Classification of Usual and Atypical Ductal Hyperplasia using Deep Learning

Breast Cancer is the leading cause of death among women. In the diagnosis of preinvasive breast cancer, some of the intraductal proliferations pose a special challenge. One such challenge includes the classification of usual ductal hyperplasia (UDH), atypical ductal hyperplasia (ADH), and ductal carcinoma in situ (DCIS). Prognosis and diagnosis of intraductal breast lesions pinpointing invasive and non-invasive cancer lesions is a monotonous task due to their similar structures in biopsy samples of breast. Expert’s differential diagnosis between breast ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) is of great importance in arriving at an optimal treatment plan. In this research work, we present AI based solution to classify the most confusing stages of breast cancer, that is Usual ductal hyperplasia (UDH), Atypical ductal hyperplasia (ADH) and Ductal carcinoma in situ (DCIS) in breast biopsy. Statistical analysis is carried out to justify the results. Data set having images of normal mammary glands, invasive ductal lesions, non-invasive ductal lesions which include Usual ductal hyperplasia (UDH), Atypical ductal hyperplasia (ADH), Ductal carcinoma in situ (DCIS) in breast biopsy specimens are employed in this study. Statistical analysis is performed to provide the statistical evidence of experiments. Analysis of Variance (ANOVA) as a feature selection method is used to select best features. The proposed method has performed quite significantly, where scores are statistically and medically acceptable. The highest accuracy recorded is 83.33%. When tested for robustness, the accuracy has dropped down to 65.25% allowing one to carry out further investigations.