Ensuring Adequate Development and Appropriate Use of Artificial Intelligence in Pediatric Medical Imaging

Diagnostic performance of artificial intelligence for pediatric pulmonary nodule detection in computed tomography of the chest

PURPOSE

To test the performance of a commercially available adult pulmonary nodule detection artificial intelligence (AI) tool in pediatric CT chests.

METHODS

30 consecutive chest CTs with or without contrast of patients ages 12-18 were included. Images were retrospectively reconstructed at 3 mm and 1 mm slice thickness. AI for detection of lung nodules in adults (Syngo CT Lung Computer Aided Detection (CAD)) was evaluated. 3 mm axial images were retrospectively reviewed by two pediatric radiologists (reference read) who determined the location, type, and size of nodules. Lung CAD results at 3 mm and 1 mm slice thickness were compared to reference read by two other pediatric radiologists. Sensitivity (Sn) and positive predictive value (PPV) were analyzed.

RESULTS

The radiologists identified 109 nodules. At 1 mm, CAD detected 70 nodules; 43 true positive (Sn = 39 %), 26 false positive (PPV = 62 %), and 1 nodule which had not been identified by radiologists. At 3 mm, CAD detected 60 nodules; 28 true positive (Sn = 26 %), 30 false positive (PPV = 48 %) and 2 nodules which had not been identified by radiologists. There were 103 solid nodules (47 measuring < 3 mm) and 6 subsolid nodules (5 measuring < 5 mm). When excluding 52 nodules (solid < 3 mm and subsolid < 5 mm) based on algorithm conditions, the Sn increased to 68 % at 1 mm and 49 % at 3 mm but there was no significant change in the PPV measuring 60 % at 1 mm and 48 % at 3 mm.

CONCLUSION

The adult Lung CAD showed low sensitivity in pediatric patients, but better performance at thinner slice thickness and when smaller nodules were excluded.

Artificial intelligence CAD tools in trauma imaging: a scoping review from the American Society of Emergency Radiology (ASER) AI/ML Expert Panel

BACKGROUND

AI/ML CAD tools can potentially improve outcomes in the high-stakes, high-volume model of trauma radiology. No prior scoping review has been undertaken to comprehensively assess tools in this subspecialty.

PURPOSE

To map the evolution and current state of trauma radiology CAD tools along key dimensions of technology readiness.

METHODS

Following a search of databases, abstract screening, and full-text document review, CAD tool maturity was charted using elements of data curation, performance validation, outcomes research, explainability, user acceptance, and funding patterns. Descriptive statistics were used to illustrate key trends.

RESULTS

A total of 4052 records were screened, and 233 full-text articles were selected for content analysis. Twenty-one papers described FDA-approved commercial tools, and 212 reported algorithm prototypes. Works ranged from foundational research to multi-reader multi-case trials with heterogeneous external data. Scalable convolutional neural network-based implementations increased steeply after 2016 and were used in all commercial products; however, options for explainability were narrow. Of FDA-approved tools, 9/10 performed detection tasks. Dataset sizes ranged from < 100 to > 500,000 patients, and commercialization coincided with public dataset availability. Cross-sectional torso datasets were uniformly small. Data curation methods with ground truth labeling by independent readers were uncommon. No papers assessed user acceptance, and no method included human-computer interaction. The USA and China had the highest research output and frequency of research funding.

CONCLUSIONS

Trauma imaging CAD tools are likely to improve patient care but are currently in an early stage of maturity, with few FDA-approved products for a limited number of uses. The scarcity of high-quality annotated data remains a major barrier.