Method and computing system for training an artificial neural network to detect prostate cancer from TURP pathology images

By employing primary and secondary learning with annotated prostate cancer images and non-cancerous TURP images, the method effectively detects prostate cancer lesions in TURP images, overcoming scale and tissue misdiagnosis issues.

JP7593692B2Active Publication Date: 2024-12-03DEEP BIO
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023551750
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2021-03-08
Filing Date
2022-03-07
Publication Date
2024-12-03
Estimated Expiration
2042-03-07

AI Technical Summary

Technical Problem

Training a machine learning model to detect prostate cancer from TURP pathological images is challenging due to the large scale of the images, the small size of cancer lesions, and the presence of non-prostate tissues or cauterized tissues that can lead to misdiagnosis.

Method used

A method involving primary and secondary learning using prostate needle biopsy and TURP images, where primary learning uses annotated prostate cancer images and secondary learning includes non-prostate or cauterized tissue regions without cancer lesions, employing neural networks like U-Net, DeepLabv3+, or Mask R-CNN.

Benefits of technology

The method enables effective detection of prostate cancer lesions in TURP images with high sensitivity and specificity, addressing the challenges of image scale and tissue composition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007593692000001
    Figure 0007593692000001
  • Figure 0007593692000002
    Figure 0007593692000002
  • Figure 0007593692000003
    Figure 0007593692000003
Patent Text Reader

Abstract

A method and a computing system for training an artificial neural network that can effectively detect prostate cancer lesions in TURP pathology images by taking into account tissue morphology that appears characteristically in the TURP pathology images are disclosed. According to one aspect of the present invention, there is provided a method for training an artificial neural network for detecting prostate cancer from pathological images of TURP, the method comprising: a step of a neural network training system acquiring a plurality of primary training pathological images (wherein each of the plurality of primary training pathological images is a pathological image of a prostate needle biopsy or a pathological image of a prostate resection); a step of the neural network training system performing a primary training of an artificial neural network for detecting prostate cancer from a pathological image using the plurality of primary training pathological images (wherein the artificial neural network for detecting prostate cancer from a pathological image); a step of the neural network training system acquiring a plurality of TURP pathological images; and a step of the neural network training system performing a secondary training of the primarily trained artificial neural network using the plurality of TURP pathological images, wherein each of the plurality of TURP pathological images necessarily includes at least one of a non-prostate tissue region or an ablated prostate tissue region, and does not include any lesion region of prostate cancer.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a method for training an artificial neural network for detecting prostate cancer from TURP pathology images, and a computing system for performing the same, and more particularly to a method for training an artificial neural network that can effectively detect prostate cancer lesions in TURP pathology images, taking into account tissue morphology that appears characteristically in the TURP pathology images, and a computing system for performing the same. [Background technology]

[0002] One of the main tasks of pathology or pathology department is to read slide images of patient's biological tissues and make a diagnosis to determine the condition or symptoms of a particular disease. In the conventional pathology diagnosis method, a pathologist visually observes and reads diagnostic pathology slide images made from a specimen through an optical microscope. The beginning of digital pathology is the method of converting pathology slides into digital images using a microscope camera connected to a computer, then observing and reading them on a monitor. Recently, with the emergence of digital slide scanners, the method of converting the entire pathology slide into one digital image to make a pathology slide image, then observing and reading it on a computer monitor has become widespread.

[0003] Recently, with the development of machine learning, there have been many attempts to automate tasks such as image recognition or classification using computing systems. In particular, attempts have been made to automate diagnoses that were previously performed by skilled medical personnel using neural networks (e.g., deep learning methods using Convolution Neural Networks (CNNs)), which are a type of machine learning. In particular, diagnosis using deep learning using neural networks (e.g., CNNs) does not simply automate the experience and knowledge of traditionally skilled medical personnel, but rather finds characteristic elements through self-learning and derives the desired answer, and may even find features of disease factors from images that are unknown to skilled medical personnel.

[0004] In general, disease diagnosis through a neural network using biological images uses fragments of biological images (e.g., slide images of biological tissue), i.e., patches (also called tiles). That is, for each patch, a skilled medical professional annotates the state of a specific disease (e.g., the presence or absence of cancer), and a neural network is trained using a plurality of such annotated patches as training data. In this case, a convolutional neural network can be used as the neural network.

