Breast Cancer Histopathology Image Classification using ADAptive Moment (ADAM) optimization Estimation
Mradul Kumar Jain, Brij Mohan Singh, Mridula Singh · 2023
Breast cancer’s diagnosis have numerous imaging techniques like ultrasound, X-rays, MRI and many more out of those biopsy imaging is the most certain way to determine whether a breast tumor is cancerous or not, but the main problem faced by oncologist during by medical imaging is whether the person has tumor or not or the whether the tumor is cancerous as cancerous can be dangerous if not determined on times and can increase the risk of spreading to other cells and tissues. Nowadays deep learning architectures are most preferably used by the classification procedure in medical field for the diagnosis of the disease. This works aims to classify the breast cancer’s image into cancerous and non-cancerous for differentiation of benign cases from malignant cases by using a 3 convolutional module for the process of feature extraction, and two fully connected dense layers with an output layer for classification by tuning the hyperparameters on scaled images. The hypermeter we tuned to get the better result is dropout rate and the classification is performed and confusion matrix is plotted for the problem. Applied procedure is able to find benign cancer cases from malignant cancer cases of breast cancer and achieving the accuracy with training data 99.25%. On the other hand accuracy with testing data is 93% with precision percentage of 94 and fi score value of 93%. In the last this mechanism is obtaining ROC AUC value 9S.5% which is significant improvement over previous work.