Related Experiment Videos
Explainable Lightweight Model Using Low-Rank and Convolutional Block Attention for Pancreatic Cancer Diagnosis
Vishesh Tanwar1, Bhisham Sharma1, Dhirendra Prasad Yadav2
1Chitkara University Institute of Engineering and Technology, Chitkara University, Rajpura, Punjab, India.
Summary
A new hybrid deep learning model significantly improves early pancreatic cancer (PC) detection using CT images. This AI framework achieves high accuracy with fewer resources, aiding clinical diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Early and accurate detection of pancreatic cancer (PC) is a critical clinical challenge.
- Current diagnostic methods for PC face limitations in speed and accuracy.
Purpose of the Study:
- To develop a novel, computationally efficient hybrid deep learning framework for automated pancreatic cancer classification from CT images.
- To enhance diagnostic performance in PC detection.
Main Methods:
- A hybrid deep learning framework integrating MobileNetV3Small, a convolutional block attention module, and Low-rank Attention with Shared Efficient Representations (LASER).
- Transformer encoder for capturing long-range dependencies and a cross-type interaction (CTI) module for feature fusion.
- Automated classification of CT images.
Main Results:
- Achieved 99.34% accuracy, 0.9996 AUC-ROC, 0.9897 Cohen's Kappa, and 0.9859 MCC on 18,942 CT images.
- Outperformed ResNet50, EfficientNetB0, and ViT variants with only 1.26 million parameters.
- Demonstrated high diagnostic performance with reduced computational requirements.
Conclusions:
- Explainability analyses (Grad-CAM, Grad-CAM++, attention visualization) confirm the model focuses on clinically relevant regions.
- The proposed framework offers a promising approach for accurate and efficient pancreatic cancer detection.