Downsampling CT Perfusion Imaging for Pancreatic Cancer: A Blood Flow and Cell Density Analysis
Abstract
Purpose
Desmoplasia around pancreatic cancer reduces blood perfusion and contrast distribution volume (DV), a biomarker of cell density. This study investigates a low-frequency clinical computed tomography perfusion (CTP) protocol to quantify treatment-induced increases in perfusion and decreases in DV. By extending temporal sampling from 1–3s intervals to 7–10s intervals while maintaining comparable accuracy, this approach enables integration of CTP into standard three-phase CT imaging for pancreatic cancer, avoiding additional contrast administration and radiation dose.
Methods
Five patients with borderline-resectable pancreatic cancer (BRPC) were treated with chemotherapy and Stereotactic Ablative Body Radiotherapy (SABR) with 27–30Gy/3fx. CTP scans were acquired at baseline, post-chemotherapy, post-SABR (6hrs after 1st fraction and 3-4wks after completing SABR). Contrast agent was administered, and scans were taken every 2.5s over 1 min during the vascular phase, followed by scans every 15s for 2 mins, comprising the interstitial phase–totaling 32 scans per timepoint. A deconvolution-based CTP software (GE HealthCare, Waukesha, US) generated parametric maps for blood flow (BF) and DV. The vascular phase was downsampled to 50%, 33%, and 25% of the native 0.4Hz sampling frequency for each patient. Regions of interest (ROIs) were drawn at the hypoxic core of the tumour, the rim around the core, and the remainder of the tumour over 6-7 slices. BF and DV values for each sampling scheme were analyzed using linear regression against the fully sampled study results.
Results
Among the five patients analyzed, the 50% sampling scheme demonstrated the strongest blood flow and distribution volume correlation with the fully sampled scheme (R2BF=0.94 and R2DV=0.87), compared to the 33% (R2BF=0.78 and R2DV=81) and 25% sampling schemes (R2BF=0.86 and R2DV=0.79).
Conclusion
Our study demonstrates scanning intervals of 5–7.5s may be acceptable relative to 2.5s to maintain satisfactory accuracy in estimated BF and DV in pancreatic lesions.