Evaluation of Image Quality of State-of-art CT Vendors in the Norwegian Market

Abstract

This study focus in the low-contrast detectability properties of two scanners examples in the state-of-the-art from Oslo Hospital Interventional center, as examples of different technologies, being one of them Dual-Energy source CT. We considered different noise filter possibilities and several iterative reconstruction options for the images from both vendors.

It is phantom-based (so does not use real patients) but it considers some of its related challenges, as it is the overweight trend in population. Beside the increased difficulty for and adequate image quality scan, this is an important risk factor in medicine. Several rings where added to phantom at each of the CT scans protocols used.

Image quality in diagnostics, refers to Diagnostic quality. First we considered some mathematical metrics and noise studies, and pursued measurements of Contrast to Noise Ratio.

Later, we wanted to check how this parameter relaters to observers perception, and we pursued several experiments. We took special attention onto the confidence intervals and the standard deviation value among observers.

The goal is to deeply study the acquisition methods of CT devices, then consider the workflow of data from CT to computers (at the selection of useful images), and finally, to have an insight onto the factors that can affect the transition from “vision to decision” at the medical sector.

Two different interpretations of the same image, could lead onto a different diagnose by a Doctor. This study explores the differences between the answers of Image Processing students, and Radiography students, and the distribution of data according to a normal (Gaussian) distribution, through some statistical methods as Mann-Whitney U-Test, Anderson Darling Test, and Student t-Test.

Results presented here could contribute to some dose reduction decisions, that is the one of main dilemmas in Radiology nowadays: To obtain an adequate image quality for a sucessful diagnose, while keeping dose contribution ALARA (as low as reasonably achievable).

In terms of low contrast detectability, Toshiba Acquilion performance is good for the phantom without ring, while the iterative reconstruction algorithm VEO by GE Works better for the phantom with medium rings, with the drawback of a higher computational time. For the big rings, results depend on the particular dose level used. Exploration of new AiDR-3D image reconstruction by Toshiba is advisable for future work.