Computer-aided detection of small pulmonary nodules in multidetector spiral computed tomography (MSCT) in children

Purpose: Retrospective evaluation of computer-aided detection software (CAD) for automated detection (LungCAD, Siemens Medical solutions, Forchheim, Germany) and volumetry (Lung-CARE) of pulmonary nodules in dose-reduced pediatric MDCT. Materials and Methods: 30 scans of 24 children (10.4 +/- 5.9 years, 13 girls, 11 boys, 39.7 +/- 29.3 kg body weight) were performed on a 16-MDCT for tumor staging (n=18), inflammation (n=9), other indications (n=3). Tube voltage 120 kVp and effective mAs were adapted to body weight. Slice thickness 2 mm, increment 1 mm. A pediatric radiologist (U1), a CAD expert (U2) and an inexperienced radiologist (U3) independently analyzed the lung window images without and with the CAD as a second reader. In a consensus decision U1 and U2 were the reference standard. Results: Five examinations had to be excluded from the study due to other underlying lung disease. A total of 24 pulmonary nodules were found in all data sets with a minimal diameter of 0.35 mm to 3.81 mm (mean 1.7 +/- 0.85 mm). The sensitivities were as follows: U1 95.8 % and 100 % with CAD; U2 91.7% U3 66.7%. U2 and U3 did not detect further nodules with CAD. The sensitivity of CAD alone was 41.7% with 0.32 false-positive findings per examination. Interobserver agreement between U1/U2 regarding nodule detection with CAD was good (k = 0.6500) and without CAD very good (k = 0.8727). For the rest (U1/U3; U2/U3 with and without CAD), it was weak (k=0.0667-0.1884). Depending on the measured value (axial measurement, volume), there is a significant correlation (p=0.0026-0.0432) between nodule size and CAD detection. Undetected pulmonary nodules (mean 1.35 mm; range 0.35 - 2.61 mm) were smaller than the detected ones (mean 2.19 mm; range 1.35 - 3.81 mm). No significant correlation was found between CAD findings and patient age (p=0.9263) and body weight (p=0.9271) as well as nodule location (subpleural, intraparenchymal; p = 1.0) and noise/SNR. Conclusion: In our study with 2 mm slice thickness and very small lesion sizes, the analyzed CAD algorithm for detection and volumetry of pulmonary nodules has limited application in pediatric dose-reduced 16-MDCTs. Determination of lesion size is possible even in the case of false-negatives.