[0005] On the other hand, transurethral resection of the prostate (TURP) is often performed to treat benign diseases such as benign prostatic hyperplasia, and it is necessary to pathologically determine the presence or absence of prostate cancer in the specimen collected during TURP.

[0006] It is very difficult to train a machine learning model to detect prostate cancer from TURP pathology images using only TURP pathology image data with annotated prostate cancer lesion areas for the following reasons:

[0007] First, the size of the pathological images of TURP is very large. On average, N (N is an integer equal to or greater than 2) 2 cm x 3 cm glass slides are prepared for each patient, which are scanned at 400x magnification to create digital images, resulting in an image of size N x 80,000 x 120,000 pixels. Second, the size of the prostate cancer lesion area in the pathological images of TURP is very small. The rate at which prostate cancer is found in TURP patients is less than 20%, and even when prostate cancer is found, it is often found only in one local area of ​​the pathological image. In other words, while the size and capacity of the training data are very large, the ratio of the area labeled with prostate cancer to the entire tissue is very small, so there is a possibility that a problem will arise in that a very large amount of data is required to train a machine learning model with sufficient sensitivity using only TURP pathological image data.

[0008] In addition, using specimen pathology image data collected by prostate needle biopsy or radical prostatectomy, on which the lesion area of ​​prostate cancer is annotated, to train a machine learning model for detecting prostate cancer from a pathology image of TURP may be problematic for the following reasons. Due to the characteristics of TURP, tissue (bladder, urothelium, etc.) or cauterized prostate tissue, rather than the prostate, is present, which does not appear in a specimen of prostate needle biopsy or resection. Therefore, when training a machine learning model for detecting prostate cancer from a pathology image of TURP using specimen pathology image data collected by prostate needle biopsy or radical prostatectomy, a problem may occur in which tissue (bladder, urothelium, etc.) or cauterized prostate tissue, rather than the prostate, is misdiagnosed as cancer. Summary of the Invention [Problem to be solved by the invention]

[0009] The technical problem to be solved by the present invention is to provide a method for training a machine learning model that can detect prostate cancer in TURP pathology images with high performance by solving the above problems. More specifically, the technical problem to be solved by the present invention is to provide an efficient method for training a machine learning model that can effectively detect prostate cancer lesions in TURP pathology images by taking into account tissue morphology that appears characteristically in TURP pathology images. [Means for solving the problem]

[0010] According to one aspect of the present invention, the present invention includes a step of a neural network learning system acquiring a plurality of primary learning pathology images (wherein each of the plurality of primary learning pathology images is either a prostate needle biopsy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by a prostate needle biopsy, or a prostatectomy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by a radical prostatectomy), a step of the neural network learning system primarily learning an artificial neural network for prostate cancer determination using the plurality of primary learning pathology images (the artificial neural network for prostate cancer determination is an artificial neural network for detecting prostate cancer from a pathology image), and a step of the neural network learning system learning a plurality of primary learning pathology images (wherein each of the plurality of primary learning pathology images is either a prostate needle biops ...). The present invention provides a method for training an artificial neural network for detecting prostate cancer from a pathology image of a transurethral prostate cancer (TURP), the method comprising: acquiring a plurality of TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by TURP; and a step in which the neural network training system performs a second training on the primarily trained artificial neural network using the plurality of TURP pathology images, the method being characterized in that each of the plurality of TURP pathology images necessarily includes at least one of a non-prostate tissue region or a cauterized prostate tissue region, and does not include any prostate cancer lesion region.

[0011] In one embodiment, the artificial neural network may be one of U-Net, DeepLabv3+, Mask R-CNN, and DenseNet.

[0012] According to another aspect of the present invention, there is provided a method for providing a diagnosis result for a pathology image of a TURP to be diagnosed via an artificial neural network trained by the method for training the artificial neural network, the method comprising: a computing system acquiring a pathology image of the TURP to be diagnosed; and the computing system outputting a prostate cancer detection result diagnosed by the artificial neural network based on the pathology image of the TURP to be diagnosed.

[0013] According to another aspect of the invention there is provided a computer program recorded on a medium for being installed on a data processing device and for carrying out the method described above.

[0014] According to another aspect of the present invention, there is provided a computer readable recording medium having recorded thereon a computer program for performing the method described above.

[0015] According to another aspect of the invention there is provided a neural network training system comprising a processor and a memory storing a computer program which, when executed by the processor, causes the neural network training system to perform the method of training an artificial neural network as described above.

