Aorta localization in Computed Tomography images: A YoloV9 segmentation approach

In this study, state-of the art segmentation methods were tested for use in biomedical CTA images to precisely highlight aorta instances. A set of YOLO architectures was trained on a newly annotated dataset of 120 transverse view images and detailed cross-validation results were obtained to measure the detection and precise localization ability of the models. The recent YOLOV9 algorithm flavors V9-C and V9-E were studied and compared to the previous V8-X model. The V9 version did score remarkable results to Mean Average Precision metrics, achieving 76.15% segmentation mAP50-95 score in the YOLOV9-C testing.