AUTO-FOLLICULOMETRY CLINICAL ASSISTANT (AFCA): A U-NET BASED APPROACH FOR ULTRASOUND IMAGE SEGMENTATION

The process of ovarian follicle measurement in IVF treatments is traditionally performed manually using ultrasound images, which is time-consuming and prone to inter-observer variability. This project proposes an AIbased system, Auto-Folliculometry Clinical Assistant (AFCA), to automate follicle detection and measurement. The objective is to improve diagnostic accuracy, reduce manual effort, and ensure consistent clinical results. The system uses a U-Net-based deep learning model with a ResNet backbone for robust image segmentation. This is paired with preprocessing techniques such as CLAHE for contrast enhancement, and post-processing using OpenCV for contour detection and precise size estimation. The proposed method enables pixel-level segmentation of follicles, allowing for the precise calculation of their dimensions. Experimental results show that the system reduces scan analysis time by approximately 50% and vastly improves measurement consistency. This approach demonstrates the strong potential of artificial intelligence in enhancing medical imaging workflows and supporting clinical decision-making.