KR20210104383A - System and method for storing and executing deep learning of medical information - Google Patents

Abstract

According to the present invention, a system for storing and executing deep learning of medical information comprises: an image information acquisition unit for acquiring medical image information of a subject using a medical imaging device; a genome information acquisition unit for acquiring genome information on the subject; a batch storage unit for collectively storing the medical image information and the genome information in a file format capable of collectively storing the medical image information and the genome information; and a deep learning execution unit configuring a deep learning model using the collectively stored medical image information and genome information as input data, and performing a deep learning process on medical information using the configured deep learning model.

Description

System and method for storing and executing deep learning of medical information

It relates to technology for storing medical image information and genomic information for constructing deep learning learning data and performing deep learning.

The general public, who lack professional knowledge in the medical field, have a health problem or want to obtain more detailed information after receiving a diagnosis result from a hospital There are many cases of going to a medical department or visiting a hospital based on Internet information.

However, in this process, if the user fails to enter the appropriate department or disease name due to lack of information, or if the search engine fails to recommend an appropriate hospital or department even though the user's medical information is entered correctly, another department or hospital is searched. And, by looking for it, you will lose time and money. Moreover, if the appropriate time for treatment is missed while repeating this unnecessary process, the patient's health may deteriorate further.

In particular, when special treatment is needed, the general public often do not even know how to treat the disease, where specialized hospitals are, or even which departments to find and treat.

Therefore, there is a need for a medical information recommendation system that provides medical information that can be used easily even by ordinary people who do not have sufficient medical knowledge and that sufficiently reflects the user's needs.

Republic of Korea Patent Publication No. 10-2019-0132290 (published on November 27, 2019)

The problem to be solved by the present invention is to store the medical image information and the genome information collectively, and perform a deep learning process on the medical information using the collectively stored medical image information and the genome information. It relates to a system and method for performing running.

A system for storing and performing deep learning of medical information for solving the above problem includes: an image information acquisition unit for acquiring medical image information of a subject using a medical imaging device; a genomic information acquisition unit for acquiring genomic information about the subject; a batch storage unit for collectively storing the medical image information and the genome information in a file format capable of collectively storing; and a deep learning performing unit that configures a deep learning model using the collectively stored medical image information and the genome information as input data, and performs a deep learning process on medical information using the configured deep learning model. do.

The image information acquisition unit includes any one or more of Chest X-Ray, CT, MRI, PET, ultrasound, angiography, and OCT image information based on DICOM standard images, and endoscope, skin disease, skin cancer, It is characterized in that the medical image information including any one or more of ophthalmic and pathological image information is obtained.

The genomic information acquisition unit is characterized in that it acquires genomic information including any one or more of FASTA/FASTQ format expressing a nucleotide sequence or a protein sequence as a character string, or SAM and BAM information expressing a nucleotide sequence.

The batch storage unit stores the DICOM standard image and the genome information in a DSQ file format (DICOM and SEQUENCE), and stores the JPEG image and the genome information in a JSQ file format (JPEG and SEQUENCE).

The collective storage unit classifies and stores the diagnosis result according to a string code including English and numbers for each disease, stores open genome data or genome data collected from clinical or public health centers, and anonymizes and anonymizes images and genome information collected from patients. Synchronization processing is characterized.

The deep learning performing unit, using the configured deep learning model, is characterized in that the diagnosis of a disease for a target subject is performed using only the medical image information that does not include the genomic information.

A method for storing medical information and performing deep learning for solving the above problems includes: acquiring medical image information of a subject using a medical imaging device; obtaining genomic information about the subject; Composing the medical image information and the genome information in a file format capable of collectively storing the information; and constructing a deep learning model using the collectively stored medical image information and the genome information as input data, and performing a deep learning process on the medical information using the configured deep learning model.

The medical image information includes any one or more of Chest X-Ray, CT, MRI, PET, ultrasound, angiography, and OCT image information based on DICOM standard images, and endoscope, skin disease, and skin cancer based on JPEG images. , ophthalmic, characterized in that it includes any one or more of pathological image information.

