Fig. 5
A detailed explanation of the output obtained at each step during development of the noise correction algorithm. (a) An arterial input function (AIF) of a digital perfusion phantom (DPP) with zero baseline and temporal sampling rate of 1.5 s, generated by averaging the AIFs from a cohort of 59 pancreatic ductal adenocarcinoma (PDAC) patients. (b) An example of the simulated impulse response function (IRF) of a DPP. (c) An example of noise-free and noise-added tissue attenuation curves (TACs) for a DPP generated for both tissue types, PDAC and non-neoplastic pancreatic parenchyma. (d) An example of TACs from non-neoplastic pancreatic parenchyma and PDAC regions of a PDAC patient for comparison with the simulated TACs. (e) An example of an image from the DICOM series of a DPP made up of 16,128 (28 sets of ground-truth values * 576 random noise samples) TACs. The top right corner of the phantom image represents the AIF. (f) An example of a blood flow (BF) map obtained from a DPP using commercially available CT perfusion software (syngo.via Body perfusion VB10B, Siemens Healthineers).
