COMPUTER-AIDED LESION DETECTION FOR BRAIN PET IMAGES

摘要

We proposed a new method for the automatic detection of lesions in brain PET images based on the asymmetry across segments of the segmented PET data. Per-pixel asymmetry feature detection is experimentally compared with our per-segment approach. The proposed per-segment method is shown to produce fewer false positives in the brain PET data examples presented. Also, the algorithm is shown to provide accurate demarcation of the tumour in each plane of PET data. Therefore, this method offers potential advantages in region-of-interest (ROI) extraction for kinetic modelling and quantitative analysis.