The genomic information is characterized in that it is a FASTA/FASTQ format expressing a nucleotide sequence or a protein sequence as a character string, or includes at least one of SAM and BAM information expressing a nucleotide sequence.

The batch storing step stores the DICOM standard image and the genome information in a DSQ file format (DICOM and SEQUENCE), and stores the JPEG image and the genome information in a JSQ file format (JPEG and SEQUENCE). do.

The step of storing the batch includes classifying and storing the diagnosis result according to a string code including English and numbers for each disease,

It stores public genomic data or genomic data collected from clinical or public health centers, and anonymizes and synchronizes images and genomic information collected from patients.

The step of performing the deep learning process is characterized in that, using the configured deep learning model, the disease diagnosis is performed on the target subject using only the medical image information that does not include the genome information.

According to the present invention, by collectively storing the medical image information and the genome information, and performing a deep learning process on the medical information using the collectively stored medical image information and the genome information, medical image and genome data It allows you to periodically update and manage deep learning learning data while saving.

1 is a reference diagram for explaining a system for storing medical information and performing deep learning according to the present invention.
2 is a block diagram illustrating a system for storing medical information and performing deep learning according to the present invention.
3 is a flowchart of an embodiment for explaining a method for storing medical information and performing deep learning according to the present invention.

Hereinafter, with reference to the drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them.

The present invention may be embodied in many different forms and is not limited to the embodiments described herein.

In order to clearly explain the present invention, parts irrelevant to the description are omitted, and the same reference numerals are given to the same or similar elements throughout the specification. In the following, the provision or arrangement of an arbitrary component on the “upper (or lower)” or “top (or below)” of the substrate means that any component is provided or disposed in contact with the upper surface (or lower surface) of the substrate. means that Further, it is not limited to not include other components between the substrate and any components provided or disposed on (or under) the substrate.

1 is a reference diagram for explaining a system for storing medical information and performing deep learning according to the present invention.

Medical information uses medical imaging equipment (HW) to generate and store image data. Using the PACS (SW) program, the image data is classified, stored and managed.

By applying computer vision technology, extracting feature values from medical images, applying image processing technology, processing and standardizing medical images, applying pattern recognition technology, artificial intelligence technology-based The data is analyzed and stored as data in a computerized format of genomic data (Genomics, Bioinformatics). In addition, by applying encryption technology (Encryption), protection of medical data and user access control are performed. Using server (HW) equipment, clinicians and researchers access and view data, etc.

In addition, storage (HW) equipment is used to store and manage a large amount of data. In addition, by applying artificial intelligence (deep learning) technology, it performs diagnostic decision-making assistance based on a large amount of diagnostic information.

2 is a configuration block diagram of an embodiment for explaining a system for storing medical information and performing deep learning according to the present invention.

Referring to FIG. 2 , the system 100 for storing medical information and performing deep learning includes an image information acquisition unit 110 , a genome information acquisition unit 120 , a batch storage unit 130 , and a deep learning execution unit 140 . include

The image information acquisition unit 110 acquires medical image information of a subject by using a medical imaging device. The image information acquisition unit 110 obtains any one or more of Chest X-Ray, CT, MRI, PET, ultrasound (heart, liver, thyroid, blood vessels, etc.), angiography (intervention), and OCT image information based on the DICOM standard. It is possible to obtain medical image information including any one or more of endoscope, skin disease, skin cancer, ophthalmology, and pathology image information based on the JPEG image.

The genome information acquisition unit 120 acquires genome information about the subject. The genomic information acquisition unit 120 may acquire genomic information including any one or more of the FASTA/FASTQ format expressing the nucleotide sequence or the protein sequence as a character string, or SAM and BAM information expressing the nucleotide sequence.

SAM is a text file and can be viewed immediately because it is saved in string format. Since BAM is a binary format, it cannot be viewed immediately. However, because it is compressed, the capacity is small. The BAM file is the same as the SAM file, but Reference Sequence Names and Lengths are included in the header.

The batch storage unit 130 configures and stores the medical image information and the genome information in a file format capable of collectively storing the information.

