Adrenal Tumor Vessels Segmentation Using Convolutional Neural Network in Computed Tomography Angiography

Abstract

The adrenal glands are important endocrine glands in humans. They are in complex environments with thin vessels around them. It's meaningful to get the accurate dissection before surgery. However, images used in hospitals are now unable to help doctors with many surgeries, which are produced by digital subtraction angiography. In this study, we used a 3D U-Net model to segment the adrenal tumor vessels in 3D computed tomography angiography slices. The model was evaluated by dice similarity coefficient (DSC) and mean intersection over union (MIoU) with the manually labeled ground truth. The DSC in this model is 94.69% and the MIoU is 90.22%.