Device and method for processing medical image based on neural network

Abstract

The apparatus for processing a medical image based on a neural network according to the present invention, which recognizes a small intestine region from a medical image acquired by a capsule endoscope, comprises: a memory loaded with an organ recognition algorithm for performing organ recognition from the medical image; and a processor for applying the medical image to an organ recognition algorithm such that the small intestine region is recognized. The organ recognition algorithm includes: a convolutional neural network algorithm for recognizing organs contained in the medical image as stomach, small intestine, and large intestine to recognize small intestine regions therefrom; and a temporal filtering algorithm associated with the convolutional neural network algorithm to reduce images that may be misrecognized by the convolutional neural network algorithm.

Description

Technical Field

[0001] The present disclosure relates to a device and method for processing medical images based on a neural network, and more particularly, to a device and method for processing medical images based on a neural network, which classifies organs in internal medical images acquired by a capsule endoscope. Background Art

[0002] In general, capsule endoscopy is used to diagnose various small intestinal diseases. In the diagnosis of small intestinal diseases, organ classification is required to distinguish the small intestinal region in the internal medical image obtained by the capsule endoscopy.

[0003] This organ classification technology for medical images has been disclosed in Korean Patent No. 10-2237198 (the invention name is "AI-based interpretation service system for medical images", published on April 1, 2021). In this technology, organ classification is performed based on a classification model for a medical image including multiple organs to be read.

[0004] On the other hand, a capsule endoscope travels in the body for 8 to 12 hours, taking dozens of still images per second to obtain internal medical images. In the case of the digestive tract, more than about 50,000 still images are taken. Therefore, clinicians need to read a large number of still images when diagnosing small intestinal diseases.

[0005] However, reading a large number of still images is tedious, so errors may occur in the diagnosis results. Therefore, research and development are being conducted to distinguish small intestinal regions using artificial intelligence and neural network algorithms. Most related technologies are used to distinguish organs in a single image, so it is difficult to find out the frame in which the organ changes in the image. Nevertheless, the actual clinical environment uses images to make diagnoses. Therefore, the problem arises that small intestinal regions classified based on a single image are not very useful in the diagnosis of small intestinal diseases. Summary of the invention

[0006] Technical issues

[0007] One aspect of the present disclosure is to provide an apparatus and method for processing medical images based on a neural network, wherein an organ change frame is predicted from an image, thereby selectively acquiring an image of a small intestine region.

[0008] Technical Solution

[0009] According to an embodiment of the present disclosure, a device for processing medical images based on a neural network classifies a small intestine region from a medical image acquired by a capsule endoscope, the device comprising: a memory loaded with an organ classification algorithm for performing organ classification on the medical image; and a processor configured to classify the small intestine region by applying the medical image to the organ classification algorithm, the organ classification algorithm comprising: a convolutional neural network algorithm configured to distinguish the small intestine region by classifying organs contained in the medical image into a stomach, a small intestine, and a colon; and a temporal filtering algorithm associated with the convolutional neural network algorithm and configured to reduce images misclassified by the convolutional neural network algorithm.

[0010] The convolutional neural network algorithm may include a ResNet model trained with 2D images, and the temporal filtering algorithm may include a hybrid temporal filter having a Savitzky-Golay filter and a median filter.

[0011] In the process of training the convolutional neural network algorithm, the stomach, small intestine and colon can be labeled into three categories by reading multiple 2D images, and a training set can be established using the labeled multiple 2D images. The convolutional neural network algorithm can be trained based on the training set, so that the convolutional neural network algorithm can predict the small intestine area.

[0012] In the process of training the convolutional neural network algorithm, after labeling, the labeled multiple 2D images can be classified into a training set, a validation set, and a test set by random selection, and each of the classified sets can include normal data and abnormal data.

[0013] During the training of the convolutional neural network algorithm, the 2D image ratio of the stomach, small intestine, and colon can be adjusted to 1:2:1 to adjust the imbalance between the organs.

[0014] In the process of adjusting the ratio of the 2D images, the normal stomach image and the abnormal stomach image may be enhanced by applying horizontal and vertical flipping thereto, and the normal small intestine image, the abnormal small intestine image, the normal colon image, and the abnormal colon image may be downsampled at a preset ratio.

