Detection of Breast Cancer from Histopathology Images Using Deep Learning
Anjali Madhwani, Pankaj Kumar, Anurag Dubey, Umakant Mandawkar · Zenodo (CERN European Organization for Nuclear Research) · 2021
The stated system presents pre-trained Convolution Neutral Network (CNN) model which is convolutional Neutral Network to verify pre-segmented Breast Cancer mass mammogram tumour as benign or malignant. Based on detailed researched and analysis, to overcome the limitations of infrequency of available training datasets, Data augmentation, particular pre-processing & transfer learning is applied to achieve results. To tackle the classification issues noted above, this processed system is built on a modified version of DESNET 201. The suggested architecture has undergone extensive training and testing. The Convolution Neutral Network (CNN) was trained using data from the RGB colour model, which included 2480 benign and 5429 malignant cases. The achieved accuracy is 0.97%, the precision achieved for benign is 0.99% and recall rate is 0.83%. An achieved precision for malignant 0.83% following recall rate is 0.99 %. Overall, the presented DENSENET201 model excelled the previously proposed method for this system in terms of accuracy.