12173
clinical study

Hospital-Wide Integration of a Natural Language Processing Algorithm To Detect Inferior Vena Cava Filters in Imaging Reports and Improve Device Removal Rates

Materials & Methods

This study explored the potential of an AI-powered natural language processing (NLP) algorithm to identify retrievable inferior vena cava (IVC) filters in imaging reports and improve retrieval rates. The AI algorithm (Aidoc) was used to analyze CT and X-ray reports over a two-month period in 2022 and 2023.

Results

A total of 30,745 reports from 2022, and 17,099 from 2023, detected IVC filters in 107 and 80 patients, respectively. The algorithm demonstrated 100% specificity for filter detection. In 2022, 61% of detected filters were eligible for removal at the time of imaging, but only 21% had been removed within a year. In 2023, 58% of filters were eligible for retrieval, with a median dwelling time of 4.2 years.

Conclusions

With the low rates of IVC filter retrieval, AI-driven NLP demonstrated effectiveness in identifying eligible patients. As a quality improvement initiative, the study plans to follow-up with patients identified by the AI algorithm in 2023 to encourage them to schedule filter removal evaluations and assess the impact on retrieval rates.

Explore the Latest AI Insights, Trends and Research