[0015] Downsampling may include: downsampling the normal small intestine image and the normal colon image at ratios of 2 / 3 and 1 / 3, respectively; and downsampling the abnormal small intestine image and the abnormal colon image at ratios of 3 / 4 and 3 / 7, respectively.

[0016] In the process of training the convolutional neural network algorithm, the convolutional neural network algorithm can be verified based on the verification set after training the convolutional neural network algorithm.

[0017] The device for processing medical images based on a neural network can test the convolutional neural network algorithm and the time filtering algorithm based on a test set after training the convolutional neural network algorithm.

[0018] In the process of applying the temporal filtering algorithm, a binary classification is performed.

[0019] The binary classification may include: applying the small intestine class obtained by the convolutional neural network algorithm to a Savitzky-Golay filter and a median filter; and distinguishing the frames of the small intestine from the frames of the stomach and colon by mapping values ​​greater than 1 obtained by adding the result values ​​of the Savitzky-Golay filter and the median filter and dividing by two to 1 and mapping the obtained values ​​less than 0 to 0.

[0020] The organ classification algorithm can predict organ change frames from medical images to classify the small intestine region.

[0021] The organ classification algorithm can be configured to: apply the small intestine class obtained by the convolutional neural network algorithm to the time filtering algorithm and specify the probability after time filtering as a threshold; and distinguish the frames of the small intestine from the frames of the stomach and colon based on the threshold.

[0022] The threshold may be 0.87.

[0023] The temporal filtering algorithm can correct the class probabilities of the frames based on the organ probabilities derived from adjacent frames by the convolutional neural network algorithm.

[0024] On the other hand, according to an embodiment of the present disclosure, a method for processing medical images based on a neural network, wherein a small intestine region is classified from a medical image acquired by a capsule endoscope, the method comprising: inputting the medical image into an organ classification algorithm; and classifying the small intestine region from the medical image through the organ classification algorithm, the organ classification algorithm comprising: a convolutional neural network algorithm, which is configured to distinguish the small intestine region by classifying organs contained in the medical image into a stomach, a small intestine, and a colon; and a temporal filtering algorithm, which is associated with the convolutional neural network algorithm and is configured to reduce images misclassified by the convolutional neural network algorithm.

[0025] Beneficial Effects

[0026] According to the present disclosure, the apparatus and method for processing medical images based on a neural network has the effect of significantly shortening the reading time of clinicians because small intestine images are automatically acquired only through organ classification.

[0027] Furthermore, according to the present disclosure, the apparatus and method for processing medical images based on a neural network can identify the starting and ending areas of the small intestine with high accuracy, thereby having great advantages when combined with other techniques for diagnosing lesions in the small intestine.

[0028] The technical effects of the present disclosure are not limited to the aforementioned effects, and other unexplained technical effects may become apparent to those skilled in the art from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a schematic diagram of an apparatus for processing medical images based on a neural network according to an embodiment,

[0030] Figure 2 is a flowchart schematically illustrating a training method and a testing method of an organ classification algorithm in an apparatus for processing medical images based on a neural network according to an embodiment,

[0031] Figure 3 is a flowchart showing a detection process of an organ classification algorithm in an apparatus for processing medical images based on a neural network according to an embodiment,

[0032] Figure 4 is a conceptual diagram showing the result of an organ classification algorithm in an apparatus for processing a medical image based on a neural network according to an embodiment,

[0033] Figure 5 4 shows data obtained by analyzing an organ classification algorithm using a gradient weighted class activation map in a device for processing medical images based on a neural network according to an embodiment,

[0034] Figure 6 showing errors in the transition from the stomach to the small intestine and in the transition from the small intestine to the colon in the organ classification algorithm of the apparatus for processing medical images based on a neural network according to an embodiment,

[0035] Figure 7 is a flowchart illustrating a method for processing a medical image based on a neural network according to an embodiment. DETAILED DESCRIPTION

[0036] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. However, the present disclosure is not limited to the embodiments set forth herein, but may be implemented in various ways. These embodiments are provided only to complete the present disclosure and enable those of ordinary skill in the art to understand the scope of the present disclosure. In the accompanying drawings, the shape of the elements, etc. may be exaggerated for clear description, and the same reference numerals refer to the same elements.

[0037] Figure 1 is a schematic diagram of an apparatus for processing medical images based on a neural network according to an embodiment.

