Abstract: Polycystic ovary syndrome (PCOS) significantly impacts women of reproductive age and is recognized as a leading cause of female infertility. It involves a disruption in hormone levels, which leads to irregular menstruation and the appearance of multiple small cysts in the ovaries. Timely identification of PCOS plays a key role in managing its manifestations and lowering the risk of associated health issues. Moreover, the efficiency of classical detection models was reduced due to the lack of advanced feature extraction schemes. Therefore, this study introduces an Explainable Artificial Intelligence+LeNet-Deep Maxout Network (Explainable AI +LeDMNet) for PCOS detection. Firstly, input ultrasound image is preprocessed with Gaussian filtering. After that, the segmentation of follicles from the ovary is conducted using Superpixel-based Fast Fuzzy C-Means Clustering (SFFCM), where region-level iterative strength is modified using chord distance. After segmentation, image augmentation techniques, including brightness, rotation, and translation, are applied to the segmented follicle regions to enhance dataset diversity, improve model robustness, and prevent overfitting. Following augmentation, the feature extraction process is performed, which identifies the patterns and correlations to enabling early detection and personalized treatment plans. Finally, the Explainable AI +LeDMNet is devised for PCOS detection. wherein LeDMNet is developed using LeNet and Deep Maxout Network (DMN). The evaluation outcomes expose that, Explainable AI +LeDMNet has gained accuracy, True Positive Rate (TPR), True Negative Rate (TNR), precision, and F1-score as 91.867%, 91.755%, 91.858%, 90.939%, and 90.302%.

Key words: PCOS detection')">PCOS detection, Fast Fuzzy C-Means Clustering Explainable AI, LeNet, Deep Maxout Network