Logo

American Heart Association

  8
  0


Final ID: TP233

Machine Learning Model to Predict Changes in Neuroimaging Profiles Following Delayed Transfer for Stroke Thrombectomy

Abstract Body: Introduction: Patient selection for endovascular thrombectomy (EVT) relies heavily on neuroimaging characteristics. While nearly all stroke patients undergo neuroimaging during acute triage, many patients do not initially present to EVT-capable centers, and interhospital transfer for the EVT procedure can result in significant time delays. Whether patients with prolonged transfer times should undergo repeat imaging prior to EVT is an area of active debate. To guide clinicians on when to re-image EVT-candidates, we sought to construct a machine learning algorithm (MLA) to predict vessel recanalization, ischemia progression, and imaging stability.
Methods: This was a retrospective study of EVT candidates with internal carotid or middle cerebral artery occlusions arriving at a single comprehensive stroke center 1.5 to 6.0 hours from initial neuroimaging. Clinical and radiographic information was collected. A sub-cohort was used to train a gradient-boosted, tree-based MLA to create a model to predict infarct progression, imaging stability, and vessel recanalization on repeat imaging based on initial clinical and imaging characteristics. The performance of the MLA was then independently assessed in an independent validation cohort.
Results: Three hundred seventeen patients were included; 69.4% of patients had stable imaging, 14.5% had ischemia progression (ASPECTS decay of 2 or more), and 16.1% had vessel recanalization. Two hundred twelve were assigned to the training cohort, and 105 to the validation cohort. In the MLA derived from the training cohort, NIH stroke scale improvement, stroke onset to initial imaging time, intravenous thrombolysis administration, initial ASPECTS, and collateral score were the most important predictors for repeat imaging outcomes. In the validation cohort, the MLA had AUCs of 0.81 (95%CI 0.72-0.90), 0.82 (95%CI 0.72-0.91), and 0.89 (95%CI 0.77-1.00) for imaging stability, ischemia progression, and vessel recanalization, respectively. The MLA also had excellent F1 scores (0.87 and 0.95 for stability and no recanalization) and calibration (Brier score of 0.17 and 0.08 for stability and no recanalization, respectively).
Conclusion: An MLA to guide clinical decision making for EVT-candidates who experienced significant transfer delays performed well in predicting imaging stability, and it can be used to identify patients who will not likely benefit from repeat imaging.
  • Skorseth, Paige  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Colasurdo, Marco  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Chen, Huanwen  ( NIH/Georgetown , Washington , District of Columbia , United States )
  • Rewinkel, Scott  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Kim, Daniel  ( Oregon Health and Science University , Portland , Oregon , United States )
  • Amin, Sonesh  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Shakal, Scott  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Priest, Ryan  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Nesbit, Gary  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Clark, Wayne  ( Oregon Stroke Center at OHSU , Portland , Oregon , United States )
  • Author Disclosures:
    Paige Skorseth: DO NOT have relevant financial relationships | Marco Colasurdo: DO NOT have relevant financial relationships | Huanwen Chen: DO NOT have relevant financial relationships | Scott Rewinkel: DO NOT have relevant financial relationships | Daniel Kim: DO NOT have relevant financial relationships | Sonesh Amin: DO NOT have relevant financial relationships | Scott Shakal: DO NOT have relevant financial relationships | Ryan Priest: DO NOT have relevant financial relationships | Gary Nesbit: DO NOT have relevant financial relationships | Wayne Clark: DO NOT have relevant financial relationships
Meeting Info:
Session Info:

Neuroendovascular Posters II

Thursday, 02/06/2025 , 07:00PM - 07:30PM

Poster Abstract Session

More abstracts on this topic:
10-Year Trend Analysis of Medicare Payment in Stroke Inpatient Hospital Admission

Wong Ka-ho, Krothapalli Neeharika, Littig Lauren, Champagne Alison, Majersik Jennifer, Reddy Vivek, De Havenon Adam

A novel reproducible low-cost model of acute myocardial infarction in swine

Li Yichen, Zheng Zilong, Tang Weijie, Chen Wangping, Yang Jinfu, Fan Chengming

More abstracts from these authors:
Large Vessel Recanalization with Thrombolysis Prior to Thrombectomy: Incidence, Outcomes, and Associations With Collateral Status

Skorseth Paige, Colasurdo Marco, Chen Huanwen, Rewinkel Scott, Kim Daniel, Amin Sonesh, Shakal Scott, Priest Ryan, Nesbit Gary, Clark Wayne

Determinants of proceduralist-associated differences in angiographic outcomes following stroke thrombectomy

Chen Huanwen, Gandhi Dheeraj, Colasurdo Marco, Njonkou Tchoquessi Rosy Linda, Iyyangar Ananya, Skorseth Paige, Rewinkel Scott, Majmundar Shyam, Cherian Jacob, Miller Tim, Sheth Sunil

You have to be authorized to contact abstract author. Please, Login
Not Available

Readers' Comments

We encourage you to enter the discussion by posting your comments and questions below.

Presenters will be notified of your post so that they can respond as appropriate.

This discussion platform is provided to foster engagement, and simulate conversation and knowledge sharing.

 

You have to be authorized to post a comment. Please, Login or Signup.


   Rate this abstract  (Maximum characters: 500)