Performance Analysis in Children of Traditional and Deep-Learning CT Lung Nodule Computer-Aided Detection Systems Trained on Adults.
Researchers
Journal
Modalities
Models
Abstract
Background: Although primary lung cancer is rare in children, chest CT is commonly performed to assess for lung metastases in children with cancer. Lung nodule computer-aided detection (CAD) systems have been designed and studied primarily using adult data, and such systems’ efficacy on pediatric patients is poorly understood. Objective: To evaluate the diagnostic performance in children of traditional and deep-learning CAD systems trained with adult data for the detection of lung nodules on chest CT scans and to compare such systems’ ability to generalize to children versus to other adults. Methods: This retrospective study included pediatric and adult chest CT test sets. The pediatric test set comprised 59 CT scans in 59 patients (30 male, 29 female; mean age, 13.1 years; age range, 4-17 year), obtained from November 30, 2018 to August 31, 2020; lung nodules were annotated by fellowship-trained pediatric radiologists as reference standard. The adult test set was the publicly available adult Lung Nodule Analysis (LUNA) 2016 subset 0, containing 89 deidentified scans with previously annotated nodules. The test sets were processed through the FlyerScan (traditional) and Medical Open Network for Artificial Intelligence (MONAI) (deep learning) lung nodule detection CAD systems, which had been trained on separate sets of adult CT scans. Sensitivity and false positive (FP) frequency were calculated for nodules measuring 3-30 mm; non-overlapping 95% CIs indicated significant differences. Results: Operating at two FPs per scan, FlyerScan and MONAI exhibited significantly lower detection sensitivities on pediatric testing data of 68.4% (197/288, 95% CI: 65.1-73.0) and 53.1% (153/288, 95% CI: 46.7-58.4), respectively, than on adult LUNA 2016 subset 0 testing data of 83.9% (94/112, 95% CI: 79.1-88.0) and 95.5% (107/112, 95% CI: 90.0-98.4), respectively. Mean nodule size was smaller (p<.001) in the pediatric (5.4±3.1 mm) than adult (11.0±6.2 mm) testing data. Conclusions: Adult-trained traditional and deep-learning-based lung nodule CAD systems had significantly lower detection sensitivity on pediatric data than on adult data at matching false-positive frequency. The performance difference may relate to smaller size of pediatric lung nodules. Clinical Impact: The results indicate a need for pediatric-specific lung nodule CAD systems trained on pediatric-specific data.