Performance of three model-based iterative reconstruction algorithms using a ct task-based image quality metric

In this study we evaluated the task-based image quality of a low contrast clinical task for the abdomen protocol (e.g., pancreatic tumour) of three different CT vendors, exploiting three model-based iterative reconstruction (MBIR) levels. We used three CT systems equipped with a full, partial, advanced MBIR algorithms. Acquisitions were performed on a phantom at three dose levels. Acquisitions were reconstructed with a standard kernel, using filtered back projection algorithm (FBP) and three levels of the MBIR. The noise power spectrum (NPS), the normalized one (nNPS) and the task-based transfer function (TTF) were computed following the method proposed by the American Association of Physicists in Medicine task group report-233 (AAPM TG-233). Detectability index (d') of a small lesion (small feature; 100 HU and 5-mm diameter) was calculated using non-prewhitening with eye-filter model observer (NPWE).The nNPS, NPS and TTF changed differently depending on CT system. Higher values of d' were obtained with advanced-MBIR, followed by full-MBIR and partial-MBIR.Task-based image quality was assessed for three CT scanners of different vendors, considering a clinical question. Detectability can be a tool for protocol optimisation and dose reduction since the same dose levels on different scanners correspond to different d' values.