PCOS-Diagnosenet: A Cross-Modality Attention-Based Framework for Early Detection of Polycystic Ovary Syndrome Using Lightweight CNNS And Patient Biometrics

Polycystic Ovary Syndrome (PCOS) is a polygenic disease and is characterized by heterogeneous signs and symptoms that can vary greatly in combination and severity. This paper presents a novel hybrid diagnosis system, namely PCOS-DiagnoseNet, using an attention-fused MobileNetV2 based model and clinical data encoder for the robust detection of PCOS. We integrate ultrasound image features and clinical measurements like hormone levels and patient history by a low-cost attention based fusion mechanism. On a multimodal dataset, the results obtained by the proposed model (AUC= 93.2%) were better than those of traditional single-input models. Its small footprint permits integration with mobile platforms for applications like point-of-care diagnostics. The research highlights the revolutionary scale of cross-modality AI for women's health diagnostics.