CT- DImQ: An open-access platform for image quality assessment of CT systems- application to CT values and noise...

To present an open access platform for CT image quality assessment combined with anthropomorphic phantoms. To promote the accessibility of anthropomorphic phantoms, including 3D-printed affordable ones, in regular QC analysis of CT images reconstructed with filtered back projection (FBP), iterative (IR) and Deep-Learning (DLR) reconstruction algorithms.

The Python-based platform CT-DImQ enables the advanced evaluation of CT values distribution through histogram analysis and noise characterization (noise maps and noise power spectrum (NPS)). The interactive GUIs allow the user to perform the CT data analysis adapting the output to their own requirements. The use case for CT-DImQ was CT anthropomorphic phantom data (thorax phantom with 3D-printed TPU lung vessel trees and PA-12 nodules) obtained under clinical conditions and two dose levels reconstructed with FBP, IR and DLR.

Histograms showed in general an underestimation of CT-values for the 3D printed materials and a overestimation for Teflon (spine surrogate), in DLR images, while air and PMMA-soft tissue were the same as with IR and FBP. Noise maps showed a strong noise reduction in uniform phantom parts, and to a lesser extent in vessel and nodules edges. NPS showed noise reduction with increasing dose, especially with FBP and iterative reconstruction, and differences in the noise reduction frequency dependence among reconstruction methods, which was corroborated visually by the more uniform appearance of vessels and nodules in DLR images.

CT-DImQ combined with anthropomorphic phantoms represents a step towards advanced image quality evaluation in CT, closer to the clinical reality with patients and able to evaluate advanced reconstruction methods trained with patient data.