Extraction of an input function from dynamic micro-PET images using wavelet packet based sub-band decomposition independent component analysis

► We used WP SDICA to extract input functions with zero or one plasma sample. ► We used FCM and population-based approaches to compare with our method. ► The one-sample WP SDICA approach is more accurate than two other approaches. ► The error of the MRFDG estimated by the one-sample WP SDICA is the smallest (7.0%).