Exploring frequency differences of physiological processes to enhance dynamic FDG-PET without blood function

Due to the difference between the glucose utilization rates of malignant and normal tissue, the observed dynamic FDG-PET time activity curve (TAC) of malignant tissue increases in time, whereas, the normal tissue´s TAC washouts with time. This paper aims at exploiting this temporal difference in frequency domain to characterize the behaviors of the individual physiological function, defined as the convolution of the input blood function and an exponential function with decay factor β, for malignance and normal tissue, respectively. We proved that for a given dynamic acquisition time duration T, there exists a value β*, differing for each patient, such that the physiological functions monotonically increase in the acquisition period T if the decay factor β<β*, otherwise decrease in T if β>β*. Based on the facts that FDG uptake, which reflects glucose metabolic rate, increases in tumor cell while decreases in normal tissue, we claim that the physiological functions with β<β* are attributed to malignances, whereas those with β>β* are associated to normal tissues. In this paper, we first map such temporal differences in malignancy and normal tissue to frequency domain, then design a low pass filter whose cut-off frequency is selected to suppress the frequency components mainly attributed to the normal tissues. The dependence of the cutoff frequency of the low pass filter on the decay factor of the physiological functions is also addressed. An image enhancement method based on the frequency domain property of the TAC is proposed and tested with both digital phantom and clinical data. One advantage of this method is that no blood function is needed for image enhancement process. Our results from clinic and phantom data show that the tumor to background contrast can be substantially improved.