[0016] According to another aspect of the present invention, there is provided a computing system for providing a diagnosis result for a pathology image of a TURP to be diagnosed, the computing system comprising a processor and a memory for storing a computer program, the computer program, when executed by the processor, causing the computing system to perform a method for providing a diagnosis result for the pathology image of the TURP to be diagnosed via an artificial neural network trained by the above-mentioned artificial neural network training method.

[0017] According to another aspect of the present invention, there are provided a method for detecting prostate cancer by a transurethral resection of a prostate, comprising: acquiring a plurality of prostate needle biopsy pathology images, the plurality of prostate resection pathology images being images obtained by scanning slides of pathology specimens collected by a prostate needle biopsy, by a neural network training system; acquiring a plurality of prostatectomy pathology images, the plurality of prostate needle biopsy pathology images being images obtained by scanning slides of pathology specimens collected by a radical prostatectomy, by the neural network training system; primarily training an artificial neural network for prostate cancer detection by the neural network training system using the plurality of prostate needle biopsy pathology images and the plurality of prostatectomy pathology images (the artificial neural network for prostate cancer detection is an artificial neural network for detecting prostate cancer from pathology images); The present invention provides a method for training an artificial neural network for detecting prostate cancer from a pathology image of a transurethral prostate cancer (TURP), the method comprising: acquiring a plurality of TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by TURP; and a step in which the neural network training system performs a second training on the primarily trained artificial neural network using the plurality of TURP pathology images, the plurality of TURP pathology images necessarily including at least one of non-prostate tissue regions or cauterized prostate tissue regions, and including no prostate cancer lesion regions. Effect of the Invention

[0018] According to the technical idea of ​​the present invention, an efficient method and system can be provided for training a machine learning model that can effectively detect prostate cancer lesions in TURP pathology images, taking into account tissue morphology that appears characteristically in TURP pathology images. [Brief description of the drawings]

[0019] In order to more fully understand the drawings referred to in the detailed description of the invention, a brief description of each of the drawings is provided.

[0020] [Figure 1] 1 is a schematic diagram illustrating an environment in which an artificial neural network training method and a method for providing a determination result for a pathological specimen according to the technical concept of the present invention are performed; [Diagram 2] 1 is a flowchart illustrating a neural network training method according to an embodiment of the present invention. [Diagram 3] 1 is an example of a pathology image of a prostate needle biopsy specimen. [Figure 4a] An example includes an area of ​​ablated prostate tissue. [Figure 4b] FIG. 13 is a close-up of the ablated prostate tissue area. [Figure 5a] This is an example where non-prostatic tissue areas are included. [Figure 5b] FIG. 13 is a magnified view of the non-prostatic tissue area. [Figure 6] FIG. 13 is a diagram illustrating an example of a method for providing a determination result according to an embodiment of the present invention. [Figure 7] FIG. 1 is a diagram showing a schematic configuration of an artificial neural network training system according to an embodiment of the present invention. [Figure 8] 1 is a diagram showing a schematic configuration of a determination result providing system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Since the present invention can be modified in various ways and can have various embodiments, a specific embodiment is shown in the drawings and described in detail. However, it is not intended to limit the present invention to the specific embodiment, and it should be understood that it includes all modifications, equivalents, and alternatives within the spirit and technical scope of the present invention. In the description of the present invention, if it is determined that a specific description of related publicly known technology makes the gist of the present invention unclear, the detailed description will be omitted.

[0022] Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. Terms such as first and second do not indicate a particular order, but are used only to distinguish one component from another.

[0023] The terms used in this application are merely used to describe specific embodiments and are not intended to limit the present invention. The singular expressions include the plural expressions unless otherwise clearly meant in the context.

[0024] As used herein, terms such as "comprise" or "have" are intended to specify the presence of features, numbers, steps, operations, components, parts, or combinations thereof described herein, and are to be understood as not precluding the presence or additional possibility of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0025] Also, in this specification, when a component "transmits" data to another component, it means that the component can transmit data directly to the other component, or can transmit data to the other component via at least one other component. Conversely, when a component "transmits" data to another component, it means that the component also transmits data to the other component without via the other component.

[0026] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described in detail with reference to the accompanying drawings, in which the same reference numerals indicate the same elements.

[0027] FIG. 1 is a schematic diagram showing an environment in which an artificial neural network training method and a method for providing a determination result for a pathological specimen according to the technical concept of the present invention are carried out.

