Voting Ensemble SVM Model for Deep CNN Based Breast Histopathology Classification
Jyoti Chowdhary, Praveen Sankaran · 2022 IEEE 19th India Council International Conference (INDICON) · 2022
Breast cancer is the second most common cause of cancer-related mortality among women. Histopathology image analysis forms a powerful tool for analyzing the tumor type. A convolutional neural network (CNN) is used to extract features from histopathology images, and the test images are classified using support vector machine classifiers. A voting ensemble technique that combines the estimated class probability scores from individual support vector machine (SVM) models is applied to increase the final accuracy. Each different model is evaluated using 10-fold cross validation. The average performance of each algorithm from random trial testing in a publicly available histopathology dataset using our method provides a maximum patch level accuracy of 98.36% with specificity 98.73% and sensitivity of 98.21% which are comparable to state-of-the-art.