AI for fracture diagnosis in clinical practice: Four approaches to systematic AI-implementation and their impact on AI-effectiveness

Purpose: Artificial Intelligence (AI) has been shown to enhance fracture-detection-accuracy, but the most effective AI-implementation in clinical practice is less well understood. In the current study, four approaches to AI-implementation are evaluated for their impact on AI-effectiveness. Materials and Methods: Retrospective single-center study based on all consecutive, around-the-clock radiographic examinations for suspected fractures, and accompanying clinical-practice radiologist-diagnoses, between January and March 2023. These image-sets were independently analysed by a dedicated bone-fracture-detection-AI. Findings were combined with radiologist clinical-practice diagnoses to simulate the four AI-implementation methods deemed most relevant to clinical workflows: AI-standalone (radiologist-findings not consulted); AI-problem-solving (AI-findings consulted when radiologist in doubt); AI-triage (radiologist-findings consulted when AI in doubt); and AI-safety net (AI-findings consulted when radiologist diagnosis negative). Reference-standard diagnoses were established by two senior musculoskeletal-radiologists (by consensus in cases of disagreement). Radiologist- and radiologist + AI diagnoses were compared for false negatives (FN), false positives (FP) and their clinical consequences. Experience-level-subgroups radiologists-in-training-, non-musculoskeletal-radiologists, and dedicated musculoskeletal-radiologists were analysed separately. Results: 1508 image-sets were included (1227 unique patients; 40 radiologist-readers). Radiologist results were: 2.7 % FN (40/1508), 28 with clinical consequences; 1.2 % FP (18/1508), 2 received full-fracture treatments (11.1 %). All AI-implementation methods changed overall FN and FP with statistical significance (p < 0.001): AI-standalone 1.5 % FN (23/1508; 11 consequences), 6.8 % FP (103/1508); AI-problem-solving 3.2 % FN (48/1508; 31 consequences), 0.6 % FP (9/1508); AI-triage 2.1 % FN (32/1508; 18 consequences), 1.7 % FP (26/1508); AI-safety net 0.07 % FN (1/1508; 1 consequence), 7.6 % FP (115/1508). Subgroups show similar trends, except AI-triage increased FN for all subgroups except radiologists-in-training. Conclusion: Implementation methods have a large impact on AI-effectiveness. These results suggest AI should not be considered for problem-solving or triage at this time; AI standalone performs better than either and may be a source of assistance where radiologists are unavailable. Best results were obtained implementing AI as safety net, which eliminates missed fractures with serious clinical consequences; even though false positives are increased, unnecessary treatments are limited.