General deep learning model for detecting diabetic retinopathy
Abstract Background Doctors can detect symptoms of diabetic retinopathy (DR) early by using retinal ophthalmoscopy, and they can improve diagnostic efficiency with the assistance of deep learning to select treatments and support personnel workflow. Conventionally, most deep learning methods for DR d...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
BMC
2021-11-01
|
Series: | BMC Bioinformatics |
Subjects: | |
Online Access: | https://doi.org/10.1186/s12859-021-04005-x |
_version_ | 1798034873585238016 |
---|---|
author | Ping-Nan Chen Chia-Chiang Lee Chang-Min Liang Shu-I Pao Ke-Hao Huang Ke-Feng Lin |
author_facet | Ping-Nan Chen Chia-Chiang Lee Chang-Min Liang Shu-I Pao Ke-Hao Huang Ke-Feng Lin |
author_sort | Ping-Nan Chen |
collection | DOAJ |
description | Abstract Background Doctors can detect symptoms of diabetic retinopathy (DR) early by using retinal ophthalmoscopy, and they can improve diagnostic efficiency with the assistance of deep learning to select treatments and support personnel workflow. Conventionally, most deep learning methods for DR diagnosis categorize retinal ophthalmoscopy images into training and validation data sets according to the 80/20 rule, and they use the synthetic minority oversampling technique (SMOTE) in data processing (e.g., rotating, scaling, and translating training images) to increase the number of training samples. Oversampling training may lead to overfitting of the training model. Therefore, untrained or unverified images can yield erroneous predictions. Although the accuracy of prediction results is 90%–99%, this overfitting of training data may distort training module variables. Results This study uses a 2-stage training method to solve the overfitting problem. In the training phase, to build the model, the Learning module 1 used to identify the DR and no-DR. The Learning module 2 on SMOTE synthetic datasets to identify the mild-NPDR, moderate NPDR, severe NPDR and proliferative DR classification. These two modules also used early stopping and data dividing methods to reduce overfitting by oversampling. In the test phase, we use the DIARETDB0, DIARETDB1, eOphtha, MESSIDOR, and DRIVE datasets to evaluate the performance of the training network. The prediction accuracy achieved to 85.38%, 84.27%, 85.75%, 86.73%, and 92.5%. Conclusions Based on the experiment, a general deep learning model for detecting DR was developed, and it could be used with all DR databases. We provided a simple method of addressing the imbalance of DR databases, and this method can be used with other medical images. |
first_indexed | 2024-04-11T20:50:24Z |
format | Article |
id | doaj.art-abd51fbc76304968aaa59243518d4b12 |
institution | Directory Open Access Journal |
issn | 1471-2105 |
language | English |
last_indexed | 2024-04-11T20:50:24Z |
publishDate | 2021-11-01 |
publisher | BMC |
record_format | Article |
series | BMC Bioinformatics |
spelling | doaj.art-abd51fbc76304968aaa59243518d4b122022-12-22T04:03:52ZengBMCBMC Bioinformatics1471-21052021-11-0122S511510.1186/s12859-021-04005-xGeneral deep learning model for detecting diabetic retinopathyPing-Nan Chen0Chia-Chiang Lee1Chang-Min Liang2Shu-I Pao3Ke-Hao Huang4Ke-Feng Lin5Department of Biomedical Engineering, National Defense Medical CenterGraduate Institute of Applied Science and Technology, National Taiwan University of Science and TechnologyDepartment of Ophthalmology, Tri-Service General Hospital, National Defense Medical CenterDepartment of Ophthalmology, Tri-Service General Hospital, National Defense Medical CenterDepartment of Ophthalmology, Tri-Service General Hospital, National Defense Medical CenterGraduate Institute of Applied Science and Technology, National Taiwan University of Science and TechnologyAbstract Background Doctors can detect symptoms of diabetic retinopathy (DR) early by using retinal ophthalmoscopy, and they can improve diagnostic efficiency with the assistance of deep learning to select treatments and support personnel workflow. Conventionally, most deep learning methods for DR diagnosis categorize retinal ophthalmoscopy images into training and validation data sets according to the 80/20 rule, and they use the synthetic minority oversampling technique (SMOTE) in data processing (e.g., rotating, scaling, and translating training images) to increase the number of training samples. Oversampling training may lead to overfitting of the training model. Therefore, untrained or unverified images can yield erroneous predictions. Although the accuracy of prediction results is 90%–99%, this overfitting of training data may distort training module variables. Results This study uses a 2-stage training method to solve the overfitting problem. In the training phase, to build the model, the Learning module 1 used to identify the DR and no-DR. The Learning module 2 on SMOTE synthetic datasets to identify the mild-NPDR, moderate NPDR, severe NPDR and proliferative DR classification. These two modules also used early stopping and data dividing methods to reduce overfitting by oversampling. In the test phase, we use the DIARETDB0, DIARETDB1, eOphtha, MESSIDOR, and DRIVE datasets to evaluate the performance of the training network. The prediction accuracy achieved to 85.38%, 84.27%, 85.75%, 86.73%, and 92.5%. Conclusions Based on the experiment, a general deep learning model for detecting DR was developed, and it could be used with all DR databases. We provided a simple method of addressing the imbalance of DR databases, and this method can be used with other medical images.https://doi.org/10.1186/s12859-021-04005-xSMOTEOverfittingDecision treeNasnet-largeTransfer learning |
spellingShingle | Ping-Nan Chen Chia-Chiang Lee Chang-Min Liang Shu-I Pao Ke-Hao Huang Ke-Feng Lin General deep learning model for detecting diabetic retinopathy BMC Bioinformatics SMOTE Overfitting Decision tree Nasnet-large Transfer learning |
title | General deep learning model for detecting diabetic retinopathy |
title_full | General deep learning model for detecting diabetic retinopathy |
title_fullStr | General deep learning model for detecting diabetic retinopathy |
title_full_unstemmed | General deep learning model for detecting diabetic retinopathy |
title_short | General deep learning model for detecting diabetic retinopathy |
title_sort | general deep learning model for detecting diabetic retinopathy |
topic | SMOTE Overfitting Decision tree Nasnet-large Transfer learning |
url | https://doi.org/10.1186/s12859-021-04005-x |
work_keys_str_mv | AT pingnanchen generaldeeplearningmodelfordetectingdiabeticretinopathy AT chiachianglee generaldeeplearningmodelfordetectingdiabeticretinopathy AT changminliang generaldeeplearningmodelfordetectingdiabeticretinopathy AT shuipao generaldeeplearningmodelfordetectingdiabeticretinopathy AT kehaohuang generaldeeplearningmodelfordetectingdiabeticretinopathy AT kefenglin generaldeeplearningmodelfordetectingdiabeticretinopathy |