Liver Lesion Detection from MR T1 In-Phase and Out-Phase Fused Images and CT Images Using YOLOv8

Unusual formations in the liver called lesions can happen for a number of different reasons. Some are malignant, while others are non-cancerous. Timely detection of the lesion and its treatment is necessary for malignant lesions. In this paper, we attempt to detect liver lesions using state-of-the-art single-stage object detection technology YOLOv8. On two distinct datasets, each of which contained 38 MR T1 fused image samples and 41 CT samples. The model is first evaluated on both datasets separately and then evaluated by combining both datasets. The findings demonstrate that the model is capable of accurately and quickly detecting liver abnormalities. The model performs best on the MRI dataset with an average detection time of 62.96 ms with an accuracy of 83.6%. Our research aims to shed light on single-stage object detection models’ unrealized potential in medical image analysis. We hope the findings contribute to developing a real-time tumor/lesion detection technology that gives accurate results in real time.