Deep learning model for detecting high-grade dysplasia in colorectal adenomas

Objective: Early detection and removal of suspicious polyps during routine colonoscopies play an important role in reducing the risk of colorectal cancer. Patient management and follow-up are determined by the type of polyps removed and the degree of dysplasia present on histological evaluation. Whe...

Full description

Saved in:
Bibliographic Details
Main Authors: Eric Steimetz, Zeliha Celen Simsek, Asmita Saha, Rong Xia, Raavi Gupta
Format: Article
Language:English
Published: Elsevier 2025-04-01
Series:Journal of Pathology Informatics
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2153353925000264
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849387461128486912
author Eric Steimetz
Zeliha Celen Simsek
Asmita Saha
Rong Xia
Raavi Gupta
author_facet Eric Steimetz
Zeliha Celen Simsek
Asmita Saha
Rong Xia
Raavi Gupta
author_sort Eric Steimetz
collection DOAJ
description Objective: Early detection and removal of suspicious polyps during routine colonoscopies play an important role in reducing the risk of colorectal cancer. Patient management and follow-up are determined by the type of polyps removed and the degree of dysplasia present on histological evaluation. Whereas discerning between a benign polyp and a dysplastic one is a trivial task, distinguishing between tubular adenomas with low-grade dysplasia (LGD) and high-grade dysplasia (HGD) is a challenging task. In this study, we trained a deep learning model to distinguish between colorectal adenomas with LGD and HGD. Design: We retrieved 259 slides of adenomatous polyps taken between January 2011 and October 2024. Slides with HGD were reviewed by a subspecialty-trained GI pathologist. After excluding discordant and duplicate cases, 200 slides remained: 71 (35.5%) with HGD and 129 (64.5%) with LGD. The slides were divided into training (160 slides, 80%) and test (40 slides, 20%) sets. After patch generation and stain normalization, a ResNet34 model (pre-trained on ImageNet) was trained using 5-fold cross-validation. Slide classification was determined by aggregating patch-level predictions. Results: The model's slide-level prediction accuracy was 95.0%, correctly classifying all but 2 out of 40 slides. The model achieved an area under the receiver operating characteristic curve score of 0.981 and an F1 score of 0.923. Conclusions: This study demonstrates that deep learning models can accurately distinguish between colonic adenomas with LGD and HGD. Training on a larger dataset could increase the accuracy and generalizability of the model and should be a focus of further studies.
format Article
id doaj-art-58e2158230d0439084b2d41b8e596274
institution Kabale University
issn 2153-3539
language English
publishDate 2025-04-01
publisher Elsevier
record_format Article
series Journal of Pathology Informatics
spelling doaj-art-58e2158230d0439084b2d41b8e5962742025-08-20T03:53:51ZengElsevierJournal of Pathology Informatics2153-35392025-04-011710044110.1016/j.jpi.2025.100441Deep learning model for detecting high-grade dysplasia in colorectal adenomasEric Steimetz0Zeliha Celen Simsek1Asmita Saha2Rong Xia3Raavi Gupta4Department of Pathology, SUNY Downstate Health Sciences University, Brooklyn, NY, USA; Corresponding author at: 450 Clarkson Avenue, MSC 25, Brooklyn, NY 11203, USA.Department of Pathology, SUNY Downstate Health Sciences University, Brooklyn, NY, USADepartment of Pathology, SUNY Downstate Health Sciences University, Brooklyn, NY, USADepartment of Pathology, New York University School of Medicine, New York, NY, USADepartment of Pathology, SUNY Downstate Health Sciences University, Brooklyn, NY, USAObjective: Early detection and removal of suspicious polyps during routine colonoscopies play an important role in reducing the risk of colorectal cancer. Patient management and follow-up are determined by the type of polyps removed and the degree of dysplasia present on histological evaluation. Whereas discerning between a benign polyp and a dysplastic one is a trivial task, distinguishing between tubular adenomas with low-grade dysplasia (LGD) and high-grade dysplasia (HGD) is a challenging task. In this study, we trained a deep learning model to distinguish between colorectal adenomas with LGD and HGD. Design: We retrieved 259 slides of adenomatous polyps taken between January 2011 and October 2024. Slides with HGD were reviewed by a subspecialty-trained GI pathologist. After excluding discordant and duplicate cases, 200 slides remained: 71 (35.5%) with HGD and 129 (64.5%) with LGD. The slides were divided into training (160 slides, 80%) and test (40 slides, 20%) sets. After patch generation and stain normalization, a ResNet34 model (pre-trained on ImageNet) was trained using 5-fold cross-validation. Slide classification was determined by aggregating patch-level predictions. Results: The model's slide-level prediction accuracy was 95.0%, correctly classifying all but 2 out of 40 slides. The model achieved an area under the receiver operating characteristic curve score of 0.981 and an F1 score of 0.923. Conclusions: This study demonstrates that deep learning models can accurately distinguish between colonic adenomas with LGD and HGD. Training on a larger dataset could increase the accuracy and generalizability of the model and should be a focus of further studies.http://www.sciencedirect.com/science/article/pii/S2153353925000264GI polypsTubular adenomaHigh-grade dysplasiaDeep learningWhole-slide images
spellingShingle Eric Steimetz
Zeliha Celen Simsek
Asmita Saha
Rong Xia
Raavi Gupta
Deep learning model for detecting high-grade dysplasia in colorectal adenomas
Journal of Pathology Informatics
GI polyps
Tubular adenoma
High-grade dysplasia
Deep learning
Whole-slide images
title Deep learning model for detecting high-grade dysplasia in colorectal adenomas
title_full Deep learning model for detecting high-grade dysplasia in colorectal adenomas
title_fullStr Deep learning model for detecting high-grade dysplasia in colorectal adenomas
title_full_unstemmed Deep learning model for detecting high-grade dysplasia in colorectal adenomas
title_short Deep learning model for detecting high-grade dysplasia in colorectal adenomas
title_sort deep learning model for detecting high grade dysplasia in colorectal adenomas
topic GI polyps
Tubular adenoma
High-grade dysplasia
Deep learning
Whole-slide images
url http://www.sciencedirect.com/science/article/pii/S2153353925000264
work_keys_str_mv AT ericsteimetz deeplearningmodelfordetectinghighgradedysplasiaincolorectaladenomas
AT zelihacelensimsek deeplearningmodelfordetectinghighgradedysplasiaincolorectaladenomas
AT asmitasaha deeplearningmodelfordetectinghighgradedysplasiaincolorectaladenomas
AT rongxia deeplearningmodelfordetectinghighgradedysplasiaincolorectaladenomas
AT raavigupta deeplearningmodelfordetectinghighgradedysplasiaincolorectaladenomas