Vector Conversion Based PCOS Detection in data segmentation using Multi Task Learning by Dynamic Deep Learning Architecture

ABSTRACT: Polycystic ovarian syndrome (PCOS), the most prevalent endocrine abnormality in women who are fertile, interferes with hormone secretion over time, leading to a large number of ovarian cysts and other serious health problems. However, the doctor's experience plays a significant role in the accuracy of interpretations, which makes the practical clinical diagnostic approach for PCOS essential. Therefore, a PCOS prediction model powered by artificial intelligence might be a workable supplement to the labor-intensive and prone to error diagnosis technique. This research proposes a novel technique in clinical data-based PCOS detection by dimensionality reduction with segmentation using a deep learning model. Here the input PCOS clinical data has been collected and processed for removing missing values and dimensionality reduction based on vector conversion with Kernel Principal Component Analysis. Then the data quality is enhanced with annotation coverage for PCOS detection using a dynamic Bayesian hidden Markov v model. The experimental analysis is performed for the PCOS clinical dataset in terms of prediction accuracy, validation accuracy, RMSE, precision, and F-1 score. The proposed method obtained an overall 97% F-1 score and prediction accuracy of 98%, validation accuracy of 98%, RMSE of 1%, and precision of 99%.