Robust Semi-Automated Arterial Input Function Identification Using Self Organizing Maps
Reddick W, Jain J
St. Jude Children's Research Hospital
Absolute quantification of cerebral blood flow and volume using dynamic-susceptibility contrast MRI relies on deconvolution of the arterial input function - commonly estimated from signal changes in a major artery. We present an automated technique to determine the local AIF by the selection of best-candidate arterial pixels using a Kohonen Self-Organizing Map. The technique, validated across five patients over three exams each, identified not just the blood vessels but also yielded unique classes for artery and vein. The technique leads to accurate calculation of the perfusion maps and may aid in normalizing results over various exams and patients.