PCOS (Polycystic Ovary Syndrome) Detection using CNN and Transfer Learning

Abstract

The objective of this research is to create a model using CNN Transfer Learning techniques e.g., MobileNet-V2, EfficientNet-B2, and ResNet-50 for the prediction of PCOS. Polycystic Ovary Syndrome (PCOS) is a type of medical condition that causes infertility issues in women globally. The way our model detects PCOS using specific computer programs is the thing our study mainly focuses on. Precise diagnosis using ultrasound scans containing multiple cysts is a well-founded approach. So, to overcome the challenges faced in manual detection, our proposed model possesses an extended machine learning technique used to classify PCOS detection. Our approach includes the training and testing of 2644 images in a dataset utilizing a Convolutional Neural Network (CNN) with transfer learning techniques. The experiments on different models in our study do not perform the same. MobileNet-V2 model do not perform so well on this type of detection of PCOS with an accuracy of 62.30%. Furthermore, the efficientNet-B2 model performs with an accuracy of 99.18%. ResNet-50 is a star model which performs very well according to our study and model with the highest accuracy of 99.59% which makes the process more efficient and attested. A web application is also created for the fast detection of PCOS as our star model with highest accuracy is integrated as backend of the web app. Recommendations for future development and conclusions are also included in the report.