Randomized Prospective Trial Suggests AI Improves Lung Cancer Screening

Artificial intelligence (AI) improved detection of actionable lung nodules without leading to higher rates of false positives, the study found.

Artificial intelligence (AI) appears to improve the accuracy of lung cancer screening, according to a new study. The report found pulmonary-nodule detection rates were higher when experienced radiologists were assisted by computer-aided detection (CAD) software.

Investigators say the results are important data points that could lead to increased integration of AI into radiography interpretation and potentially result in many more patients benefiting from early detection. The results were published in Radiology.

The study authors explained that AI has already been adopted as a tool to aid detection and diagnosis of several cancer types, and AI-based CAD systems have been shown to improve the accuracy of radiologists’ readings. Yet, they said the existing evidence supporting the use of these technologies in lung cancer has significant limitations. Most previous research was performed retrospectively and outside of the real-world clinic setting. Thus, they said, the previous research was subject to the potential for both selection bias and performance bias. In the new study, the investigators sought to avoid those limitations.

“The purpose of our randomized controlled trial was to investigate whether commercial AI-based CAD software could improve the detection rate of actionable lung nodules on chest radiographs in health checkup participants,” they said.

The open-label trial was performed at a single health care center between July 2020 and December 2021, and 10,476 participants were randomized 1:1 to have their radiographs read with the help of AI or without. The 3 radiologists who interpreted the results had between 13 and 36 years of experience. The median age of the participants was 59 years, and 5121 were men.

The results showed a 0.59% rate of detecting actionable nodules in the AI group compared with a 0.25% rate in the non-AI group (odds ratio [OR], 2.4; 95% CI, 1.3%-4.7%; P = .008). Similarly, the rate of malignant lung nodules was 0.15% in the AI group, while no cases were found in the non-AI group. This higher accuracy was achieved without increasing the risk of false positives, the authors found. In the AI group, the false-positive referral rate was 45.9% vs 56.0% in the non-AI group. The positive-report rates were 2.3% and 1.9%, respectively.

The investigators noted that nearly half of patients underwent CT regardless of their radiography results, so the results of the study are indicative solely of the impact of AI on the diagnostic performance of chest radiography and not on overall rates of diagnosis.

“Nevertheless, the improved detection rate of actionable lung nodules with a similar false-referral rate suggests that using AI-based CAD may improve lung cancer diagnosis without imposing an additional radiation hazard,” they wrote.

An accompanying editorial noted there are important limitations to the study. This was a single-center study that involved a small number of radiologists, so it is not clear from this evidence whether AI would work equally well on a more diverse population or with a larger set of radiologists interpreting the radiography. Also, it is not clear from the study whether using AI would affect the time it takes to read the images.

However, the author said that despite those limitations, the trial marks “an important step forward in demonstrating the utility of AI-based CAD.”

“While some questions remain to be answered, this article provides much-needed randomized controlled trial data regarding the potential utility of AI-based CAD to help radiologists detect pulmonary nodules in clinical practice,” he concluded.


Nam JG, Hwang EJ, Kim J, et al. AI improves nodule detection on chest radiographs in a health screening population: a randomized controlled trial. Radiology. Published online February 7, 2023. doi:10.1148/radiol.221894

Related Videos
Samyukta Mullangi, MD, MBA.
Screenshot of Erik Jaklitsch smiling
Sigrun Hallmeyer, MD, Advocate Health
Kevin Davies, PhD, Editing Humanity: The CRISPR Revolution and the New Era of Genome Editing
Olalekan Ajayi, PharmD, MBA, 2023-2024 ACCC president-elect
Sashi Naidu, MD, Carolina Blood and Cancer Care Associates
Michael Cohen, MD, FAAO
Timothy Murphy, MD, Rocky Mountain Cancer Centers
Ken Cohen interview
Timothy Murphy, MD, Rocky Mountain Cancer Centers
Related Content
© 2023 MJH Life Sciences
All rights reserved.