Göğüs Kanseri Kitle Tespitinde Derin Öğrenme Modellerinin Karşılaştırılması: YOLOv8 ve U-Net

Early detection of diseases is critical to the success of the treatment process, especially in life-threatening conditions such as cancer. In diseases such as breast cancer, early mass detection can be decisive for the effectiveness of the treatment process. This study compares the performance of YOLOv8 and U-Net models for mass detection in breast images. In the first stage, both models are evaluated on CBIS-DDSM and INbreast datasets. The results show that the YOLOv8 model outperforms U-Net in precision metrics. In the CBIS-DDSM dataset, YOLOv8 achieved a precision value of 0.800123, while U-Net achieved 0.762345. In the INbreast dataset, YOLOv8 achieved a precision value of 0.785234, while U-Net achieved a value of 0.742345. These findings show that YOLOv8 provides more successful and faster results, especially in object detection tasks, and is more efficient in areas where fast decisions need to be made, such as medical imaging. Future studies can develop hybrid solutions by combining the strengths of both models and optimize model speeds to achieve faster and more accurate results in medical diagnostics.