Estimating Flow Direction of Circle of Willis Using Dynamic Arterial Spin Labeling Magnetic Resonance Angiography - AJNR 2024 May

This study presents an innovative end-to-end machine learning pipeline for analyzing blood flow direction in the Circle of Willis (CoW) using Dynamic Arterial Spin Labeling Magnetic Resonance Angiography (ASL dMRA).

Key achievements:

  1. Developed an end-to-end ML pipeline that processes 4D time-series ASL dMRA data to predict blood flow direction in the CoW.

  2. Built a computational model incorporating graph theory, bootstrap strategy, and ensemble learning to analyze complex 4D vascular structures and time-series data.

  3. Achieved high performance, with 92.8% accuracy in predicting flow direction compared to 3D Phase Contrast MRI as a reference.

  4. Successfully automated the analysis of high-dimensional medical imaging data, reducing the need for manual intervention.

End-to-end Processing Pipeline

Graph Modeling of Circle of Willis

 

Ensemble & Resample to Acquire Flow Direction (ML)