Use of Artificial Intelligence to Enhance Mammographic
Tumour Size Evaluation in Breast Cancer

A multidisciplinary team of researchers are completing a study that will develop and evaluate an AI based algorithm for estimating breast cancer tumour size and extent on preoperative mammography, and to compare its accuracy with standard radiologist interpretation. By improving the accuracy, the way tumors are assessed before surgery, this research aims to help surgeons remove the cancer more completely the first time and reduce the need for additional surgeries. 
 

Accurate preoperative assessment of tumour extent is critical to surgical success. Mammography remains the most widely used imaging modality, yet its accuracy in estimating true tumour size is modest and frequently affected by breast density and tumour biology. Underestimation can lead to positive margins, while overestimation may result in unnecessary tissue removal or mastectomy. 
 

Although AI has demonstrated strong performance in screening mammography, its role in preoperative tumour size estimation has been minimally studied. Given rapid advances in deep learning methodologies, further evaluation is warranted. Understanding whether AI can improve mammographic tumour accuracy, and identifying factors that affect its performance has the potential to meaningfully reduce re‑excision rates, optimize surgical planning, and improve patient outcomes. 

 
Collaborators: Dr. Rasika Rajapakshe, Dr. Angel Arnout, Dr. Michael Partrick, Dr. Patricia Lasserre 
 

Students: Alex Niazi, Ruben Ispiryan