A STUDY ON PARADIGM SHIFT IN DIAGNOSING DISEASES USING ARTIFICIAL INTELLIGENCE: CASE STUDY ON POLYCYSTIC OVARIAN SYNDROME

Abstract

Polycystic Ovarian Syndrome is a complicated multifactorial disease identified as irregular menses, stubborn weight gain, insulin resistance, Hair loss, and in extreme cases, infertility and ovarian cancer. There are multiple diagnostic criteria but none of them covers all the aspects of PCOS and hence it becomes all the more important to perform absolutely correct diagnosis. In this case, Artificial intelligence has significantly played an important role. The present study discusses CNN-Convolutional Neural Network model using ultrasound images of infected and normal ovaries. The dataset was obtained from Open source database- Kaggle. With an accuracy of 74.5%, the model worked satisfactorily. To analyze the model, a normalized confusion matrix was also prepared, to rule out any confusion faced by model. Since the CNN Architecture was pretty straightforward and simple, accuracy was limited to 75%, however, with addition of more model layers, the accuracy can be increased. The amount of dataset can also be increased to improve efficiency of the model. With time, more techniques have been developed that can enhance the quality of healthcare in world. Smart AI like XAI or explainable AI is self-interpretable and helps working with complex algorithms. Precision medicine is also developing and innovative self-diagnosing applications can be built for diseases like PCOS which progress on symptoms that are easily identifiable and non-invasive.