AI Tool Aids in Detection of Incidental Pulmonary Embolism and Reduces Time to Diagnosis in Cancer Patients

A recent study published in Radiology has shown that the use of artificial intelligence (AI) software in detecting incidental pulmonary embolism (IPE) at CT can significantly reduce the time to diagnosis and improve diagnostic accuracy for cancer patients. The research, led by Laurens Topff and a team of experts from multiple institutions, evaluated the diagnostic efficacy of AI software in prioritizing radiologist reading worklists and shortening the time to diagnosis of IPE.

The research team, comprising specialists in radiology, oncology, and AI technology, has vast experience in the development and evaluation of AI tools for various medical applications. The AI software used in the study is a regulatory-cleared application designed to automatically detect and prioritize IPE on routine chest CT scans with intravenous contrast agent in adult oncology patients.

The study included 11,736 CT scans of 6,447 adult oncology patients and assessed detection and notification times (DNTs) during three time periods: routine workflow without AI, human triage without AI, and worklist prioritization with AI. The prevalence of IPE was found to be 1.3%, 1.4%, and 1.0% for the respective time periods. The AI software demonstrated 91.6% sensitivity, 99.7% specificity, 99.9% negative predictive value, and 80.9% positive predictive value.

During the prospective evaluation, AI-based worklist prioritization reduced the median DNT for IPE-positive examinations to 87 minutes, compared to 7,714 minutes for routine workflow and 4,973 minutes for human triage. The radiologists' missed rate of IPE was significantly reduced from 44.8% without AI to 2.6% when assisted by the AI tool (P < .001).

The use of AI software in detecting IPE at chest CT in cancer patients showed high diagnostic accuracy and significantly shortened the time to diagnosis, particularly in settings with a backlog of examinations. This study is part of a growing body of research exploring the potential benefits of AI-assisted workflows in radiology departments. Previous studies have investigated AI-based prioritization tools for detecting intracranial hemorrhage at CT, acute pathologic abnormalities on chest radiographs, and pulmonary embolism on dedicated CT pulmonary angiograms, yielding varying results.

The findings of the current study not only demonstrate the capability of AI to improve patient care through faster diagnosis and treatment but also pave the way for further research into the development and implementation of AI technologies in radiology and other medical fields.