YOLOv12 Deep Learning Model Achieves 99.3% Accuracy in Acute Myeloid Leukemia Cell Classification

A deep learning model has demonstrated exceptional performance in classifying Acute Myeloid Leukemia cells, achieving 99.3% accuracy in both validation and testing phases. The YOLOv12 model was applied to this challenging medical classification task, which involves distinguishing between visually similar blood cell types. Researchers employed two distinct segmentation approaches based on cellular and nuclear features, preprocessing images using Hue channel and Otsu thresholding techniques. The most effective method combined YOLOv12 with Otsu thresholding applied to cell-based segmentation. Acute Myeloid Leukemia represents one of the most dangerous forms of blood cancer, where accurate classification remains difficult due to morphological similarities between different cell types. This research addresses the multiclass classification of AML cells through advanced computer vision techniques. The study was published on arXiv, a platform for scientific preprints, under the computer vision and pattern recognition category. The work contributes to medical imaging and diagnostic technology development.