Diagnostic Feasibility and Physical Assessment of Lung CT Images Reconstructed From Ultra-Low-Dose Positioning Scan for Computer-Aided Nodule Detection.

Authors

Abstract

To evaluate the diagnostic feasibility of computer-aided detection (CAD) of pulmonary nodules on CT images reconstructed from volumetric ultra-low-dose (ULD) positioning scans compared with standard-dose CT. In this retrospective study, 160 adults with at least 1 pulmonary nodule underwent volumetric ULD positioning scans and standard-dose CT. One target nodule per patient was analyzed. CAD sensitivity was evaluated on 5-mm volumetric ULD positioning scans and 5-mm standard-dose CT according to nodule size, morphology, and malignancy. An additional analysis was performed using 1-mm volumetric ULD positioning scans, with 1-mm standard-dose CT serving as the reference standard. Physical image characteristics were assessed using phantoms. A total of 160 nodules (median size: 5 mm) were analyzed. CAD sensitivity was 71.3% (114/160) for 5-mm volumetric ULD positioning scans and 85.0% (136/160) for 5-mm standard-dose CT (P<0.01), with reduced sensitivity most apparent for small nodules. With 1-mm volumetric ULD positioning scans, sensitivity improved to 88.8% (142/160). However, noninferiority to reference 1-mm standard-dose CT was not demonstrated using a 15-percentage-point margin (P=0.084). The median estimated effective dose was 0.18 mSv for volumetric ULD positioning scans and 4.67 mSv for standard-dose CT. Phantom assessment showed lower spatial resolution and increased low-frequency noise for volumetric ULD positioning scans. Thinner-slice reconstruction improved CAD sensitivity on volumetric ULD positioning scans; however, noninferiority to standard-dose CT was not demonstrated. Further optimization of image quality and CAD performance is warranted.

Tags

Topics