The study examines a diverse population with a near-equal gender distribution (66% males, 34% females) and an age range of 19 to 80 years, averaging 39 years. This inclusivity enhances the generalizability of findings. Clinical data reveals that lower limb fractures dominate, comprising 53.7% of cases, particularly ankle fractures (21.7%), while upper limb fractures account for 46.3%, with hands being most common (18%). AI’s performance in fracture detection was compared with CT scans, where it identified 121 fractures among 300 cases, showing a sensitivity of 91.13% and specificity of 95.45%. The AI tool’s overall accuracy stood at 93.67%, indicating its reliability in diagnosing fractures. However, a significant performance gap between AI and CT scans was noted. Importantly, 5.7% of AI cases were flagged as uncertain, emphasizing the necessity for radiologist verification. These insights affirm the potential of AI in clinical settings while advocating for human oversight in diagnosis.
Source link
