Quantifying quality: correlation phantom and clinical scan CT noise levels to contruct a Quality Reference Level
Conclusion This study represents a proof of concept method to develop and refine a Noise Reference Level for a given protocol and scanner. This represents a first step in establishing an automated Performance Reference Level for a given scan protocol that incorporates both radiation dose and image quality. Background The use of ionizing radiation in computerized tomography mandates risk-benefit understanding. However, this consideration may weight the presumed risk of the ionizing radiation at the expense of the benefits of the examination, with protocol optimization often focusing on dose reduction to the potential detriment of image quality, hence diagnostic value. In fact, the standardized Diagnostic Reference Levels (DRL) are based solely on dose metrics. The purpose of this study is to begin to systematically evaluate an efficient tool for image quality, an automated noise metric in CT to eventually develop a performance metric consisting of both radiation dose and image quality. Evaluation Using an IRB-exempt protocol, automated noise values were determined for 1904 contrast enhanced abdominopelvic (AP) and unenhanced chest CT scans on two different scanner models. These data were used to calculate a Noise Reference Level (NRL) by calculating a noise median across a 25-35 cm patient size range, defining the NRL as this median ±20%. A variable diameter phantom was also scanned utilizing both protocols on both scanners, and the resultant noise-size curves for the phantom were compared to those utilizing clinical scan data. Discussion Utilizing our novel noise evaluation tool, noise-patient size curves demonstrate agreement between clinical scan and phantom data. Noise measurements were varied based on the region scanned, with the NRL interval between 25 and 37 HU for AP CT, and between 10 and 15 HU for chest CT. This demonstrates the utility of noise monitoring across large numbers of scans to generate a target noise range based on region scanned and scanner equipment, and to identify and evaluate image quality outliers. This automated tool can thus be used to compare performance in a single scanner and between scanners for similar sized patients independent of interpreting radiologist subjective biases.
