Reliable Detection of Polycystic Ovary Syndrome Using a Hybrid Deep Learning Approach: CNN-LSTM-GRU Integration
Sruthi SanilKumar · International Journal of Electrical and Electronics Engineering · 2025
Polycystic Ovary Syndrome (PCOS) is a prevalent hormonal disorder in women of reproductive age categorized by the presence of numerous tiny cysts on the ovaries, greater levels of androgen and irregular menstrual cycles. PCOS detection involves identifying and categorizing ovarian health conditions employing medical imaging modalities. Accurate detection is essential for appropriate treatment and inhibition of related health problems. Challenges such as the intrinsic complexity of ovarian morphological characteristics and differences in image quality due to transformations in acquisition settings or noise significantly affect the accuracy of detection systems. Conventional methods severely depend on manual image examination and feature extraction, often leading to variations and limited reliability. This research focuses on creating a hybrid Deep Learning (DL) system for the unfailing detection of PCOS employing ultrasound images. The system was assessed on the Kaggle PCOS dataset containing 3,856 images categorized into "infected" and "not infected" cases. Preprocessing and data augmentation techniques were utilized to increase variability in data, followed by feature extraction operating the hybrid model. The proposed system merges Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) architectures to collect spatial and temporal features expertly. The system achieved outstanding results with an accuracy of 98.50%, precision of 98.59%, recall of 98.55% and an F1 score of 98.48%. These results underscore the efficacy of the hybrid framework in responding to the challenges of PCOS detection, providing an efficient and better solution for clinical applications.