[0028] Referring to FIG. 1, an artificial neural network training method according to one embodiment of the present invention can be performed by a neural network training system 100, and a method for providing a determination result for a pathological specimen according to one embodiment of the present invention can be performed by a system for providing a determination result for a pathological specimen 200 (hereinafter referred to as the "determination result providing system").

[0029] The artificial neural network training system 100 can train an artificial neural network 300. The artificial neural network 300 can be a neural network for providing diagnostic information for pathological specimens obtained by transurethral resection of the prostate. The pathological specimens can be biopsies obtained from various organs of the human body and biological tissues removed by surgery.

[0030] In particular, the artificial neural network 300 may be an artificial neural network for inputting a pathological image of TURP and detecting prostate cancer from the input pathological image of TURP. The determination result providing system 200 can determine whether or not prostate cancer is detected from the pathological image of TURP using the trained artificial neural network 300. In this specification, the pathological image of TURP may refer to an image or a part thereof obtained by scanning a slide of a pathological specimen collected by transurethral resection of the prostate. The pathological image of TURP may refer to a whole slide image obtained by scanning a pathological slide, and in some embodiments, may refer to patches or tiles obtained by dividing the whole pathological image of TURP into a predetermined unit size.

[0031] In one embodiment, the neural network 300 may be a machine learning model trained to output a probability value for the presence or absence of prostate cancer. The neural network 300 can output a numerical value indicating a judgment result for a target specimen (e.g., the possibility of the presence or absence of a disease), i.e., a probability value, based on data input via an input layer.

[0032] In one embodiment, artificial neural network 300 may be a machine learning model that inputs an entire slide image and determines whether or not a lesion due to prostate cancer is present or detects a lesion area, or in some embodiments, may be a machine learning model that inputs a patch of a slide image and determines whether or not a lesion due to prostate cancer is present or detects a lesion area in the patch.

[0033] As used herein, an artificial neural network is an artificially constructed neural network based on the operating principles of human neurons, includes a multi-layer perceptron model, and may refer to a collection of information representing a set of design items that define the artificial neural network.

[0034] In one embodiment, the artificial neural network 300 may be or include a convolutional neural network that is primarily used for image analysis. In particular, U-Net (Ronneberger 2015), DeepLabv3+ (Chen 2018), Mask R-CNN (He 2017), etc., which are known to be effective for region segmentation, may be used. In cases where a fine region detection function in pixel units is not required, classification models such as ResNet (He 2016) and DenseNet (Huang 2017), which are trained to determine the presence or absence of prostate cancer in small patch image units, may be used.

[0035] The neural network learning system 100 and / or the judgment result providing system 200 may be a computing system, which is a data processing device having computing power for realizing the technical idea of ​​the present invention, and may generally include not only a server, which is a data processing device to which a client can be connected via a network, but also computer devices such as personal computers and mobile terminals.

[0036] The neural network learning system 100 and / or the judgment result providing system 200 may be realized by any physical device, but an average expert in the technical field of the present invention can easily deduce that the neural network learning system 100 and / or the judgment result providing system 200 according to the technical idea of ​​the present invention can be realized by organically combining multiple physical devices as necessary.

[0037] The neural network training system 100 can train the artificial neural network 300 based on training data generated from a plurality of pathological specimens. The neural network training system 100 can generate individual training data using an image obtained by scanning a slide of a pathological specimen or a part (i.e., a patch) of the scanned slide image, and input the individual training data to an input layer of the neural network 300 to train the neural network 300. In this case, the pathological specimen may be a specimen obtained by prostate needle biopsy or prostatectomy, or a specimen obtained by TURP procedure.

[0038] Meanwhile, the trained neural network 300 may be stored in the determination result providing system 200, and the determination result providing system 200 can make a determination on a given TURP specimen to be diagnosed using the trained artificial neural network.

[0039] As shown in Fig. 1, the neural network learning system 100 and / or the judgment result providing system 200 may be realized in the form of a subsystem of a given parent system 10. The parent system 10 may be a server. The server 10 means a data processing device having a computing capability for realizing the technical idea of ​​the present invention, and an average expert in the technical field of the present invention can easily deduce that not only a data processing device to which a client can be connected via a network, but also any device capable of performing a specific service, such as a personal computer, a mobile terminal, etc., can be defined as a server.

