Early detection of PCOD using machine learning techniques
Pratik P. Wagh, Megha Panjwani, Amrutha S. V. · 2021
Polycystic Ovary Disease (PCOD) is a common disorder among women with no exact cure known to date. It shows various symptoms and may even contribute to long term health problems. It is largely ignored, due to a lack of awareness and usually detected when women try for conception. To confirm the diagnosis various hormonal blood tests along with an ultrasound scan is required which leads to multiple trips to well-equipped hospitals in urban areas. Currently, the whole process is expensive for people with a poor background, especially for people living in rural areas. Our contribution is to predict PCOD as accurately as possible, thus we used both Tabular data consisting of metabolic and clinical parameters and ultrasound scan images of patients for prediction. We found CatBoost as the best performing model with F1-score 88.68% for Tabular data and YOLOv2 with F1-score of 85.86% for ultrasound scan images. A Related contribution is to create an application that can be used by doctors to upload the tabular data and ultrasound scan images to predict PCOD and save the patients data for future use.