When used to flag anomalies in combined chest and musculoskeletal X-rays, imaging AI can relieve overstretched emergency radiology teams.
In the process, the software can reduce misdiagnosis rates, according to a study published online in the European Journal of Radiology Open [1].
Radiologist Alexandre Parpaleix, MD, PhD, and colleagues at Valenciennes General Hospital in France reviewed 1,772 cases of patients who received emergency X-rays of any body part except for spine, skull and abdomen.
The researchers had a senior MSK-specialized radiologist adjudicate 172 cases (9.7% of the sample) that had discordances between first reads from ED physicians and final reads from the radiology department.
The umpiring MSK radiologist had access to all relevant clinical records.
Meanwhile the team used commercially available AI software to triage patients based on the X-rays and see how it handled the cases with discrepant readings.
They found the AI effectively tied the ED physicians on sensitivity and achieved 90.1% accuracy on the 172 cases that were misdiagnosed by those same readers.
In their discussion, the authors note that the tested AI model performed similarly to those used in previous studies. However, they add, the present research may have been the first to combine MSK and chest X-rays.
Even while excluding spine, skull and abdomen imaging, the MSK/chest combination allowed for a wider range of cases to be covered in the radiographic workflow, Parpaleix and co-authors state.
In addition, they found no significant difference in the AI’s performance across age and body-part subgroups. This “is of importance for the widespread use of AI in the daily clinical environment,” the authors write.
They call the technology’s potential for reducing misdiagnosis in the ED “promising,” underscoring that the AI and emergency physicians did not fail on the same cases, “highlighting their expected synergy.”
More:
Given the high negative predictive value of the AI solution and the high recall rate for misdiagnosis, we foresee the implementation of an AI-based 24/7/365 emergency radiographic workflow without replacing radiological interpretation but rather securing patient management through its use as both a rule-out (triage) and a computer-aided diagnosis (add-on) solution.”
Building on this scenario, Parpaleix et al. suggest prioritizing doubtful exams for STAT radiologist review.
“Under this hypothesis, AI could alleviate emergency radiographic workflow constraints,” they write.
The study is available in full for free.
Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.