Cerebral Vasculature Segmentation

Introduction

Cerebral vasculature segmentation in 3DRA and MRA modalities continues to be an open problem in the field of biomedical imaging, and there is still no standard solution. In this work, we proposed a vessel segmentation method based on the Swin transformer architecture with a composite loss function and dedicated pre and post-processing pipeline. We evaluated our method on the SHINY-ICARUS (3DRA) and SMILE-UHURA (MRA) dataset and showed that it achieves state-of-the-art performance in terms of common segmentation evaluation metrics such as Dice and JAC. Our method has the potential to significantly improve the efficiency and accuracy of vasculature segmentation in clinical practice.

Result

3DRA Modality

Joint 1st place in ISBI 2023 SHINY Challenge. Our team name is FunPixel.

MRA Modality