Skip to main content
SIR 2025 Event Main banner
BSCI - Consideration Cases Ad 1
Merit/DPSSME Siege
Back

SIR 2025

Venous Interventions

Scientific Session

Session: AI, Machine Learning, Simulation, and Imaging 1

Time Will Tell: An Analysis of Exam-to-Needle Time Following Phased Implementation of AI Diagnosis and Mobile Alert of Acute Pulmonary Embolism

Monday, March 31, 2025
4:12 PM – 4:21 PM CT
Location: Music City Center, 205 A

    Presenting Author(s)

  • Jacob T. Ramsey, MD

    Resident Physician, PGY4
    Jefferson Einstein Hospital, United States

    • Author/Co-author(s)

  • MD

    Mark DiMaggio, DO

    Resident Physician
    Jefferson Einstein Hospital, United States

  • Jung H. Yun, MD

    IR/DR Resident
    Jefferson Einstein Philadelphia Hospital, United States

  • AS

    Avi Sharma, MD

    Director of AI & Body Radiologist
    Jefferson Einstein Hospitals, United States

  • MC

    Matthew Callaghan, MD

    Attending Interventional Radiologist
    Einstein Medical Center, Jefferson Health, United States

Purpose:

This study compares the exam-to-needle time for acute pulmonary embolism (PE) treated with percutaneous thrombectomy before and after introducing an AI mobile alert.


Materials and Methods:

A single-center retrospective review, comparing patients who underwent percutaneous thrombectomy for acute PE before and after AI Mobile Alert (1/2019–6/2020 and 01/2023–4/2024, respectively). Subcategories were intermediate-high and high-risk patients, based on imaging and clinical criteria. The AI solutions used were CT-based PE detection, right ventricle-to-left ventricle ratio analysis, and a mobile-based alert to notify PERT physicians of actionable findings (Aidoc, Tel Aviv). The primary endpoint was exam-to-needle time, defined as the time between the start of the diagnostic CT and percutaneous thrombectomy. Exclusion criteria included study-to-needle time exceeding 4 hours (deemed nonurgent cases based upon hemodynamic stability and lab trends) and PE not diagnosed by CT. Wilcoxon rank sum and Kruskal-Wallis tests were used for statistical analysis. 

Results:

A total of 77 patients were treated: 38 in the Pre-Mobile Alert group and 39 in the Post-Mobile Alert group. The average exam-to-needle time was 148 minutes pre-alert versus 119 minutes post-alert (p = 0.028). Risk-stratified analysis revealed the following exam-to-needle times (Pre vs. Post): High-Risk (156 vs. 119, p = 0.086) and Intermediate-High Risk (143 vs. 125, p = 0.460) but there was no significant difference by severity or between all groups combined (p = 0.148). 

Conclusion:

Introduction of the mobile alert significantly reduced exam-to-needle time for all-comer patients undergoing PE thrombectomy. While analysis of risk-stratified subcategories did not demonstrate a statistically significant improvement in exam-to-needle time, there was a trend towards a faster response. Ultimately, we found the timely alert of IR physicians by the Mobile Alert improved synergy amongst Radiology, Emergency and Intensive Care departments in our institution during the timely care of acute PE. Limitations include concurrent advancements in thrombectomy techniques and a small patient cohort.