Related Experiment Video
Updated: Jul 26, 2025
06:44
Automated Dissection Protocol for Tumor Enrichment in Low Tumor Content Tissues
Published on: March 29, 2021
2.6K
SnapEnsemFS: a snapshot ensembling-based deep feature selection model for colorectal cancer histological analysis
Soumitri Chattopadhyay1, Pawan Kumar Singh1, Muhammad Fazal Ijaz2
1Department of Information Technology, Jadavpur University, Jadavpur University Second Campus, Plot No. 8, Salt Lake Bypass, LB Block, Sector III, Salt Lake City, Kolkata, 700106, West Bengal, India.
Scientific Reports
|June 19, 2023
Summary
This study introduces a novel deep learning framework for automated colorectal cancer detection using histological images. The method achieves high accuracy, improving early diagnosis and potentially reducing manual analysis burdens.
Area of Science:
- Oncology
- Computer Science
- Medical Imaging
Background:
- Colorectal cancer is a leading cause of cancer death, necessitating early diagnosis for effective treatment.
- Manual analysis of histological slides for colorectal cancer screening is time-consuming and prone to human error.
- Automated computer-aided detection (CAD) systems can enhance the efficiency and accuracy of cancer diagnosis.
Purpose of the Study:
- To develop and evaluate an automated deep learning-based CAD framework for colorectal cancer detection from histological images.
- To improve upon existing methods by employing a computationally efficient snapshot ensemble technique combined with feature optimization.
- To provide a visually explainable AI tool for medical practitioners in colorectal cancer screening.
Main Methods:
- A novel snapshot ensemble method was utilized, extracting deep features from the penultimate layer of a Convolutional Neural Network (CNN).
- Particle Swarm Optimization (PSO) was applied for dimensionality reduction of extracted deep features to mitigate redundancy.
- The proposed framework was evaluated on a public colorectal cancer histology dataset using a five-fold cross-validation scheme.
Main Results:
- The proposed method achieved a highest accuracy of 97.60% and an F1-Score of 97.61%.
- The framework demonstrated superior performance compared to existing state-of-the-art methods on the same dataset.
- Class activation maps provided visual explainability, supporting the clinical utility of the CAD system.
Conclusions:
- The developed deep learning framework offers a highly accurate and efficient solution for automated colorectal cancer detection.
- The snapshot ensemble approach with PSO-based feature optimization presents a computationally viable alternative to traditional ensemble methods.
- The visually explainable CAD system can assist medical professionals in the screening of colorectal histology, aiding in early disease detection.
