سامانه همایش‌ها TSTA | ثبت‌نام و ارسال مقاله کنفرانس‌ها و کنگره‌های ملی

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

.Deep Learning-based Segmentation of Human Sperm Heads using YOLOv8 and SAM

نویسندگان :

Hadis Aligoo Zanjany¹ Maryam Pashaiasl² Ata Jodeiri³

1- دانشگاه علوم پزشکی تبریز 2- دانشگاه علوم پزشکی تبریز 3- دانشگاه علوم پزشکی تبریز

کلمات کلیدی :

Sperm segmentation،Deep learning،YOLOv8،Segment Anything Model،SAM

چکیده :

Accurate assessment of sperm quality plays a vital role in diagnosing male infertility. One of the major challenges in this process is the precise segmentation of sperm heads in microscopic images, which is typically performed manually and is both time consuming and error prone. In this work, we propose a hybrid deep learning framework for automated sperm head segmentation. The method employs YOLOv8 for initial sperm head localization, while its bounding boxes are used as prompts for the Segment Anything Model (SAM) to generate refined segmentation masks. Experimental results demonstrate that the hybrid YOLOv8→SAM approach outperforms standalone models, achieving an IoU of 78.62% and a Dice score of 70.78%. These findings highlight the potential of hybrid frameworks to significantly improve segmentation accuracy and robustness in biomedical imaging applications.