[0038] like Figure 1As shown, the apparatus 100 for processing medical images based on a neural network according to an embodiment (hereinafter referred to as a "processing apparatus") can predict organ change frames from a medical image 11 taken by a wireless capsule endoscope 10, thereby providing only a small intestine image 12 to be analyzed to a reading doctor 30. In this case, the medical image 10 may include an image provided in real time from the wireless capsule endoscope 10, or a pre-stored image provided from a database (not shown).

[0039] The processing device 100 may classify the medical image 11 into three parts: the stomach, the small intestine, and the colon, and provide only the small intestine image 12 to the reading doctor 30. Conventionally, medical images may be classified into four parts: the esophagus, the stomach, the small intestine, and the colon. However, when performing organ classification on the medical image 11 in four parts, data imbalance occurs due to an insufficient amount of learning data, thereby reducing the performance of the algorithm. Therefore, the processing device 100 may perform organ classification in three parts excluding the esophagus part. In this case of classifying the organ into three parts, the processing device 100 may find the starting area and the ending area of ​​the small intestine based on the landmarks on the boundary between the stomach and the small intestine and the landmarks on the boundary between the small intestine and the colon.

[0040] To this end, the processing device 100 may include a memory 110 and a processor 120 .

[0041] The memory 110 is loaded with an organ classification algorithm 111 that performs organ classification on the medical image 11. In addition, the processor 120 may selectively extract the small intestine image 12 from the medical image 11 by inputting the medical image 11 to the organ classification algorithm 111.

[0042] First, the organ classification algorithm 111 provided in the processing device 100 can be based on a combination of a 2D convolutional neural network (CNN) algorithm 111a and a temporal filtering algorithm 111b. The organ classification algorithm 111 can be trained with images instead of videos. Below, the method of training and designing the organ classification algorithm 111 will be described in detail with reference to the accompanying drawings.

[0043] Figure 2 is a flowchart schematically illustrating a training method and a testing method of an organ classification algorithm in an apparatus for processing medical images based on a neural network according to an embodiment. In addition, Figure 3 is a flowchart showing a detection process of an organ classification algorithm in an apparatus for processing medical images based on a neural network according to an embodiment, Figure 4 is a conceptual diagram illustrating a result of an organ classification algorithm in an apparatus for processing a medical image based on a neural network according to an embodiment.

[0044] like Figures 2 to 4As shown, according to an embodiment, a wireless capsule endoscope (WCE) image 13 taken by a wireless capsule endoscope 10 can be used to train and design an organ classification algorithm 111. However, this is only for describing the embodiment and there is no limitation on the type of learning data. Here, the WCE image 13 can be provided in a Joint Photographic Experts Group (JPEG) format with a matrix size of 320*320 and 3 frames per second (FPS).

[0045] Then, in order to train and design the organ classification algorithm 111, image labeling is performed for each organ (S210). During the image labeling process, the reading doctor can perform three-classification labeling for the stomach, small intestine, and colon by manually reading all the WEC images 13 to be read. Then, the labeled WEC images are classified into a training set S1, a validation set S2, and a test set S3 (S220). Random selection can be applied to image classification. In this case, each of the sets S1, S2, and S3 can include both normal data and abnormal data.

[0046] Then, in order to train and design the organ classification algorithm 111, the 2D convolutional neural network algorithm 111a is trained with the training set S1 (S230). The 2D convolutional neural network algorithm 111a may refer to an algorithm for predicting the probability of the small intestine region. While using the ResNet50 model as the backbone network, the 2D convolutional neural network algorithm 111a is trained to distinguish between the stomach, small intestine, and colon.

[0047] In this case, in order to adjust the imbalance between organs, data augmentation or downsampling can be performed so that the ratio of stomach, small intestine and colon images can be 1:2:1. For example, the stomach images of normal patients and abnormal patients can be enhanced twice by horizontal and vertical flipping. In addition, the small intestine images and colon images of normal patients can be downsampled at ratios of 2 / 3 and 1 / 3, respectively, and the small intestine images and colon images of abnormal patients can be downsampled at ratios of 3 / 4 and 3 / 7, respectively. The 2D convolutional neural network algorithm 111a can be trained by an adaptive moment estimation (ADAM) optimizer with a learning rate of 0.001 and a cross entropy loss.

