Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning
Endoscopic resection is recommended for gastric neoplasms confined to mucosa or superficial submucosa. The determination of invasion depth is based on gross morphology assessed in endoscopic images, or on endoscopic ultrasound. These methods have limited accuracy and pose an inter-observer variabili...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-06-01
|
Series: | Journal of Clinical Medicine |
Subjects: | |
Online Access: | https://www.mdpi.com/2077-0383/9/6/1858 |
id |
doaj-509c878a5cb24d0283abcfa13d1fa6fd |
---|---|
record_format |
Article |
spelling |
doaj-509c878a5cb24d0283abcfa13d1fa6fd2020-11-25T03:24:46ZengMDPI AGJournal of Clinical Medicine2077-03832020-06-0191858185810.3390/jcm9061858Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-LearningBum-Joo Cho0Chang Seok Bang1Jae Jun Lee2Chang Won Seo3Ju Han Kim4Medical Artificial Intelligence Center, Hallym University Medical Center, Anyang 14068, KoreaInstitute of New Frontier Research, Hallym University College of Medicine, Chuncheon 24253, KoreaInstitute of New Frontier Research, Hallym University College of Medicine, Chuncheon 24253, KoreaMedical Artificial Intelligence Center, Hallym University Medical Center, Anyang 14068, KoreaDivision of Biomedical Informatics, Seoul National University Biomedical Informatics (SNUBI), Seoul National University College of Medicine, Seoul 03080, KoreaEndoscopic resection is recommended for gastric neoplasms confined to mucosa or superficial submucosa. The determination of invasion depth is based on gross morphology assessed in endoscopic images, or on endoscopic ultrasound. These methods have limited accuracy and pose an inter-observer variability. Several studies developed deep-learning (DL) algorithms classifying invasion depth of gastric cancers. Nevertheless, these algorithms are intended to be used after definite diagnosis of gastric cancers, which is not always feasible in various gastric neoplasms. This study aimed to establish a DL algorithm for accurately predicting submucosal invasion in endoscopic images of gastric neoplasms. Pre-trained convolutional neural network models were fine-tuned with 2899 white-light endoscopic images. The prediction models were subsequently validated with an external dataset of 206 images. In the internal test, the mean area under the curve discriminating submucosal invasion was 0.887 (95% confidence interval: 0.849–0.924) by DenseNet−161 network. In the external test, the mean area under the curve reached 0.887 (0.863–0.910). Clinical simulation showed that 6.7% of patients who underwent gastrectomy in the external test were accurately qualified by the established algorithm for potential endoscopic resection, avoiding unnecessary operation. The established DL algorithm proves useful for the prediction of submucosal invasion in endoscopic images of gastric neoplasms.https://www.mdpi.com/2077-0383/9/6/1858artificial intelligenceconvolutional neural networksendoscopygastric neoplasms |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Bum-Joo Cho Chang Seok Bang Jae Jun Lee Chang Won Seo Ju Han Kim |
spellingShingle |
Bum-Joo Cho Chang Seok Bang Jae Jun Lee Chang Won Seo Ju Han Kim Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning Journal of Clinical Medicine artificial intelligence convolutional neural networks endoscopy gastric neoplasms |
author_facet |
Bum-Joo Cho Chang Seok Bang Jae Jun Lee Chang Won Seo Ju Han Kim |
author_sort |
Bum-Joo Cho |
title |
Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning |
title_short |
Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning |
title_full |
Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning |
title_fullStr |
Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning |
title_full_unstemmed |
Prediction of Submucosal Invasion for Gastric Neoplasms in Endoscopic Images Using Deep-Learning |
title_sort |
prediction of submucosal invasion for gastric neoplasms in endoscopic images using deep-learning |
publisher |
MDPI AG |
series |
Journal of Clinical Medicine |
issn |
2077-0383 |
publishDate |
2020-06-01 |
description |
Endoscopic resection is recommended for gastric neoplasms confined to mucosa or superficial submucosa. The determination of invasion depth is based on gross morphology assessed in endoscopic images, or on endoscopic ultrasound. These methods have limited accuracy and pose an inter-observer variability. Several studies developed deep-learning (DL) algorithms classifying invasion depth of gastric cancers. Nevertheless, these algorithms are intended to be used after definite diagnosis of gastric cancers, which is not always feasible in various gastric neoplasms. This study aimed to establish a DL algorithm for accurately predicting submucosal invasion in endoscopic images of gastric neoplasms. Pre-trained convolutional neural network models were fine-tuned with 2899 white-light endoscopic images. The prediction models were subsequently validated with an external dataset of 206 images. In the internal test, the mean area under the curve discriminating submucosal invasion was 0.887 (95% confidence interval: 0.849–0.924) by DenseNet−161 network. In the external test, the mean area under the curve reached 0.887 (0.863–0.910). Clinical simulation showed that 6.7% of patients who underwent gastrectomy in the external test were accurately qualified by the established algorithm for potential endoscopic resection, avoiding unnecessary operation. The established DL algorithm proves useful for the prediction of submucosal invasion in endoscopic images of gastric neoplasms. |
topic |
artificial intelligence convolutional neural networks endoscopy gastric neoplasms |
url |
https://www.mdpi.com/2077-0383/9/6/1858 |
work_keys_str_mv |
AT bumjoocho predictionofsubmucosalinvasionforgastricneoplasmsinendoscopicimagesusingdeeplearning AT changseokbang predictionofsubmucosalinvasionforgastricneoplasmsinendoscopicimagesusingdeeplearning AT jaejunlee predictionofsubmucosalinvasionforgastricneoplasmsinendoscopicimagesusingdeeplearning AT changwonseo predictionofsubmucosalinvasionforgastricneoplasmsinendoscopicimagesusingdeeplearning AT juhankim predictionofsubmucosalinvasionforgastricneoplasmsinendoscopicimagesusingdeeplearning |
_version_ |
1724600017164959744 |