[0040] Alternatively, depending on the embodiment, the neural network learning system 100 and the determination result providing system 200 may be realized in a form separate from each other.

[0041] 2 is a flow chart illustrating a neural network training method according to an embodiment of the present invention. An artificial neural network 300 may be trained by the training method of FIG 2, and as described above, the artificial neural network 300 is an artificial neural network for detecting prostate cancer from pathology images.

[0042] The artificial neural network 300 is a neural network that diagnoses prostate cancer using pathological images of TURP, but in the process of training the artificial neural network 300, not only pathological images of TURP but also images of specimens taken by prostate needle biopsy or prostatectomy can be used.

[0043] 2, the neural network learning system 100 can acquire a plurality of primary learning pathology images (S100). Here, each of the plurality of primary learning pathology images may be either a prostate needle biopsy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostate needle biopsy, or a prostatectomy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by radical prostatectomy.

[0044] Meanwhile, each of the multiple primary learning pathology images may be annotated in advance with lesions caused by prostate cancer, and the annotated information may be used as labels for the learning data.

[0045] A prostate needle biopsy is a method of taking a living tissue sample from the prostate using a needle. An example of a pathological image of a prostate needle biopsy sample is shown in Figure 3.

[0046] The neural network training system 100 can perform primary training of the artificial neural network 300 using a plurality of primary training pathology images (S110).

[0047] Since annotated prostate needle biopsy pathology images or annotated prostatectomy pathology images can be relatively easily obtained, the neural network training method primarily trains the artificial neural network 300 using annotated prostate needle biopsy pathology images or annotated prostatectomy pathology images that can be easily obtained. A certain ratio or more of the multiple primary training pathology images may be configured as images containing lesion areas caused by prostate cancer, and in this way, the sensitivity of the artificial neural network 300 can be increased to a certain level or higher.

[0048] Meanwhile, in the case of TURP pathology images, due to the nature of the TURP procedure, there are areas of tissue (bladder, urothelium, etc.) rather than the prostate or areas of ablated prostate tissue, which do not appear in prostate needle biopsy specimens or resection specimens. Therefore, in order to increase the specificity of the primarily trained artificial neural network 300 for TURP pathology images, the following process must be further performed.

[0049] Also, referring to FIG. 2, the neural network learning system 100 can acquire a plurality of TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by transurethral resection of the prostate (S120).

[0050] In this case, each of the multiple TURP pathology images used in the secondary learning must include at least one of non-prostate tissue regions or cauterized prostate tissue regions, and must not include any prostate cancer lesion regions at all. Figures 4a and 5a are diagrams showing examples of TURP pathology images used in the secondary learning. Figure 4a is an example including a cauterized prostate tissue region, and Figure 4b is an enlarged view of the cauterized prostate tissue region. Figure 5a is an example including a non-prostate tissue region, and Figure 5b is an enlarged view of the non-prostate tissue region.

[0051] In one embodiment, some of the multiple TURP pathology images used for secondary learning may be images that include non-prostate tissue areas but no prostate cancer lesion areas, and the remaining images may be images that include ablated prostate tissue areas but no prostate cancer lesion areas.

[0052] In this way, since the multiple TURP pathology images used in the secondary learning should not contain any prostate cancer lesion areas, the multiple TURP pathology images used in the secondary learning can be collectively labeled as normal (benign) and are learning data that do not require annotation of other lesion areas.

[0053] Thereafter, the neural network training system 100 can secondarily train the primarily trained artificial neural network 300 using a plurality of TURP pathology images (S130).

[0054] FIG. 6 is a diagram showing an example of a method for providing a determination result according to an embodiment of the present invention.

[0055] 6, the determination result providing system 200 can acquire a pathological image of a predetermined TURP to be determined (S200). The pathological image of the TURP to be determined is an image obtained by scanning a slide of a predetermined pathological specimen to be determined that is collected by transurethral resection of the prostate (TURP).

[0056] The determination result providing system 200 inputs a pathological image of the TURP to be determined to the artificial neural network 300, and can output a prostate cancer detection result determined by the artificial neural network 300 based on the pathological image of the TURP to be determined (S210).

[0057] FIG. 7 is a diagram showing a schematic configuration of an artificial neural network learning system 100 according to an embodiment of the present invention, and FIG. 8 is a diagram showing a schematic configuration of a determination result providing system 200 according to an embodiment of the present invention.

