IQ-PAT: An innovative tool for noise power spectrum and task-based transfer function assessment in patient CT images

To develop a new tool (IQ-PAT) for noise power spectrum (NPS) and task-based transfer function (TTF) evaluation on patient images and validate this tool on an image quality phantom.

IQ-PAT was developed to compute automatically NPS and TTF in the thoracic aorta. First aorta mask and mesh were generated by a segmentation tool. Regions of interest (ROIs) were drawn on 100 slices in order to compute NPS. For TTF assessment, the barycenter of the aorta on 100 consecutive slices was calculated and used to extract and align radial profiles for Edge Spread Function (ESF) computation. Profiles affected by anatomical variability or pathology were automatically excluded. TTF was then derived by differentiating the ESF and applying the Fourier Transform. Validation was done on a Mercury v4.0 phantom (26 and 31 cm section, 14.5 mGy, 120 kVp, gated aortic protocol). Images were reconstructed with iDose4 (level 4-6), 1 mm slices, and kernels A-C. IQ-PAT was compared to ImQuest and applied to 20 patients.

Phantom data showed that NPS and TTF values from IQ-PAT closely matched those from ImQuest. For each combination of reconstruction kernel and phantom diameter, noise magnitude, average spatial frequency (f_av) and TTF at 50% (f₅₀) values were not significantly different between IQ-PAT and ImQuest (all p > 0.05). In patients, IQ-PAT exhibited trends consistent with phantom results across reconstruction settings.

IQ-PAT provided reliable and consistent assessments of NPS and TTF on phantom images, with comparable results to ImQuest. Early application to patient data demonstrated the method's feasibility in clinical conditions.