Artificial intelligence matches radiologists at predicting patients’ emphysema severity
Artificial intelligence tools tested by Massachusetts General Hospital were able to match trained radiologists at predicting patients’ emphysema severity, according to new research published Thursday.
Visual assessments of this clinical concern and bronchial abnormality can be subjective and prone to variation. Scientists at the Boston-based institution set out to use automation for lung assessment, hoping to eliminate discrepancies and bolster chest CT’s reproducibility at determining chronic obstructive pulmonary disease severity.
Their analysis included 113 adults who received both a noncontract CT scan and pulmonary function test performed within a two-month span. Disease severity was classified as mild, moderate or severe, measured by how much air patients could exhale during a forced breath. And two trained radiologists (blinded to the AI and clinical results), graded each individual’s emphysema on a five-point scale, experts wrote in Academic Radiology.
“Performance of the assessed AI prototypes was similar to radiologists’ subjective assessment of emphysema severity and bronchial airways,” Mannudeep Kalra, MD, with Mass General’s Department of Radiology, and colleagues wrote Oct. 14. “Although outputs from the assessed AI prototypes were robust across different scanner types, scan parameters, and reconstruction settings, there is lack of consistent performance across patients with larger body habitus due to inadequate radiation dose,” the authors cautioned.
Patients in the retrospective study underwent chest CT between 2018-2020 at a single healthcare institution, receiving a diagnosis of COPD. CT scanners from three different vendors were used in the study, while the images were processed with the AI-Rad Companion Chest prototype from Siemens Healthineers. Trained radiologists and a fully automated software from Philips assessed bronchial abnormality.
Artificial intelligence and the thoracic radiologists were able to differentiate mild, moderate and severe emphysema based on forced expiratory volume, with area under the curve scores of 0.77 and 0.76, respectively. They noted a strong positive correlation between the two. And the combo of emphysema and bronchial abnormality quantification from radiologist and AI assessments could differentiate between different severities (AUC of 0.80-0.82 and 0.87, respectively).
Marty Stempniak has covered healthcare since 2012, with his byline appearing in the American Hospital Association's member magazine, Modern Healthcare and McKnight's. Prior to that, he wrote about village government and local business for his hometown newspaper in Oak Park, Illinois. He won a Peter Lisagor and Gold EXCEL awards in 2017 for his coverage of the opioid epidemic.