HSG-Assistant: AI-Driven Framework for Enhanced Hysterosalpingography Analysis

HSG-Assistant is an AI-driven diagnostic framework designed to address the significant limitations of traditional Hysterosalpingography (HSG) imaging in diagnosing female infertility, including low resolution, high noise levels, and limited diagnostic accuracy. These challenges hinder the reliable detection of essential anatomical features, often leading to missed or inaccurate diagnoses. To overcome these issues, the framework integrates three key components: (1) a YOLO-based detection model to accurately localize and classify fallopian tube conditions with high precision (mAP@.5 of 99.5%), (2) a Multi-Scale Resolution (MSR) enhancement technique to refine image quality by addressing uneven illumination and noise while preserving fine anatomical details, and (3) a Multi-Channel Attention U-Net (MCAtt-U-Net) for detailed and accurate segmentation of complex structures, achieving an IoU of 84%. By significantly improving image clarity, segmentation precision, and diagnostic reliability, HSG-Assistant streamlines diagnostic workflows and provides clinicians with a reliable tool for enhancing infertility diagnoses and treatment planning.