Related Experiment Video
Updated: Oct 2, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
2.7K
Diagnosis of Esophageal Lesions by Multi-Classification and Segmentation Using an Improved Multi-Task Deep Learning
Suigu Tang1, Xiaoyuan Yu1, Chak-Fong Cheang1
1Faculty of Information Technology, Macau University of Science and Technology, Macau 999078, China.
Sensors (Basel, Switzerland)
|February 26, 2022
Summary
A new multi-task classification and segmentation (MTCS) model aids endoscopists in detecting esophageal lesions. This AI tool accurately classifies and segments lesions in endoscopic images, improving diagnostic accuracy and reducing physician burden.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Gastroenterology
Background:
- Accurate detection of esophageal lesions during endoscopic screening is challenging due to visual similarities.
- Endoscopists face a high workload, necessitating tools to reduce diagnostic burden and improve accuracy.
Purpose of the Study:
- To develop and evaluate a multi-task classification and segmentation (MTCS) model for esophageal lesions.
- To assist endoscopists in accurately classifying and segmenting esophageal lesions from endoscopic images.
Main Methods:
- Proposed a multi-task classification and segmentation (MTCS) model comprising an Esophageal Lesions Classification Network (ELCNet) and an Esophageal Lesions Segmentation Network (ELSNet).
- Trained and evaluated the MTCS model on a dataset of 805 esophageal images from 255 patients and 198 images from 64 patients.
Main Results:
- The MTCS model achieved a classification accuracy of 93.43%.
- The model demonstrated a segmentation performance with a Dice Similarity Coefficient of 77.84%.
Conclusions:
- The MTCS model significantly enhances endoscopist performance in esophageal lesion detection.
- This AI-powered tool accurately classifies and segments lesions, serving as a valuable assistant to minimize oversight risks.
