Self-supervised approach for organs at risk segmentation of abdominal CT images

Accurate segmentation of organs at risk is essential for radiation therapy planning. However, manual segmentation is time-consuming and prone to inter and intra-observer variability. This study proposes a self-supervision based attention UNet model for OAR segmentation of abdominal CT images. The model utilizes a self-supervision mechanism to train itself without the need for manual annotations. The attention mechanism is used to highlight important features and suppress irrelevant ones, thus improving the model’s accuracy. The model is evaluated on a dataset of 100 abdominal CT scans and compared its perfor mance with state-of-the-art methods. Our results show that the proposed model got comparable performance in terms of the dice similarity coefficient. More over, the inference time is much faster than traditional manual segmentation methods, making it a promising tool for clinical use.