The batch storage unit 130 integrates the acquired medical image information and the genome information using a Picture Archiving and Communications System (PACS) medical imaging program to classify and store the collected information.

The batch storage unit 130 stores the data structured in a file format capable of collectively storing the image and the genome information. For example, the batch storage unit 130 stores a file including a DICOM standard image and genome sequence information in a DSQ file format (DICOM and SEQUENCE). In addition, the batch storage unit 130 stores a file including a JPEG image and genome sequence information in a JSQ file format (JPEG and SEQUENCE). Such a file format is named MIAS (Medical Image and Sequence).

In addition, the batch storage unit 130 classifies and stores the diagnosis result according to a character string code including English and numbers for each disease. In addition, the batch storage unit 130 stores public genomic data or genomic data collected from clinical or public health centers. In addition, the batch storage unit 130 anonymizes and synchronizes the images and genome information collected from the patient and stores them.

The batch storage unit 130 checks the data additionally generated and stored at a specific time every day. The batch storage unit 130 applies a multi-threading operation to check new data, periodically checks medical images and genome data, and updates (updates) the data.

When storing/reading data, encryption/decryption is processed. The batch storage unit 130 may use the following method as an encryption algorithm.

Block encryption methods: DES, 3-DES, AES, SPEED (KISA), ARIA,

Stream encryption: RC4, A5/1, A5/2, A5/3

Asymmetric key algorithms: RSA, ECC, DSS

Hash Algorithms: MD5, SHA-0, SHA-1, SHA-2 (SHA-224, SHA-256, SHA-384, SHA-512)

The deep learning performing unit 140 configures a deep learning model using the collectively stored medical image information and the genome information as input data, and performs a deep learning process on medical information using the configured deep learning model.

The deep learning performing unit 140 configures a deep learning technology-based diagnostic model reflecting the added data by learning the medical image information and the genomic information in the deep learning training model. Thereafter, the deep learning performing unit 140 may receive image information that does not include genomic information and perform a corresponding disease diagnosis by using the deep learning model.

3 is a flowchart of an embodiment for explaining a method for storing medical information and performing deep learning according to the present invention.

The system for storing medical information and performing deep learning acquires the subject's medical image information using a medical imaging device (step S200).

The medical image information includes any one or more of Chest X-Ray, CT, MRI, PET, ultrasound, angiography, and OCT image information based on DICOM standard images, and endoscope, skin disease, and skin cancer based on JPEG images. , ophthalmology, and any one or more of pathological image information.

After step S200, the system for storing medical information and performing deep learning acquires genomic information for the subject (step S202).

The genomic information includes either a FASTA/FASTQ format expressing a nucleotide sequence or a protein sequence as a character string, or any one or more of SAM and BAM information expressing a nucleotide sequence.

After step S202, the system for storing and performing deep learning of medical information configures and stores the medical image information and the genome information in a file format capable of collectively storing (step S204).

The batch storing step stores the DICOM standard image and the genome information in a DSQ file format (DICOM and SEQUENCE), and stores the JPEG image and the genome information in a JSQ file format (JPEG and SEQUENCE).

In addition, the step of collectively storing the diagnosis result by classifying and storing according to a string code including English and numbers for each disease, storing public genome data or genome data collected from clinical or public health centers, etc., and images and genomes collected from patients Anonymize and synchronize information.

After step S204, the system for storing and performing deep learning of medical information configures a deep learning model using the stored medical image information and the genome information as input data, and deep learning on medical information using the configured deep learning model The process is performed (step S206).

In the performing the deep learning process, a disease diagnosis is performed on a target subject using only the medical image information that does not include the genomic information by using the configured deep learning model.

The present invention can be applied to various playback devices by being implemented as a software program and recorded on a computer-readable recording medium. The various playback devices may be a PC, a notebook computer, a portable terminal, or the like. For example, the recording medium may be a hard disk, flash memory, RAM, ROM, etc. built-in to each playback device, or an optical disk such as a CD-R or CD-RW, compact flash card, smart media, memory stick, or multimedia card as an external type. have.

