Doctors may find it challenging to identify breast cancer through mammography, but image processing can assist. Tumors and macrocalcifications are typically detectable using pixel intensity analysis. The issue with this approach is that its high false positive rate causes diagnostic errors and needless biopsies. In this work, we investigate the dependability of mammography analysis techniques based on pixel intensity. Otsu thresholding, K-means clustering, and “contrast-limited adaptive histogram equalization (CLAHE)” are examples of segmentation and classification techniques that frequently have flaws, according to research in the literature. These errors are caused by variations in breast tissue, noise sensitivity, and imaging artifacts. Although hybrid methods (like CNNs, SVMs, and CANs) can reduce false positives by up to 30%, they are challenging to apply for small lesions. Based on previous research, we discovered that tumors cannot be reliably classified using only pixel intensity. Combining morphological, textual, and contextual parameters is crucial for improving breast cancer detection ans reducing false positives.
Wahiba et al. (2026) studied this question.