Method for tumor recognition with short dynamic PET images: Theory and simulation study

A new method was provided in this paper to recognize tumor in PET images automatically and reduce the waiting time of patient. Based on the unsupervised clustering algorithm (ISODATA), this method diagnoses tumor with short dynamic PET images. The theoretical basis of this method is that the metabolic characteristics of different tissues, represented by time-activity curve in the region of interest, are distinctive. The computer program was developed and validated using simulated dynamic PET data with small tumor in lung. Simulation study shows that this method could recognize tumor with the short dynamic PET data in 10 min successfully, and this method was not sensitive to initial cluster center.