An artificial intelligence algorithm (Aidoc) was used to retrospectively review 14,453 conventional chest CT exams with IV contrast for the presence of incidental pulmonary emboli. All exams were performed as part of a combined CT of the chest, abdomen, and pelvis and read by thoracic radiologists. Natural language processing was used to search the exam reports to find cases of iPE detected prospectively. All cases read as positive for pulmonary emboli by the AI algorithm and NLP search were reviewed by thoracic radiologists to confirm the presence of acute PE. The most proximal level of clot and overall clot burden were assessed during this review.
A total of 254 chest CT exams had acute iPE detected by AI or by review of the radiologist original reports and confirmed by subsequent review for a prevalence of 1.8%. 218 iPE were detected prospectively by radiologists. The sensitivity of radiologists for iPE detection was 0.858. AI detected an additional 36 cases of iPE that were missed prospectively (radiologist false negative rate of 14.2%). 30 cases indicated as positive by the AI algorithm were classified as negative on further clinical review. The AI algorithm had a sensitivity for iPE detection of 0.882 and specificity of 0.998 and accuracy of 0.996. Of the 36 cases of iPE missed by the radiologist, 19 were solitary segmental or subsegmental emboli and the average Qanadli score of clot burden was 1.9. Of the 30 cases of iPE missed by the AI algorithm, one case had large central emboli, but the other cases had small emboli with 23 having solitary subsegmental emboli and an average Qanadli score of 1.5. Tumor/adenopathy adjacent to the pulmonary vessel was the most common cause of AI false positives. Beam hardening artifact, motion artifact, and image noise were also causes of AI false positives.
Acute iPE are sometimes missed during interpretation of conventional chest CT with IV contrast. The use of an AI tool significantly improved detection of iPE which had been missed prospectively. The AI tool had high accuracy with a l ow false positive rate. The iPE missed by the radiologist were small. Further studies of patient outcome are needed to assess the clinical significance of missed untreated iPE.
Aidoc experts, customers and industry leaders share the latest in AI benefits and adoption.
Explore how clinical AI can transform your health system with insights rooted in real-world experiences.
Learn how to go beyond the algorithm to develop a scalable AI strategy and implementation plan.
Explore how Aidoc can help increase hospital efficiency, improve outcomes and demonstrate ROI.
