A Novel Embryo Morphology Evaluation Based on Improved YOLOv8 Object Detection Model
Published in International Conference on Life System Modeling and Simulation, LSMS 2024 and International Conference on Intelligent Computing for Sustainable Energy and Environment, ICSEE 2024, Suzhou, China, 2025
In this study, an improved YOLOv8 object detection model (YOLOv8-C2f-CA) is proposed to automate the detection and quantification of embryonic cells, supporting embryologists in accurately assessing embryo morphology. By integrating a Coordinate Attention (CA) mechanism into the YOLOv8 architecture, our model achieved 87.8% mean average precision (mAP), 83.9% precision, and 76.4% recall, outperforming the baseline YOLOv8. The lightweight CA mechanism is incorporated into both the backbone and neck networks of YOLOv8, bolstering the model’s capacity to identify the key morphological features of embryos without substantially impacting its size or computational efficiency. This method facilitates a rapid and precise evaluation process, minimizing the need for extensive time and human resources while maintaining precise accuracy
Recommended citation: Talha, O., Zhou, W., Xu, Y., Liu, Q., Odeyemi, J. (2024). A Novel Embryo Morphology Evaluation Based on Improved YOLOv8 Object Detection Model. In: Gu, J., Hu, F., Zhou, H., Fei, Z., Yang, E. (eds) Robotics and Autonomous Systems and Engineering Applications of Computational Intelligence. LSMS ICSEE 2024 2024. Communications in Computer and Information Science, vol 2220. Springer, Singapore. https://doi.org/10.1007/978-981-96-0313-8_15
Download Paper