[0048] On the other hand, when the training of the 2D convolutional neural network algorithm 111a is completed, the organ classification algorithm 111 classifies the stomach, small intestine, and colon from the medical image 11 of the wireless capsule endoscope 10. In other words, the organ classification algorithm 111 can predict the probability of the boundary between the stomach, small intestine, and colon to distinguish the small intestine.

[0049] Then, the organ classification algorithm 111 is designed to execute a temporal filtering algorithm 111b. The temporal filtering algorithm 111b may refer to a hybrid temporal filter that is a combination of a Savitzky-Golay filter and a median filter. Therefore, the temporal filtering algorithm 111b uses only the probability of the small intestine to distinguish the boundary between the small intestine and the stomach and the boundary between the small intestine and the colon by thresholding.

[0050] In addition, the temporal filtering algorithm 111b can perform binary classification. The temporal filtering algorithm 111b can add the result values ​​of the Savitzky-Golay filter and the median filter, divide it by two, map the obtained value greater than 1 to 1, and map the obtained value less than 0 to 0. Here, when the maximum value is less than 1, the values ​​can be divided by the maximum value. For example, the filtering range can be set to 1001 frames, and the small intestine, stomach, and colon can be classified based on a threshold value of 0.87 after applying the temporal filtering algorithm 111b. Here, the minimum index of the frame predicting the small intestine can be identified as the starting area of ​​the small intestine, and the maximum index can be identified as the ending area of ​​the small intestine.

[0051] In this way, the organ classification algorithm 111 can detect transition points between organs based on the neural network algorithm to which the temporal filtering algorithm 111b is applied. In other words, the organ classification algorithm 111 classifies the image into a stomach image, a small intestine image, and a colon image through the 2D convolutional neural network algorithm 111a. Then, the organ classification algorithm 111 corrects the class probability of the frame based on the organ probability derived from the adjacent frames through the temporal filtering algorithm 111b, thereby significantly reducing the number of frames misclassified by the 2D convolutional neural network algorithm 111a.

[0052] Furthermore, the temporally filtered probability of the small intestine may be assigned as a threshold to distinguish between frames of the small intestine mapped as 1 and frames of the stomach and colon mapped as 0. Thus, the organ classification algorithm 111 may detect transition points of boundaries between organs that depend on each other in the image frames (e.g., the boundary between the stomach and the small intestine and the boundary between the small intestine and the colon).

[0053] In this manner, once the organ classification algorithm 111 is fully trained and designed, the organ classification algorithm 111 may be tested based on the test set S3 ( S240 ).

[0054] For example, during the testing of the organ classification algorithm 111, a gradient weighted class activation map (hereinafter referred to as "Grad-CAM") as an interpretable model can be applied to the ResNet50 model. Grad-CAM can be extracted from the feature map of the predicted class, and then adjusted to the image size of the wireless capsule endoscope 10, i.e., 320*320, and superimposed on the original image. Therefore, during the testing process, the performance of the trained and designed organ classification algorithm 111 can be quantitatively analyzed in terms of accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPN).

[0055] Hereinafter, the result of three-classification in the application of the organ classification algorithm according to the embodiment will be described in detail with reference to the accompanying drawings.

[0056] Figure 5 4 shows data obtained by analyzing an organ classification algorithm using a gradient weighted class activation map in a device for processing medical images based on a neural network according to an embodiment, Figure 6 Errors in the transition from the stomach to the small intestine and in the transition from the small intestine to the colon in the organ classification algorithm of the apparatus for processing medical images based on a neural network according to the embodiment are shown.

[0057] like Figure 5 and Figure 6 As shown, as a result of analyzing the organ classification algorithm 111 according to the present embodiment using Grad-CAM, Figure 5 The color map in represents the normalized predictions with red and blue referring to 1 and 0 respectively.

[0058] here, Figure 5 (a) and Figure 5 The comparison between (b) shows that the result of applying the organ classification algorithm 111 according to the embodiment is close to the result of the organ classification process of the endoscopist. In other words, it is confirmed that the trained and designed organ classification algorithm 111 classifies organs by structural information such as dark areas and mucosal vascular patterns captured along the fold direction and the trajectory direction.

[0059] In addition, it was confirmed that the organ classification algorithm 111 according to the embodiment has higher performance than the related art.

