Cargando…

Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets

Automated glaucoma detection using deep learning may increase the diagnostic rate of glaucoma to prevent blindness, but generalizable models are currently unavailable despite the use of huge training datasets. This study aims to evaluate the performance of a convolutional neural network (CNN) classi...

Descripción completa

Detalles Bibliográficos
Autores principales: Ko, Yu-Chieh, Chen, Wei-Shiang, Chen, Hung-Hsun, Hsu, Tsui-Kang, Chen, Ying-Chi, Liu, Catherine Jui-Ling, Lu, Henry Horng-Shing
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9219722/
https://www.ncbi.nlm.nih.gov/pubmed/35740336
http://dx.doi.org/10.3390/biomedicines10061314
_version_ 1784732186595819520
author Ko, Yu-Chieh
Chen, Wei-Shiang
Chen, Hung-Hsun
Hsu, Tsui-Kang
Chen, Ying-Chi
Liu, Catherine Jui-Ling
Lu, Henry Horng-Shing
author_facet Ko, Yu-Chieh
Chen, Wei-Shiang
Chen, Hung-Hsun
Hsu, Tsui-Kang
Chen, Ying-Chi
Liu, Catherine Jui-Ling
Lu, Henry Horng-Shing
author_sort Ko, Yu-Chieh
collection PubMed
description Automated glaucoma detection using deep learning may increase the diagnostic rate of glaucoma to prevent blindness, but generalizable models are currently unavailable despite the use of huge training datasets. This study aims to evaluate the performance of a convolutional neural network (CNN) classifier trained with a limited number of high-quality fundus images in detecting glaucoma and methods to improve its performance across different datasets. A CNN classifier was constructed using EfficientNet B3 and 944 images collected from one medical center (core model) and externally validated using three datasets. The performance of the core model was compared with (1) the integrated model constructed by using all training images from the four datasets and (2) the dataset-specific model built by fine-tuning the core model with training images from the external datasets. The diagnostic accuracy of the core model was 95.62% but dropped to ranges of 52.5–80.0% on the external datasets. Dataset-specific models exhibited superior diagnostic performance on the external datasets compared to other models, with a diagnostic accuracy of 87.50–92.5%. The findings suggest that dataset-specific tuning of the core CNN classifier effectively improves its applicability across different datasets when increasing training images fails to achieve generalization.
format Online
Article
Text
id pubmed-9219722
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-92197222022-06-24 Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets Ko, Yu-Chieh Chen, Wei-Shiang Chen, Hung-Hsun Hsu, Tsui-Kang Chen, Ying-Chi Liu, Catherine Jui-Ling Lu, Henry Horng-Shing Biomedicines Article Automated glaucoma detection using deep learning may increase the diagnostic rate of glaucoma to prevent blindness, but generalizable models are currently unavailable despite the use of huge training datasets. This study aims to evaluate the performance of a convolutional neural network (CNN) classifier trained with a limited number of high-quality fundus images in detecting glaucoma and methods to improve its performance across different datasets. A CNN classifier was constructed using EfficientNet B3 and 944 images collected from one medical center (core model) and externally validated using three datasets. The performance of the core model was compared with (1) the integrated model constructed by using all training images from the four datasets and (2) the dataset-specific model built by fine-tuning the core model with training images from the external datasets. The diagnostic accuracy of the core model was 95.62% but dropped to ranges of 52.5–80.0% on the external datasets. Dataset-specific models exhibited superior diagnostic performance on the external datasets compared to other models, with a diagnostic accuracy of 87.50–92.5%. The findings suggest that dataset-specific tuning of the core CNN classifier effectively improves its applicability across different datasets when increasing training images fails to achieve generalization. MDPI 2022-06-03 /pmc/articles/PMC9219722/ /pubmed/35740336 http://dx.doi.org/10.3390/biomedicines10061314 Text en © 2022 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Ko, Yu-Chieh
Chen, Wei-Shiang
Chen, Hung-Hsun
Hsu, Tsui-Kang
Chen, Ying-Chi
Liu, Catherine Jui-Ling
Lu, Henry Horng-Shing
Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title_full Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title_fullStr Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title_full_unstemmed Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title_short Widen the Applicability of a Convolutional Neural-Network-Assisted Glaucoma Detection Algorithm of Limited Training Images across Different Datasets
title_sort widen the applicability of a convolutional neural-network-assisted glaucoma detection algorithm of limited training images across different datasets
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9219722/
https://www.ncbi.nlm.nih.gov/pubmed/35740336
http://dx.doi.org/10.3390/biomedicines10061314
work_keys_str_mv AT koyuchieh widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT chenweishiang widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT chenhunghsun widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT hsutsuikang widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT chenyingchi widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT liucatherinejuiling widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets
AT luhenryhorngshing widentheapplicabilityofaconvolutionalneuralnetworkassistedglaucomadetectionalgorithmoflimitedtrainingimagesacrossdifferentdatasets