Fig. 2

Automated image processing algorithm workflow. Raw images are first motion corrected, before the heart region is detected via segmentation of candidate ventricle regions from a standard deviation map. A bounding box is created around the selected regions, constraining further processing steps to this detected heart region. Intensity correction is performed prior to independent component analysis, which detects the time-intensity signal of the LV and RV. Correlation maps are used to classify each pixel to the RV, LV, or background (see Fig. 3). Finally, bright LV pixels are thresholded to extract the final AIF measurement (see Fig. 4)