Non-annotated renal histopathological image analysis with deep ensemble learning.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Renal cancer poses a significant global health threat, with early detection crucial for improving patient outcomes.
• Current manual analysis of renal tissues is time-consuming, variable, and may miss early cancer markers.

Purpose of the Study:

• To develop an automated system for renal histopathological image analysis using deep learning.
• To enhance the early detection of renal cancer and reduce diagnostic errors.

Main Methods:

• Developed heterogeneous ensemble models using deep convolutional neural networks (CNNs) from VGG, ResNet, DenseNet, MobileNet, and EfficientNet architectures.
• Images were segmented into patches, classified as normal or tumor using pre-trained CNNs, and combined via ensemble learning for final classification.
• Selected high-performing CNNs for ensemble learning to leverage diverse model strengths.

Main Results:

• The best performing model, a five-CNN weighted averaging ensemble, achieved 99% accuracy, 98% specificity, 99% F1-score, and 98% ROC AUC.
• Demonstrated superior performance compared to existing state-of-the-art methods in renal histopathological image analysis.

Conclusions:

• The developed ensemble model shows high robustness and reliability for assisting pathologists in renal tissue analysis.
• This AI-driven approach can enhance early renal cancer detection efficiency, minimize misdiagnosis, and ultimately improve patient survival rates.