[0060] [Table 1]

[0061]

[0062]

[0063] As shown in Table 1 above, compared with the prior art, the organ classification algorithm 111 according to the embodiment shows high performance based on the combination of the 2D convolutional neural network algorithm 111a based on the ResNet50 model and the temporal filtering algorithm 111b.

[0064] Furthermore, as a result of analyzing randomly selected cases in the test set S3, it was confirmed that the temporal filtering algorithm 111b had a strong effect on distinguishing the small intestine to be analyzed. In particular, it was confirmed that after applying the temporal filtering algorithm 111b, the images misclassified by the 2D convolutional neural network algorithm 111a were significantly reduced. In addition, the problem of many misclassified frames in the colon region was solved by deriving an appropriate threshold of 0.87, thereby enabling the small intestine and colon to be distinguished.

[0065] Furthermore, the time error for each case was calculated as frame error in FPS, and the transition error between the stomach and small intestine was only 38.8 ± 25.8 seconds, and the transition error between the small intestine and colon was only 32.0 ± 19.1 seconds. Therefore, the transition time error of the organ classification algorithm 111 was very low.

[0066] On the other hand, a method of processing a medical image to be processed will be described in detail with reference to the accompanying drawings.

[0067] Figure 7 is a flowchart illustrating a method for processing a medical image based on a neural network according to an embodiment.

[0068] like Figure 7 As shown, in the image processing method according to the embodiment, an organ classification algorithm 111 is loaded into the memory 110 .

[0069] Then, when the medical image 11 to be processed is provided from the outside to the processing device 100, the processor 120 can perform organ classification by applying the medical image 11 to be processed to the organ classification algorithm 111. Therefore, the organ classification algorithm 111 can automatically classify and provide only the small intestine frame to the reading doctor 30.

[0070] In more detail, the processor 120 inputs the medical image 11 to be processed into the organ classification algorithm 111 (S710). Therefore, the organ classification algorithm 111 can input the medical image 11 to be processed into the 2D convolutional neural network algorithm, thereby classifying the stomach, small intestine and colon based on the ResNet50 model.

[0071] Furthermore, the processor 120 applies the classified small intestine class to the temporal filtering algorithm 111 using a Savitzky-Golay filter and a median filter (S720). Therefore, the boundary between the stomach and the small intestine and the boundary between the small intestine and the colon can be distinguished by thresholding.

[0072] Then, the processor 120 may provide the small intestine image 12 classified and analyzed by the organ classification algorithm 111 to the reading doctor 30 .

[0073] However, according to an embodiment, the processor 120 directly transmits the small intestine image 13 classified by the organ classification algorithm 111 to the reading doctor 30. However, this is only for describing the embodiment, and the organ classification algorithm 111 may be associated with another algorithm for subsequent processing (e.g., for identifying a lesion area), so that a lesion can be detected in the small intestine.

[0074] In this way, in the apparatus and method for processing medical images based on a neural network according to the present disclosure, only the small intestine image is automatically acquired through organ classification, thereby having the effect of significantly shortening the reading time of the clinician.

[0075] Furthermore, in the apparatus and method for processing medical images based on a neural network according to the present disclosure, the starting and ending areas of the small intestine are identified with high accuracy, thus having great advantages when combined with other techniques for diagnosing lesions in the small intestine.

[0076] Although several embodiments of the present disclosure have been described above and illustrated in the accompanying drawings, the embodiments should not be interpreted as limiting the technical spirit of the present disclosure. The scope of the present disclosure is limited only by the subject matter disclosed in the attached claims, and the technical spirit of the present disclosure can be modified and changed in various forms by ordinary technicians in the field. Therefore, such modifications and changes that are obvious to those skilled in the art will fall within the scope of the present disclosure.

Claims

1. A device for processing medical images based on a neural network, for classifying a small intestine region from a medical image acquired by a capsule endoscope, the device for processing medical images based on a neural network comprising: A memory loaded with an organ classification algorithm for classifying organs of the medical image; as well as a processor configured to classify the small intestine region by applying the medical image to the organ classification algorithm, The organ classification algorithm includes: a convolutional neural network algorithm configured to distinguish the small intestine region by classifying organs contained in the medical image into a stomach, a small intestine, and a colon; as well as A temporal filtering algorithm is associated with the convolutional neural network algorithm and is configured to reduce images misclassified by the convolutional neural network algorithm.