[0058] The artificial neural network learning system 100 and the judgment result providing system 200 refer to a logical configuration having hardware resources and / or software necessary to realize the technical idea of ​​the present invention, and do not necessarily refer to one physical component or one device. In other words, the artificial neural network learning system 100 and the judgment result providing system 200 refer to a logical combination of hardware and / or software provided to realize the technical idea of ​​the present invention, and may be realized as a set of logical configurations that are installed in devices separated from each other and perform their respective functions to realize the technical idea of ​​the present invention, if necessary. In addition, the artificial neural network learning system 100 and the judgment result providing system 200 may refer to a set of configurations that are realized separately for each function or role to realize the technical idea of ​​the present invention. The components of the artificial neural network learning system 100 and the judgment result providing system 200 may be located in different physical devices or in the same physical device. In addition, depending on the embodiment, the software and / or hardware constituting each of the components of the artificial neural network learning system 100 and the judgment result providing system 200 may be located in different physical devices, and the components located in the different physical devices may be organically combined with each other to realize each module.

[0059] In addition, in this specification, a module may refer to a functional and structural combination of hardware for carrying out the technical idea of ​​the present invention and software for driving the hardware. For example, a module refers to a logical unit of a certain code and a hardware resource for carrying out the certain code, and does not necessarily refer to physically connected code or one type of hardware, as can be easily inferred by an average expert in the technical field of the present invention.

[0060] 7, the artificial neural network training system 100 may include a storage module 110, an acquisition module 120, and a training module 130. It goes without saying that in some embodiments of the present invention, some of the above-mentioned components may not necessarily be essential components for the realization of the present invention, and in some embodiments, the artificial neural network training system 100 may include more components than the above. For example, the artificial neural network training system 100 may further include a communication module (not shown) for communicating with an external device, and a control module (not shown) for controlling the components and resources of the artificial neural network training system 100.

[0061] The storage module 110 is capable of storing the trained artificial neural network 300 .

[0062] The acquisition module 120 can acquire a plurality of primary learning pathology images, each of which may be either a prostate needle biopsy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostate needle biopsy, or a prostatectomy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostatectomy.

[0063] The acquisition module 120 can also acquire multiple TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by transurethral resection of the prostate. In this case, each of the multiple TURP pathology images necessarily includes at least one of a non-prostate tissue region or a cauterized prostate tissue region, and may not include any prostate cancer lesion region.

[0064] The learning module 130 can perform primary training of an artificial neural network for prostate cancer diagnosis using a plurality of primary training pathology images, and can perform secondary training of the primarily trained artificial neural network using a plurality of TURP pathology images.

[0065] 8, the determination result providing system 200 may include a storage module 210, an acquisition module 220, and a determination module 230. Depending on the embodiment of the present invention, some of the above-mentioned components may not necessarily be essential components for realizing the present invention, and it goes without saying that the determination result providing system 200 may include more components than the above components depending on the embodiment. For example, the determination result providing system 200 may further include a communication module (not shown) and a control module (not shown) for controlling the components and resources of the determination result providing system 200.

[0066] The storage module 210 may store the trained artificial neural network 300 .

[0067] The acquisition module 220 can acquire a pathology image of a given target TURP.

[0068] The determination module 230 can input the pathological image of the TURP to be determined to the artificial neural network 300, and output the prostate cancer detection result determined by the artificial neural network 300 based on the pathological image of the TURP to be determined.

[0069] Meanwhile, depending on the embodiment, the artificial neural network training system 100 and the judgment result providing system 200 may include a processor and a memory for storing a program executed by the processor. The processor may include a single-core CPU or a multi-core CPU. The memory may include a high-speed random access memory, and may include a non-volatile memory, such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid-state memory devices. Access to the memory by the processor and other components may be controlled by a memory controller.

[0070] Meanwhile, the method according to the embodiment of the present invention may be realized in the form of computer-readable program instructions and stored in a computer-readable recording medium, and the control program and the object program according to the embodiment of the present invention may also be stored in a computer-readable recording medium. The computer-readable recording medium includes any kind of recording device in which data readable by a computing system is stored.

[0071] The program instructions contained on the recording media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind commonly known and available to those having skill in the software arts.

[0072] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Also, computer-readable recording media can be distributed among computing systems connected via a network to store and execute computer-readable code in a distributed manner.

