Automated Selection of Arterial Input Function (AIF) Pixels in DCE-MRI
Pohlmann A, Schreiber W, Oberholzer K
Johannes Gutenberg University Medical School
When estimating the AIF from a dynamic images series a high reproducibility in identifying the most adequate arterial pixels is crucial for reliably detecting changes in perfusion. Generally used manual outlining of ROIs is prone to introducing large variations. However, recently developed automated AIF detection methods are either very specific (score/threshold based methods) and thus difficult to apply to other imaging locations or employ complicated mathematical techniques (e.g. fuzzy clustering) that are not easily implemented. We have developed a simple adaptive and flexible method that automatically selects AIF pixels, while using available prior information and requiring only minimal user interaction.