing module performs the first cropping process on the first image data according to a first command. The first command is, for example, an image cropping operation performed by a user using a mouse, or, for example, a default image cropping command, but the present disclosure is not limited thereto.

[0052] In a specific embodiment, the step of generating the first learning data in response to the first image data further includes a step of the data processing module performing a mirroring process on the first image data to generate mirrored image data. The data processing module generates the first learning data in response to the mirrored image data. In a specific embodiment, the data processing module performs the mirroring process on the first image data in response to a second command. The second command is, for example, an image mirroring operation performed by a user using a mouse, or, for example, a default image mirroring command, but the present disclosure is not limited thereto.

[0053] In a specific embodiment, the step of generating the first learning data according to the first image data further includes a step of the data processing module performing a second cropping process on the mirrored image data to generate second cropped image data. The second cropped image data has a specific image shape. The data processing module generates the first learning data according to the second cropped image data. In a specific embodiment, the data processing module performs the second cropping process on the mirrored image data according to a third command. The third command is, for example, an image cropping operation performed by a user using a mouse, or, for example, a default image cropping command, but the present disclosure is not limited thereto.

[0054] In a specific embodiment, the step of generating the first learning data in response to the first image data further includes a step of the data processing module performing a third cropping process on the second cropped image data to generate third cropped image data. The data processing module generates the first learning data in response to the third cropped image data. In a specific embodiment, the data processing module performs the third cropping process on the second cropped image data in response to a fourth command. The fourth command is, for example, an image cropping operation performed by a user using a mouse, or, for example, a default image cropping command, but the present disclosure is not limited thereto.

[0055] In a specific embodiment, the step of generating the first learning data in response to the first image data further includes a step of the data processing module performing a glare removal process on the first image data to generate glare removed image data. The data processing module generates the first learning data in response to the glare removed image data. In a specific embodiment, the data processing module performs the glare removal process on the first image data in response to a fifth command. The fifth command is, for example, a glare removal operation performed by a user using a mouse, or, for example, a default image glare removal command, but the present disclosure is not limited thereto.

[0056] In a specific embodiment, in order to obtain more training data and thus improve the accuracy of the analysis performed by the trained convolutional neural network module for the bilirubin concentration range or jaundice degree, the jaundice analysis method 700 further includes the steps of: a data processing module generating second image data according to the first image data; a data processing module generating second training data according to the second image data, and a data processing module correlating the second training data with second category data. Preferably, the second category data is first category data correlated with the first image data. In a specific embodiment, the deep learning module obtains the trained convolutional neural network module by training the target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data.

[0057] In a particular variation, the data processing module performs one of an image conversion process, an image rotation process, an image flip process, and a gap correction constant process on the first image data to generate the second image data. Note that the present disclosure does not limit the manner in which the data processing module generates the second image data in response to the first image data.

[0058] Thus, the jaundice analysis system and method of the present disclosure are shown by the attached drawings and described above. The specific embodiments of the present disclosure are merely for illustrative purposes, and various changes may be made thereto without departing from the spirit and scope of the claims of the present disclosure, which are to be included within the scope of the claims of the present disclosure. Thus, the specific embodiments of the present disclosure are not intended to limit the present disclosure, but rather the spirit and scope of the present disclosure are to be defined by the appended claims.

Claims

1. 1. A jaundice analysis system comprising: A database; a processor for accessing the database, the processor comprising: a data processing module that generates first training data in response to first image data, correlates the first training data with first category data indicative of a bilirubin concentration range of the first image data, and stores the first training data in the database; and a deep learning module for acquiring a trained convolutional neural network module by training a target convolutional neural network module using the first training data that is correlated with the first category data; A processing device including: Equipped with the first image data includes a first sclera image; the database is communicatively connected to the data processing module and / or the deep learning module; the trained convolutional neural network module of the processing device generates test data in response to input image data, the input image data including a second sclera image of the target subject; the test data indicating a bilirubin concentration range for the subject; Jaundice Analysis System.

2. The deep learning module obtains the trained convolutional neural network module by transfer learning. The jaundice analysis system according to claim 1 .

3. the data processing module executes a first cropping process on the first image data to generate first cropped image data, and generates the first learning data in accordance with the first cropped image data; The jaundice analysis system according to claim 1 .