2. The apparatus for processing medical images based on a neural network according to claim 1, wherein: The convolutional neural network algorithm includes a ResNet model trained with 2D images, and The temporal filtering algorithm comprises a hybrid temporal filter having a Savitzky-Golay filter and a median filter.

3. The apparatus for processing medical images based on a neural network according to claim 2, wherein: In the process of training the convolutional neural network algorithm, By reading multiple 2D images, the stomach, small intestine, and colon are labeled in three categories, and A training set is established using the multiple labeled 2D images, and the convolutional neural network algorithm is trained based on the training set, so that the convolutional neural network algorithm can predict the small intestine area.

4. The apparatus for processing medical images based on a neural network according to claim 3, wherein: In the process of training the convolutional neural network algorithm, After the labeling is performed, the labeled 2D images are classified into the training set, the validation set, and the test set by random selection, and Each of the classified sets includes normal data and abnormal data.

5. The apparatus for processing medical images based on a neural network according to claim 4, wherein: During the training of the convolutional neural network algorithm, the ratio of the 2D images of the stomach, small intestine, and colon was adjusted to 1:2:1 to adjust the imbalance between the organs.

6. The apparatus for processing medical images based on a neural network according to claim 5, wherein: In the process of adjusting the ratio of the 2D image, the normal stomach image and the abnormal stomach image are enhanced by applying horizontal and vertical flipping thereto, and The normal small intestine image, the abnormal small intestine image, the normal colon image, and the abnormal colon image are downsampled at a preset ratio.

7. The apparatus for processing medical images based on a neural network according to claim 6, wherein: The down sampling includes: downsampling the normal small intestine image and the normal colon image at ratios of 2 / 3 and 1 / 3, respectively; and The abnormal small intestine image and the abnormal colon image are downsampled at ratios of 3 / 4 and 3 / 7, respectively.

8. The apparatus for processing medical images based on a neural network according to claim 4, wherein: In the process of training the convolutional neural network algorithm, after training the convolutional neural network algorithm, the convolutional neural network algorithm is verified based on the verification set.

9. The apparatus for processing medical images based on a neural network according to claim 4, wherein: After training the convolutional neural network algorithm, the convolutional neural network algorithm and the time filtering algorithm are tested based on the test set.

10. The apparatus for processing medical images based on a neural network according to claim 2, wherein: During application of the temporal filtering algorithm, a binary classification is performed.

11. The apparatus for processing medical images based on a neural network according to claim 10, wherein: The two categories include: Applying the small intestine class obtained by the convolutional neural network algorithm to the Savitzky-Golay filter and the median filter; and The frames of the small intestine are distinguished from the frames of the stomach and colon by mapping a value greater than 1 obtained by adding the result values ​​of the Savitzky-Golay filter and the median filter and dividing by two to 1 and mapping the obtained value less than 0 to 0.

12. The apparatus for processing medical images based on a neural network according to claim 1, wherein: The organ classification algorithm predicts an organ change frame from the medical image to classify the small intestine region.

13. The apparatus for processing medical images based on a neural network according to claim 1, wherein: The organ classification algorithm is configured to: Applying the small intestine class obtained by the convolutional neural network algorithm to the time filtering algorithm, and specifying the probability after time filtering as a threshold; and Frames of the small intestine are distinguished from frames of the stomach and colon based on the threshold.

14. The apparatus for processing medical images based on a neural network according to claim 13, wherein: The threshold is 0.

87.

15. The apparatus for processing medical images based on a neural network according to claim 1, wherein: The temporal filtering algorithm corrects the class probabilities of frames based on organ probabilities derived from adjacent frames by the convolutional neural network algorithm.

16. A method for processing medical images based on a neural network, wherein: A small intestine region is classified from a medical image acquired by a capsule endoscope, and the method for processing the medical image based on a neural network comprises: Inputting medical images into organ classification algorithms; and Classifying the small intestine region from the medical image by using the organ classification algorithm, The organ classification algorithm includes: a convolutional neural network algorithm configured to distinguish the small intestine region by classifying organs contained in the medical image into a stomach, a small intestine, and a colon; and A temporal filtering algorithm is associated with the convolutional neural network algorithm and is configured to reduce images misclassified by the convolutional neural network algorithm.

Images