Deep Learning Based Cytogenetic Biomarker Analysis: YOLOv8 Automation of Mitotic Index in Toxicological and Radiation Studies

This poster was presented at the India International Science Festival (IISF) 2025 – Young Scientist Conclave. The work describes the development and evaluation of a deep learning–based framework using YOLOv8 for automated detection of mitotic cells and estimation of the mitotic index in cytogenetic studies relevant to toxicology and radiation research. Manual scoring of mitotic index in cytogenetic preparations is time-consuming, subjective, and not scalable for large studies. To address these limitations, a YOLOv8 model was trained and validated using labeled microscopy images of cultured human lymphocytes. The proposed approach enables rapid, reproducible identification of metaphase and blast cells, allowing automated calculation of mitotic index with high precision and recall. Model performance was assessed by comparison with expert manual scoring, demonstrating reliable detection accuracy and low false-positive rates. The results highlight the potential of deep learning–based automation to support high-throughput cytogenetic analysis, particularly in applications related to radiation biodosimetry, toxicological screening, and medical physics research. The methodology presented in this poster is intended as a scalable and reproducible tool to reduce observer variability and improve efficiency in cytogenetic workflows. An abstract version of this work was previously included in the IISF abstract book; the present poster reflects the final presented content and author order.