[0073] Examples of program instructions include not only machine code produced by a compiler, but also higher level language codes that may be executed by a device for processing information electronically, such as a computer, using an interpreter or the like.

[0074] The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

[0075] The above description of the present invention is for illustrative purposes only, and a person having ordinary skill in the art to which the present invention pertains can easily understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. Therefore, it should be understood that the above-described embodiments are illustrative in all respects and are not limiting. For example, each component described in a single form can be implemented in a distributed manner, and similarly, components described as being distributed can be implemented in a combined form.

[0076] The scope of the present invention is indicated by the claims set forth below, rather than by the above detailed description, and all modifications and variations that fall within the meaning and scope of the claims, as well as their equivalent concepts, should be construed as being included within the scope of the present invention. [Industrial Applicability]

[0077] The present invention can be used in a method, and a computing system, for training an artificial neural network to detect prostate cancer from TURP pathology images.

Claims

1. a step in which the neural network learning system acquires a plurality of primary learning pathology images, each of the plurality of primary learning pathology images being either a prostate needle biopsy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostate needle biopsy, or a prostatectomy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostatectomy; a step of the neural network learning system performing primary learning of an artificial neural network for prostate cancer diagnosis using the plurality of primary learning pathology images, the artificial neural network for prostate cancer diagnosis being an artificial neural network for detecting prostate cancer from pathology images; The neural network learning system acquires a plurality of TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by TURP; The neural network training system performs a second training on the first trained artificial neural network using the pathological images of the plurality of TURPs; Including, Each of the plurality of TURP pathology images includes A method for training an artificial neural network for detecting prostate cancer from a TURP pathology image, the image necessarily including at least one of non-prostate tissue areas or ablated prostate tissue areas, and not including any areas of prostate cancer lesions.

2. A method for providing a diagnosis result for a pathological image of a predetermined TURP to be diagnosed through an artificial neural network trained by the method for training an artificial neural network according to claim 1, comprising: A computing system acquires a pathology image of the TURP to be determined; The computing system outputs a prostate cancer detection result determined by the artificial neural network based on a pathological image of the TURP to be determined; A method comprising:

3. A computer program recorded on a medium for being installed in a data processing device and for carrying out the method according to any one of claims 1 or 2.

4. A computer-readable recording medium having a computer program recorded thereon for carrying out the method according to claim 1 or 2.

5. 1. A neural network learning system, comprising: A processor and a memory for storing a computer program, The computer program, when executed by the processor, causes the neural network training system to perform a method for training an artificial neural network; The artificial neural network training method includes: a step of acquiring a plurality of primary learning pathology images, each of the plurality of primary learning pathology images being either a prostate needle biopsy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostate needle biopsy, or a prostatectomy pathology image, which is an image obtained by scanning a slide of a pathology specimen collected by prostatectomy; a step of performing primary training of an artificial neural network for prostate cancer diagnosis using the plurality of primary training pathology images, the artificial neural network for prostate cancer diagnosis being an artificial neural network for detecting prostate cancer from pathology images; The neural network learning system acquires a plurality of TURP pathology images, which are images obtained by scanning slides of pathology specimens collected by TURP; A step of secondarily training the primarily trained artificial neural network using the plurality of TURP pathology images; Including, Each of the plurality of TURP pathology images includes A neural network learning system that necessarily includes at least one of non-prostate tissue regions or ablated prostate tissue regions, and does not include any prostate cancer lesion regions.

6. A computing system for providing a diagnosis result for a pathological image of a predetermined diagnosis target TURP, A processor; A memory for storing a computer program; Including, The computer program, when executed by the processor, causes the computing system to perform a method of providing a diagnosis result for a pathology image of the TURP to be diagnosed through an artificial neural network trained by the artificial neural network training method according to claim 1; The method for providing the determination result includes: Obtaining a pathology image of the TURP to be determined; A step of outputting a prostate cancer detection result determined by the artificial neural network based on a pathological image of the TURP to be determined; A computing system comprising:

Citation Information

Patent Citations

  • Method for image analysis, image analyzer, program, method for manufacturing learned deep learning algorithm, and learned deep learning algorithm

    JP2019148473A

  • Medical image processing apparatus, system and program

    JP2020101860A

  • System and method for two phase diagnosis using neural network

    KR1020190143510A

  • Method for diagnosing and staging prostate cancer

    US6004267A