Yolov8-ED: An Enhanced Deep Learning Framework for Early Detection of PCOS from Ultrasound Scans

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine condition among women of reproductive age, linked to infertility, metabolic malfunction, and long-term negative health consequences. The precise and prompt identification of PCOS is crucial for appropriate medical intervention and diagnosis. This paper presents Yolov8-ED, an advanced deep learning model optimized for the early identification of PCOS using multilayer ultrasound images, in which optimization is achieved through multi-level feature extraction, spatial attention mechanism, and Adam-based hyperparameter tuning. This work utilizes the PCOS Detection using ultrasound images dataset, including infected and non-infected categories within the training and testing subsets. Yolov8–ED is evaluated against state-of-the-art models like VGG 19, WaOEL, CystNet, and ResNet50, using critical metrics such as AUC-ROC, precision, accuracy, F1-score, recall, and computational efficiency. The testing findings indicate that Yolov8-ED surpasses current methodologies in accuracy, precision, recall, and F1-score, which are 95.75%, 96.8%, 95.6%, and 95.4%, respectively, with an inference time of 31 milliseconds.