Real-time automated image segmentation technique for cerebral aneurysm on reconfigurable system-on-chip

Abstract

Cerebral aneurysm is a weakness in a blood vessel that may enlarge and bleed into the surrounding area, which is a life-threatening condition. Therefore, early and accurate diagnosis of aneurysm is highly required to help doctors to decide the right treatment. This work aims to implement a real-time automated segmentation technique for cerebral aneurysm on the Zynq system-on-chip (SoC), and virtualize the results on a 3D plane, utilizing virtual reality (VR) facilities, such as Oculus Rift, to create an interactive environment for training purposes. The segmentation algorithm is designed based on hard thresholding and Haar wavelet transformation. The system is tested on six subjects, for each consists 512 × 512 DICOM slices, of 16 bits 3D rotational angiography. The quantitative and subjective evaluation show that the segmented masks and 3D generated volumes have admitted results. In addition, the hardware implement results show that the proposed implementation is capable to process an image using Zynq SoC in an average time of 5.2 ms.