12171
clinical study

Retrospective Evaluation of an Artificial Intelligence (AI) System for Enhanced Detection of Pulmonary Embolism and Potential Clinical Impact Using Chart Review

Materials & Methods

This study evaluated the effectiveness of integrating artificial intelligence into the diagnosis of pulmonary embolism (PE) using CT angiograms (CTA). The goal was to identify PE cases that might be overlooked by radiologists, assess how AI could improve detection rates and explore the potential impact on patient care.

Results

Analyzing 1,471 CTAs using the Aidoc AI algorithm, detection rates of AI and radiologists were compared. Discrepancies between the two were reviewed by a panel of two thoracic radiologists. The results revealed that radiologists initially detected PE in 10.1% of cases, while AI identified PE in 11.7% of cases, including 10 additional cases missed by radiologists. This led to a 6.8% enhanced triage rate. Notably, all missed PEs were smaller segmental or subsegmental, and no central PEs were overlooked. Further review of patient charts found no negative clinical outcomes or subsequent PE diagnoses associated with these missed cases.

Conclusions

The findings suggest that AI can help radiologists significantly improve PE awareness and potentially enhance patient outcomes, though further research with follow-up imaging may provide more insights into the clinical impact of missed diagnoses.

Explore the Latest AI Insights, Trends and Research