Related Experiment Video
Updated: Jul 29, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
A lightweight ResNet50V2-ECA model for renal cell carcinoma grading: efficiency, calibration, and state-of-the-art
1Department of Computer Science, College of Computer Science and Engineering, Taibah University, Yanbu, 46421, Saudi Arabia. mwateer@taibahu.edu.sa.
Scientific Reports
|June 11, 2026
Summary
A new lightweight deep learning model accurately classifies renal cell carcinoma (RCC) grades using Efficient Channel Attention, improving diagnostic efficiency and accuracy for better patient outcomes.
Area of Science:
- Digital pathology
- Machine learning in oncology
- Renal cell carcinoma (RCC) histopathology
Background:
- Accurate International Society of Urological Pathology (ISUP)-grade classification of renal cell carcinoma (RCC) is crucial but challenging due to histopathological subtleties and manual review limitations.
- Current deep learning models for RCC grading often employ complex attention mechanisms, leading to high computational costs and deployment barriers.
Purpose of the Study:
- To develop a lightweight and computationally efficient deep learning framework for accurate ISUP-grade classification of RCC.
- To enhance channel-wise feature discrimination with minimal overhead for improved diagnostic performance.
Main Methods:
- Introduction of a novel ResNet50V2-ECA framework integrating a single Efficient Channel Attention (ECA) block post-convolutional layers.
- Utilizing the KMC kidney histopathology dataset comprising 722 whole slide images (WSIs) and 3,442 training patches across five grades.
- Performing comprehensive ablation studies comparing various attention mechanisms and backbones, alongside comparative analysis with existing RCC grading frameworks.
Main Results:
- The ResNet50V2-ECA model achieved high accuracy (96.90%), F1-score (91.39%), and near-perfect AUC scores across all grades.
- Ablation studies confirmed ResNet50V2-ECA's superior performance over other attention mechanisms and backbones.
- Comparative analysis demonstrated the model's enhanced accuracy and efficiency against leading RCC classification frameworks.
Conclusions:
- The proposed lightweight ResNet50V2-ECA framework offers a high-precision, well-calibrated solution for RCC grading, showing significant potential for clinical deployment.
- Model reliability was validated using Monte Carlo Dropout, indicating strong confidence-accuracy alignment and high accuracy for deferred low-confidence cases.
- Further external multi-center validation is recommended to confirm the generalizability of the developed RCC grading system.