Purpose: This study aimed to evaluate image quality of ultra-low dose chest computed tomography using 6 iterative reconstruction (IR) algorithms.
Method: A lung phantom was scanned on 4 computed tomography scanners using fixed tube voltages and the lowest mAs available on each scanner, resulting in dose levels of 0.1 to 0.2 mGy (80 kVp) and 0.3 to 1 mGy (140 kVp) volume CT dose index (CTDIvol). Images were reconstructed with IR available on the scanners. Image noise, signal-to-noise ratios, contrast-to-noise ratios, uniformity, and noise power spectrum (NPS) were assessed for evaluation of image quality.
Results: Image quality parameters increased with increasing dose for all algorithms. At constant dose levels, model-based techniques improved the contrast-to-noise ratio of lesions more than the statistical algorithms. All algorithms tested at 0.1 mGy showed lower NPS peak frequencies compared with 0.39 mGy. In contrast to the statistical techniques, model-based algorithms showed lower NPS peak frequencies at the lowest doses, indicating a coarser and blotchier noise texture.
Conclusion: This study shows the importance of evaluating IR when introduced clinically.