Classification of Pancreatic Cystic Lesions Using ResNet Deep Learning Network in Confocal Laser Endomicroscopy Videos

Clara Lavita Angelina; Chien Ming Pan; Tsung Chun Lee; Ming Lun Han; Pradermchai Kongkam; Hsiu Po Wang; Chuan Yu Chang; Hsuan Ting Chang

doi:10.1016/j.procs.2024.03.015

Classification of Pancreatic Cystic Lesions Using ResNet Deep Learning Network in Confocal Laser Endomicroscopy Videos

Clara Lavita Angelina
, Chien Ming Pan
, Tsung Chun Lee
, Ming Lun Han
, Pradermchai Kongkam
, Hsiu Po Wang
, Chuan Yu Chang
, Hsuan Ting Chang

Research output: Contribution to journal › Conference article › peer-review

5 Citations (Scopus)

Abstract

Accurate classification of pancreatic cystic lesions is crucial to differentiate mucinous lesions of malignant potential. We utilized the ResNet-50 and ResNet-101 network to develop a model for classification of the pancreatic cystic lesions. A total of 50 videos, 13,425 images, from five types of pancreatic cystic lesions and utilize the image rotation and contrast reversal scheme for the training. We adopt a contrast limited adaptive histogram equalization method onto the test video. Our method can automatically classify the feature type and record the prediction results frame by frame. The method has been evaluated on 18 test videos and achieves an accuracy 94% overall.

Original language	English
Pages (from-to)	357-363
Number of pages	7
Journal	Procedia Computer Science
Volume	234
DOIs	https://doi.org/10.1016/j.procs.2024.03.015
Publication status	Published - 2024
Event	7th Information Systems International Conference, ISICO 2023 - Washington, United States Duration: Jul 26 2023 → Jul 28 2023

Keywords

Contrast Limited Adaptive Histogram Equalization
Contrast Reversal
Deep Learning
Pancreatic Cystic Lesions
ResNet-101
ResNet-50

ASJC Scopus subject areas

General Computer Science

Access to Document

10.1016/j.procs.2024.03.015

Cite this

@article{221548aac9424639ad12abb6baf52d5b,

title = "Classification of Pancreatic Cystic Lesions Using ResNet Deep Learning Network in Confocal Laser Endomicroscopy Videos",

abstract = "Accurate classification of pancreatic cystic lesions is crucial to differentiate mucinous lesions of malignant potential. We utilized the ResNet-50 and ResNet-101 network to develop a model for classification of the pancreatic cystic lesions. A total of 50 videos, 13,425 images, from five types of pancreatic cystic lesions and utilize the image rotation and contrast reversal scheme for the training. We adopt a contrast limited adaptive histogram equalization method onto the test video. Our method can automatically classify the feature type and record the prediction results frame by frame. The method has been evaluated on 18 test videos and achieves an accuracy 94\% overall.",

keywords = "Contrast Limited Adaptive Histogram Equalization, Contrast Reversal, Deep Learning, Pancreatic Cystic Lesions, ResNet-101, ResNet-50",

author = "Angelina, \{Clara Lavita\} and Pan, \{Chien Ming\} and Lee, \{Tsung Chun\} and Han, \{Ming Lun\} and Pradermchai Kongkam and Wang, \{Hsiu Po\} and Chang, \{Chuan Yu\} and Chang, \{Hsuan Ting\}",

note = "Publisher Copyright: {\textcopyright} 2023 The Authors. Published by Elsevier B.V.; 7th Information Systems International Conference, ISICO 2023 ; Conference date: 26-07-2023 Through 28-07-2023",

year = "2024",

doi = "10.1016/j.procs.2024.03.015",

language = "English",

volume = "234",

pages = "357--363",

journal = "Procedia Computer Science",

issn = "1877-0509",

publisher = "Elsevier B.V.",

}

TY - JOUR

T1 - Classification of Pancreatic Cystic Lesions Using ResNet Deep Learning Network in Confocal Laser Endomicroscopy Videos

AU - Angelina, Clara Lavita

AU - Pan, Chien Ming

AU - Lee, Tsung Chun

AU - Han, Ming Lun

AU - Kongkam, Pradermchai

AU - Wang, Hsiu Po

AU - Chang, Chuan Yu

AU - Chang, Hsuan Ting

PY - 2024

Y1 - 2024

N2 - Accurate classification of pancreatic cystic lesions is crucial to differentiate mucinous lesions of malignant potential. We utilized the ResNet-50 and ResNet-101 network to develop a model for classification of the pancreatic cystic lesions. A total of 50 videos, 13,425 images, from five types of pancreatic cystic lesions and utilize the image rotation and contrast reversal scheme for the training. We adopt a contrast limited adaptive histogram equalization method onto the test video. Our method can automatically classify the feature type and record the prediction results frame by frame. The method has been evaluated on 18 test videos and achieves an accuracy 94% overall.

AB - Accurate classification of pancreatic cystic lesions is crucial to differentiate mucinous lesions of malignant potential. We utilized the ResNet-50 and ResNet-101 network to develop a model for classification of the pancreatic cystic lesions. A total of 50 videos, 13,425 images, from five types of pancreatic cystic lesions and utilize the image rotation and contrast reversal scheme for the training. We adopt a contrast limited adaptive histogram equalization method onto the test video. Our method can automatically classify the feature type and record the prediction results frame by frame. The method has been evaluated on 18 test videos and achieves an accuracy 94% overall.

KW - Contrast Limited Adaptive Histogram Equalization

KW - Contrast Reversal

KW - Deep Learning

KW - Pancreatic Cystic Lesions

KW - ResNet-101

KW - ResNet-50

UR - https://www.scopus.com/pages/publications/85193200958

UR - https://www.scopus.com/pages/publications/85193200958#tab=citedBy

U2 - 10.1016/j.procs.2024.03.015

DO - 10.1016/j.procs.2024.03.015

M3 - Conference article

AN - SCOPUS:85193200958

SN - 1877-0509

VL - 234

SP - 357

EP - 363

JO - Procedia Computer Science

JF - Procedia Computer Science

T2 - 7th Information Systems International Conference, ISICO 2023

Y2 - 26 July 2023 through 28 July 2023

ER -

Classification of Pancreatic Cystic Lesions Using ResNet Deep Learning Network in Confocal Laser Endomicroscopy Videos

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this