The Role of Mammographic Breast Area/Microcalcification Cluster Area (BA/MCA) Ratio in the Classification of BI-RADS 4 Lesions: A Step for Development of Artificial Intelligence in Breast Cancer Patients

Abstract

Introduction: The working principle of artificial intelligence in medicine is primarily as follows: The data are collected and entered into the system, the computer uses an algorithm to gather information via these data, and finally, it analyzes this algorithm to utilize in the diagnosis and treatment of the disease. In this study, we investigated the achievement of mammographic breast area/microcalcification cluster area ratio (BA/MCA) in the grouping of BI-RADS 4 (a, b, c) lesions. We planned to contribute to the development of artificial intelligence in medicine with a simple calculation program to be attached to the mammography computer.

Methods: 125 patients who underwent surgery with the diagnosis of mammographic BI-RADS 4 lesion (could not be detected in lesion-specific ultrasonography) between 2019 and 2022 in the Department of Surgical Oncology of Health Sciences University Gulhane Medical Faculty Training and Research Hospital were retrospectively examined. The mammographic MCA was divided by the breast area and their ratio was calculated. The relationship between the ratios we found and the BI-RADS values defined by radiology was analyzed.

Results: We found the median BA/MCA value of BI-RADS 4a patients to be 24943.5, BI-RADS 4b patients to be 12609.2, and BI-RADS 4c patients to be 11547.1 (p = 0.003). According to ROC curve analysis, we detected the BA/MCA ratio for BI-RADS 4c to be 14183.34 (AUC = 0.686, p = 0.005, sensitivity 54.2%). This ratio is inversely related, and the probability of BI-RADS 4c increases in patients with a BA/MCA ratio less than 14183.34. We revealed that the malignancy rate of radiological BI-RADS 4c patients was 90%, and the cutoff value of BI-RADS 4c patients was 72%. Using both classifications together, we detected the malignancy rate to be 98%.

Conclusion: The increase in the ratio of MCA to BA might have a place in the differentiation of BI-RADS 4 lesions. We foresee that artificial intelligence could also have a place in the classification of BI-RADS lesions with software to be installed on the mammography computer.