Related Experiment Videos
Performance and generalization analysis of machine learning, deep learning, and transformer models for histopathology
1Department of Computer Science, College of Applied Sciences, King Khalid University, Abha, Saudi Arabia. wmsami@kku.edu.sa.
Scientific Reports
|May 12, 2026
Summary
Deep learning and transformer models excel in histopathology image classification, outperforming traditional machine learning. Transformer models show superior generalization for complex tissue analysis in computer-aided diagnosis.
Area of Science:
- Medical Imaging
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Histopathology image classification is vital for computer-aided diagnosis, aiding pathologists in disease detection and grading.
- Artificial intelligence, including machine learning, deep learning, and transformers, is increasingly applied to histopathological image analysis.
- A lack of systematic, unified comparisons limits understanding of these AI approaches in this domain.
Purpose of the Study:
- To conduct a comprehensive performance and generalization analysis of classical machine learning, convolutional neural network (CNN), and vision transformer models for histopathology image classification.
- To systematically compare these diverse AI models under a unified experimental framework.
- To provide practical insights for selecting optimal classification models in histopathology.
Main Methods:
- Utilized publicly available benchmark datasets for histopathology image classification.
- Employed standardized preprocessing, training protocols, and evaluation metrics for all tested models.
- Evaluated classical machine learning classifiers, CNN models, and vision transformer architectures.
Main Results:
- Deep learning and transformer-based models significantly outperformed traditional machine learning approaches in classification accuracy.
- Transformer models demonstrated enhanced generalization capabilities, particularly on complex tissue patterns.
- The study analyzed the strengths and limitations of each model category regarding accuracy, robustness, and computational cost.
Conclusions:
- Deep learning and transformer models represent advanced solutions for histopathology image classification compared to traditional methods.
- Transformer architectures offer promising generalization for complex histopathological analyses.
- This research aids in selecting appropriate AI models for diagnostic decision-support systems in medical imaging.