MammoAI: automated mammographic breast cancer classifier powered by AutoML

Awang@ Ismail, Faikah and Abdul Karim, Muhammad Khalis and Awang Kechik, Mohd Mustafa and Kamal, Izdihar and Zaidon, Siti Izzatul Akma and Abdulwahid Noor, Katham (2025) MammoAI: automated mammographic breast cancer classifier powered by AutoML. In: Negeri Sembilan International Exposition (NSIEx) & Research Symposium 2025: e-Book of Extended Abstract. Universiti Teknologi MARA, Negeri Sembilan, pp. 210-213. ISBN 9786299595373
Abstract

MammoAI is a breakthrough AI-powered system designed to revolutionize breast cancer screening through fully automated mammographic image analysis. Developed using the state-of-the-art Auto-Sklearn AutoML framework, MammoAI integrates advanced image enhancement, tumor segmentation, and optimized machine learning pipelines to classify breast tumors with high precision. This product addresses the critical need for consistent, scalable, and accurate diagnostic tools, particularly in regions with limited radiological expertise. The innovation combines Contrast Limited Adaptive Histogram Equalization (CLAHE) and Active Contour Method (ACM) for robust tumor delineation, followed by extraction of 37 key radiomic features. These are then processed through MammoAI’s smart ensemble learning engine, achieving 76.16% test accuracy and 0.86 AUC after optimization. Designed with scalability, speed, and minimal user input in mind, MammoAI holds strong commercialization potential as a clinical decision support system that augments radiologist workflow, reduces diagnostic delays, and enhances early detection outcomes.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 146033.pdf]
Text
146033.pdf
Download (8MB)
Indexing & Metrics
Download Statistics