Machine Learning-Assisted Early Detection
Abstract
Artificial intelligence (AI) has shown promise in enhancing glioblastoma (GBM) diagnosis through histopathological image analysis, yet further refinement is needed to improve accuracy across diverse patient populations. This follow-up study aims to expand on prior work by evaluating the robustness and generalizability of AI-driven texture-based analysis in GBM detection. Using an expanded dataset incorporating multi-center histopathological images, this study will assess the performance of previously validated image features—such as those derived from gray level co-occurrence matrix (GLCM) and gray level run length matrix (GLRLM)—across different histological staining techniques and imaging platforms.
Additionally, this study will integrate explainable AI (XAI) methodologies to improve interpretability, ensuring that image feature importance aligns with established pathological indicators. Machine learning models, including support vector machines (SVM) and deep learning approaches, will be retrained and validated on this expanded dataset, with a focus on maintaining high diagnostic sensitivity and specificity. By identifying potential discrepancies in model performance across patient cohorts, the study aims to refine AI-driven GBM detection for broader clinical adoption. Ultimately, the findings could pave the way for AI-assisted histopathology as a reliable, scalable tool in neuro-oncology, enabling earlier and more precise GBM diagnoses.
References
- Cheung, E. Y. W., Wu, R. W. K., Li, A. S. M., & Chu, E. S. M. (2023). AI deployment on GBM diagnosis: A novel approach to analyze histopathological images using image feature-based analysis. Cancers, 15(20), 5063.
