Journal: Scientific Reports
Article Title: Model based noise correction enhances the accuracy of pancreatic CT perfusion blood flow measurements
doi: 10.1038/s41598-025-24482-x
Figure Lengend Snippet: A flow chart illustrating the complete noise correction process for blood flow (BF) measurements, including evaluation using digital perfusion phantom (DPP) (starting from the first block) and evaluation using a clinical dataset (starting from the second block). For the DPP analysis, each GTBF value is simulated using two independent sets of 576 noise-impacted TACs, resulting in BF1 and BF2 estimates for random error calculation. This process is repeated for 28 GTBF values, totaling 16,128 TACs. For the clinical dataset, patient BF values calculated using the deconvolution model from Mayer’s study were used as input for the noise-impacted BF maps. BFD represents the noise-impacted BF measurements, which need to be corrected. IRF is the impulse response function, AIF is the arterial input function, TAC represents the tissue attenuation curve, and GTBF is the ground-truth blood flow. BFD corr (i) represents the noise-corrected BF measurement for the i th iteration. The random error and model error calculations are also shown in the flow chart. This iterative process for DPP continues until BFD corr aligns with GTBF or until the error between GTBF and corrected measurements is minimized to an acceptable threshold.
Article Snippet: All evaluations in this study were performed using a commercial deconvolution model (syngo.via, Siemens Healthineers) with fixed reconstruction parameters such as slice thickness, reconstruction kernel, and matrix size selected to reflect standard clinical CTp practice.
Techniques: Blocking Assay
