Hemostasis Today

October, 2026
October 2026
M T W T F S S
 1234
567891011
12131415161718
19202122232425
262728293031  
Hari Trivedi: Real-World AI Performance in PE Detection
Oct 27, 2026, 04:09

Hari Trivedi: Real-World AI Performance in PE Detection

Hari Trivedi, Advisor of Croviz, Associate Professor of Radiology, Biomedical Informatics, and Emergency Medicine at Emory University, shared a post on LinkedIn about a recent article he and his colleagues co-authored, published in Nature Digital Medicine, adding:

“Happy to share our recent work which was just accepted in Nature Digital Medicine on real-world performance of a commercial AI model for pulmonary embolism detection.

We tested two FDA-cleared algorithms across 17 Emory facilities for triaging PE and incidental PE across 65,000+ exams and built a custom LLM-pipeline to retrieve important subgroups like PE acuity, depth, and right heart strain from our reports.

As with our work on intracranial hemorrhage, we found that sensitivity varied greatly from 65% to 99% depending on the type of PE, and suffered most on subtle cases that radiologists find difficult.

This again highlights the need for a deeper understanding of model performance across subgroups in real-world environments, as this directly affects if and how radiologists should rely on these models.

Congrats to co-first authors Aawez Mansuri and Mohammadreza Chavoshi, and to the whole HITI Lab team.”

Title: Multi-site real-world performance of commercial AI for pulmonary and incidental pulmonary embolism detection

Authors: Aawez Mansuri, Mohammadreza Chavoshi, Theodorus Dapamede, Wasif Bala, Beatrice Brown-Mulry, Rohan Satya Isaac, Bardia Khosravi, Hanzhou Li, Frank Li, John T. Moon, Chad Robichaux, Dan I. G. Cohen-Addad, Ninad V. Salastekar, Janice Newsome, Judy W. Gichoya, Hari Trivedi

Hari Trivedi

Stay updated on all scientific advances with Hemostasis Today.