Although the embodiments of the present invention have been described as described above, the embodiments disclosed in the specification of the present invention do not limit the present invention. The scope of the present invention should be construed by the following claims, and all technologies within the scope equivalent thereto should be construed as being included in the scope of the present invention.

100: system for storing medical information and performing deep learning
110: image information acquisition unit
120: genome information acquisition unit
130: batch storage
140: deep learning execution unit

Claims (12)

an image information acquisition unit for acquiring medical image information of a subject using a medical imaging device;
a genomic information acquisition unit for acquiring genomic information about the subject;
a batch storage unit for collectively storing the medical image information and the genome information in a file format capable of collectively storing; and
Constructing a deep learning model using the collectively stored medical image information and the genome information as input data, and using the configured deep learning model to perform a deep learning process on medical information, characterized in that it comprises a deep learning performing unit A system for storing medical information and performing deep learning. The method according to claim 1,
The image information acquisition unit,
Including any one or more of Chest X-Ray, CT, MRI, PET, Ultrasound, Angiography, OCT image information based on DICOM standard image, and endoscope, skin disease, skin cancer, ophthalmology, and pathology image information based on JPEG image A system for storing and performing deep learning of medical information, characterized in that acquiring medical image information including any one or more. The method according to claim 1,
The genome information acquisition unit,
A system for storing and performing deep learning of medical information, characterized in that the FASTA/FASTQ format in which the nucleotide sequence or the protein sequence is expressed as a character string, or genomic information including any one or more of SAM and BAM information expressing the nucleotide sequence is obtained. 3. The method according to claim 2,
The batch storage unit,
Store the DICOM standard image and the genome information in DSQ file format (DICOM and SEQUENCE),
A system for storing and performing deep learning of medical information, characterized in that the JPEG image and the genome information are stored in a JSQ file format (JPEG and SEQUENCE). The method according to claim 1,
The batch storage unit,
The diagnosis result is classified and stored according to the string code including English and numbers for each disease,
It stores public genomic data or genomic data collected from clinical or public health centers, etc.
A system for storing medical information and performing deep learning, which anonymizes and synchronizes images and genome information collected from patients. The method according to claim 1,
The deep learning performing unit,
Using the configured deep learning model, the system for storing and performing deep learning of medical information, characterized in that the diagnosis of a disease is performed on a subject using only the medical image information that does not include the genome information. obtaining medical image information of a subject using a medical imaging device;
obtaining genomic information about the subject;
Composing the medical image information and the genome information in a file format capable of collectively storing the information; and
Medical information comprising the steps of constructing a deep learning model using the collectively stored medical image information and the genome information as input data, and performing a deep learning process on the medical information using the configured deep learning model of storage and deep learning. 8. The method of claim 7,
The medical image information is
Contains any one or more of Chest X-Ray, CT, MRI, PET, Ultrasound, Angiography, and OCT image information based on DICOM standard image,
A method of storing medical information and performing deep learning, characterized in that it includes any one or more of endoscopy, skin disease, skin cancer, ophthalmology, and pathological image information based on a JPEG image. 8. The method of claim 7,
The genome information is
FASTA/FASTQ format in which the nucleotide sequence or protein sequence is expressed as a string, or
A method for storing and performing deep learning of medical information, characterized in that it includes any one or more of SAM and BAM information representing the nucleotide sequence. 9. The method of claim 8,
The batch storage step includes:
Store the DICOM standard image and the genome information in DSQ file format (DICOM and SEQUENCE),
Method for storing and performing deep learning of medical information, characterized in that the JPEG image and the genome information are stored in a JSQ file format (JPEG and SEQUENCE). 8. The method of claim 7,
The batch storage step includes:
The diagnosis result is classified and stored according to the string code including English and numbers for each disease,
It stores public genomic data or genomic data collected from clinical or public health centers, etc.
A method for storing medical information and performing deep learning, characterized in that anonymizing and synchronizing images and genomic information collected from patients. 8. The method of claim 7,
The step of performing the deep learning process is,
Using the configured deep learning model, the method for storing and performing deep learning of medical information, characterized in that the diagnosis of a disease is performed on a target subject using only the medical image information that does not include the genome information.

Images