4. The data processing module executes a mirroring process on the first image data to generate mirrored image data, and generates the first learning data according to the mirrored image data. The jaundice analysis system according to claim 1 .

5. The data processing module performs a second cropping process on the mirroring image data to generate second cropped image data, and generates the first learning data according to the second cropped image data, and the second cropped image data has a specific image shape. The jaundice analysis system according to claim 4.

6. the data processing module executes a third cropping process on the second cropped image data to generate third cropped image data, and generates the first learning data in response to the third cropped image data; The jaundice analysis system according to claim 5.

7. the data processing module performs a glare removal process on the first image data to generate glare-removed image data, and generates the first learning data in response to the glare-removed image data; The jaundice analysis system according to claim 1 .

8. the data processing module generates second image data in response to the first image data, generates second training data in response to the second image data, correlates the second training data with second category data, and stores the second training data in the database; and the deep learning module acquires the trained convolutional neural network module by training the target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data. The jaundice analysis system according to claim 1 .

9. the data processing module performs one of an image conversion operation, an image rotation operation, and an image flip operation on the first image data to generate the second image data. The jaundice analysis system according to claim 8.

10. A mobile terminal for storing the input image data is further provided. The processing device comprises: and a communication module communicatively connected to the mobile terminal and the trained convolutional neural network module of the processing device, the communication module adapted to receive the input image data from the mobile terminal and to transmit the test data to the mobile terminal. The jaundice analysis system according to claim 1 .

11. A jaundice analysis method applicable to a jaundice analysis system, the jaundice analysis method comprising: A data processing module of the jaundice analysis system generates first learning data in response to first image data, and the data processing module correlates the first learning data with first category data indicating a bilirubin concentration range of the first image data; A deep learning module of the jaundice analysis system learns a target convolutional neural network module using the first learning data correlated with the first category data, thereby obtaining a trained convolutional neural network module; the trained convolutional neural network module of the jaundice analysis system generating test data in response to input image data, the input image data including a second sclera image of the subject; Including, the first image data includes a first sclera image; the test data indicating a bilirubin concentration range for the subject; Jaundice analysis method.

12. The deep learning module obtains the trained convolutional neural network module by transfer learning. The method for analyzing jaundice according to claim 11.

13. The step of generating the first learning data in response to the first image data includes: The data processing module further includes performing a first cropping process on the first image data to generate first cropped image data; The data processing module generates the first learning data in response to the first cropped image data. The method for analyzing jaundice according to claim 11.

14. The step of generating the first learning data in response to the first image data includes: The data processing module further includes performing a mirroring process on the first image data to generate mirrored image data; The data processing module generates the first learning data in response to the mirroring image data. The method for analyzing jaundice according to claim 11.

15. The step of generating the first learning data in response to the first image data includes: The data processing module further includes a step of performing a second cropping process on the mirrored image data to generate second cropped image data; The data processing module generates the first learning data in response to the second cropped image data; The second cropped image data has a specific image shape. The method for analyzing jaundice according to claim 14.

16. The step of generating the first learning data in response to the first image data includes: The data processing module further includes a step of performing a third cropping process on the second cropped image data to generate third cropped image data; The data processing module generates the first learning data in response to the third cropped image data. The method for analyzing jaundice according to claim 15.

17. The step of generating the first learning data in response to the first image data includes: The data processing module further includes a step of performing a glare removal process on the first image data to generate glare removed image data; The data processing module generates the first learning data in response to the glare removed image data. The method for analyzing jaundice according to claim 11.

18. the data processing module generating second image data in response to the first image data; the data processing module generating second training data in response to the second image data, the data processing module correlating the second training data with second category data; Further comprising: The deep learning module learns the target convolutional neural network module using the first training data correlated with the first category data and the second training data correlated with the second category data, thereby obtaining the trained convolutional neural network module. The method for analyzing jaundice according to claim 11.

19. the data processing module performs one of an image conversion operation, an image rotation operation, and an image flip operation on the first image data to generate the second image data. The method for analyzing jaundice according to claim 18.

20. a communication module of the jaundice analysis system receiving the input image data from a mobile terminal; the communication module transmitting the test data to the mobile terminal; Further comprising: The method for analyzing jaundice according to claim 11.

Images