Comparative Study of CNN and Transfer Learning Techniques in the classification of PCO Ultra Sound Images

Abstract

Reproduction is the process of giving birth to a child. A child may bring all the happiness inside a family. Now a days due to change in the life style and the food habits, the couples may not have a successful reproduction. Even though there are many reasons for infertility, PCO in female is one of the major cause. PCOS can be treated and there are many procedures in the medical field which should be followed to get reproduction. Among the medical procedure US scanning is done to identify the presence of PCO. Compared to other medical tests US scans are cost effective and at the same time presence of PCOS can be easily identified. Many machine learning algorithms are applied on segmentation and classification of these images. In the proposed work, a self defined CNN model is created and the performance of the model is analyzed with the eight other models. VGG16, RESNET, Transfer Learning models having ANN and SVM as classifiers for VGG16,RESNET and self defined models are taken here. Accuracy of self defined model with SVM is comparatively same as VGG16 and RESNET50 with SVM but still the F1 score of self defined is low when compared VGG16 with SVM.