Interpretable artificial intelligence (AI) for cervical cancer risk analysis leveraging stacking ensemble and expert knowledge

Short Note: This study addresses real-world challenges in predicting early-stage Cervical Cancer, achieving 98% accuracy and a 0.995 AUC score with a robust stacking ensemble model. Using interpretability tools such as SHAP, LIME, ELI5, and DT surrogate models enhanced transparency, identifying influential features like age, hormonal contraceptives, and STDs. Integrating expert feedback through interviews helped rank features and assess their impact, increasing eXplainability, reliability, and clinical relevance. This approach bridges AI and healthcare, promoting model acceptance and adoption in clinical practice.

An explainable ensemble approach for advanced brain tumor classification applying Dual-GAN mechanism and feature extraction techniques over highly imbalanced data

Short Note: We introduce a novel and reliable pipeline leveraging WGAN-GP and Real-ESR GANs with deep feature extraction methods to significantly enhance the accuracy of brain tumor classification from MRI images. Our approach effectively addresses challenges posed by highly imbalanced data and demonstrates strong generalizability, achieving superior accuracy compared to state-of-the-art deep learning models. These results underscore the potential of our proposed mechanism to improve computer-aided early-stage brain tumor diagnosis. Moreover, our study focuses on achieving high classification accuracy while prioritizing stable performance. By incorporating Grad-CAM, we enhance the transparency and interpretability of the overall classification process. This research identifies the most relevant and contributing parts of the input images toward accurate outcomes enhancing the reliability of the proposed pipeline.

Accepted in Machine Learning Technologies on Energy Economics and Finance – Energy and Sustainable Analytics, Publication in Studies in International Series in Operations Research & Management Science, Springer, 2024. 

Image Source: China, India, and the U.S. account for 52% of global CO2 emissions. (n.d.). Voronoi. https://www.voronoiapp.com/climate/China-India-and-the-US-Account-for-52-of-Global-CO2-Emissions-197