An Automatic Estimation of Arterial Input Function Based on Multi-Stream 3D CNN

Published on July 6, 2019

Arterial input function (AIF) is estimated from perfusion images as a basic curve for the calculus of hemodynamic variables used as surrogate markers of the vascular status of tissues. However, at present, identification of AIF is made manually leading by prior knowledge. We propose an automatic estimation of AIF in perfusion images, which is based on a multi-stream 3D CNN, which combined spatial and temporal features together to estimate AIF ROI. The model is trained by manual annotations. The proposed method was trained and tested by 100 cases of PWI images. The result was evaluated by dice similarity coefficient (DSC), which reached to 0.79. The trained model had a better performance than the traditional method. After segmentation of AIF ROI, AIF was calculated by average of all voxels in ROI. We compared AIF result with manual and traditional method, and father process parameter maps such as Tmax and rCBF. The result had a better performance. We had applied the method on a cloud platform Estroke, which can evaluation ischemic penumbra to help make treatment decision.

Read Full Article (External Site)