Studies have suggested that AI-based computer-aided detection (CAD) for chest x-rays has equal or better performance than radiologists and improves radiologist interpretations when used as a second reader. But these studies have often been conducted using datasets collected retrospectively, arbitrarily selected and often disease-rich and dichotomized in composition, resulting in consequent selection bias, noted Dr. Ju Gang Nam, a radiologist at the Seoul National University Hospital (SNUH). Reader performance tests are also conducted retrospectively with a similar performance bias, and proper integration of AI-CAD with conventional PACS is often lacking, according to Nam.
“We sought to determine whether commercial AI-based software can increase the detection rate of clinically significant and actionable lung nodules on chest x-rays in a health check-up population through a randomized controlled trial. this prospective trial we integrated AI-CAD into a commercial PACS (AI-PACS) and incorporated it into the real clinical work process,” Nam told meeting attendees. “AI-based software has improved the detection of clinically significant lung nodules on chest X-rays without increasing the false recommendation rate.”
In this single-center, double-arm, open-label clinical trial, 11,062 X-rays were taken from 10,476 people at the SNUH Health Care Screening Center. The subjects were all adults over the age of 18 who came to the medical screening center for a chest X-ray examination.
After X-ray acquisition, images were simultaneously sent to the AI-CAD server (Lunit INSIGHT CXR version 126.96.36.199) and AI-PACS server (Infinitt Healthcare), with the AI-PACS server randomizing patients 1:1 to one of the two groups; the experimental arm (IA group; subject n=5238, radiograph n=5549) or the control arm (non-IA group; subject n=5238, radiograph n=5513).
One of three designated radiologists interpreted each radiograph using a structured report format built into AI-PACS and referring to results from AI-CAD software for subjects in the AI group.
The primary endpoint was the detection rate of clinically significant pulmonary nodules confirmed by a CT scan performed within three months, where significant nodules were defined as solid nodules greater than 8 mm or subsolid nodules with a solid part more than 6 mm (Lung-RADS category 4). Pathological results were reviewed four months after the end of the trial. Detection rate was defined as the number of true positive X-rays divided by the total number of X-rays. Secondary outcomes included positivity rate, sensitivity, false referral rate, and cancer detection rate of the lung.
The researchers compared primary and secondary outcomes between the two groups. Subgroup analyzes were also performed using demographic data, medical history, and radiologist readers.
The trial demonstrated an increase in the detection rate of Lung-RADS Grade 4 nodules in the AI group (AI group vs. non-AI group; 0.59% [31/5,238] against 0.25% [13/5,238]; p = 0.008).
A chest CT scan was obtained between the two groups within three months. Among those who underwent chest CT, the prevalence of large nodules was similar, and chest X-ray performance showed increased sensitivity in the AI group. Chest X-ray of these patients in the AI group also showed higher positive and negative predictive values, with very few false positives.
There was no difference between x-ray positivity rates and misdirection rates. Twenty-four people were diagnosed with primary lung cancer (16 in the AI group and 8 in the non-AI group) and there was no significant difference in lung cancer detection rates.
Copyright © 2022 AuntMinnie.com
#improves #detection #lung #nodules #medical #screening