Detection of Invasive Ductal Carcinoma using Transfer Learning with Deep Residual Network
Srinidhi Kulkarni, Amrita Sundaray · 2021
Invasive Ductal Carcinoma (IDC) is a most com-mon subtype of Breast Cancer. This is a cancer to which any gender is sensitive. The early diagnosis can help to start timely treatment for this rapidly spreading cancer which gradually infects the other parts of the body. This malignancy can be detected by the pathologists by some specialised pathological tasks. But this can be a tedious and time-consuming task. So, to speed up the detection, computational techniques are being employed in the pathology labs. The Deep Learning based approaches prove to be a reliable solution as they are generally accurate and less error prone. In this we propose a classification model which involves the pretrained Residual model for the feature extraction. The IDC datasets are generally huge, so most of the researchers train their models over a subset of the dataset. But in this paper a fold based training approach was used to train the model over the entire dataset. The model seems to be performing better than the other works and is reliable as it is trained and over the complete dataset distribution.