• Center on Health Equity and Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

AI Beats Radiologists in Detecting Lung Cancer, Study Finds

Article

Researchers recently conducted a study in which they trained an artificial intelligence (AI) deep learning tool to detect lung cancer tumors in computed tomography scans. The algorithm's evaluation was then compared with that of 6 radiologists, and the results showed that the AI was more accurate when prior CT imaging was not available.

Lung cancer, responsible for the death of nearly 160,000 Americans per year, is the most common cause of cancer death in the United States. Early detection is key for both stopping the spread of tumors and improving patient outcomes, and one way to accomplish this is through screening.

Screening for the disease by using low-dose computed tomography (CT), an alternative to chest x-rays, has been shown to reduce mortality by 20% to 43%. However, there is room for improvement as screening results aren’t always accurate—there tends to be high rates of false positives and false negatives because certain spots on the lungs can be mistaken for malignancies, while others are thought to be benign.

In order to increase the accuracy of such tests, researchers conducted a study in which they trained an artificial intelligence (AI) deep learning algorithm to detect lung tumors in CT scans.1 The algorithm learns by example, and the researchers trained the system by using more than 42,000 patient CT scans (both current and prior scans) to predict the risk of lung cancer. These CT scans were from patients whose diagnoses were known; some already had lung cancer, some did not, and some had lumps that later turned cancerous.

“When prior [CT] imaging was not available, our model outperformed all 6 radiologists with absolute reductions of 11% in false positives and 5% in false negatives. When prior [CT] imaging was available, the model performance was on-par with the same radiologists,” wrote the authors.

The algorithm was tested against 6716 cases with known diagnoses and was found to be 94% accurate. The authors stressed that while the tool is still in the early stages of testing, the intention is for it to help radiologists diagnose patients, not replace them. In terms of the future of the tool, researchers will look to confirm the accuracy of AI diagnoses in larger cohorts.

Lung cancer screening has garnered much more attention recently as various studies have been published in the last year on the outcomes of screening guidelines and programs.

While most guidelines call for the screening of smokers, a study published in April 2019 found that lung cancer in people who have never smoked is more common than was once thought and in fact, is on the rise.2 Major contributing factors to the development in lung cancer in never-smokers include second-hand smoke, occupational carcinogen exposure, and outdoor pollution. Interestingly, on a global level, the use of solid fuels for indoor cooking and second-hand smoke exposure disproportionately affect women.

“Drawing attention to the contribution of underlying risk factors to lung cancer in never-smokers presents opportunities to reinforce efforts to tackle other major public health challenges. For example, the impact of passive smoking and air pollution on lung cancers adds weight to the government’s ambitions to improve air quality and the public, clinicians and policy makers must all be aware of this relationship,” Mick Peake, MD, clinical director of the Centre for Cancer Outcomes University College London Hospitals Cancer Collaborative, said in a statement.

Reference

1. Ardila D, Kiraly A, Bharadwaj S, et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography [published online May 20, 2019]. Nat Med. nature.com/articles/s41591-019-0447-x.

2. Bhopal A, Peake MD, Gilligan D, Cosford P. Lung cancer in never-smokers: a hidden disease [published online April 25, 2019]. J R Soc Med. doi: 10.1177/0141076819843654.

Related Videos
dr parth rali
Edward W. Boyer, MD, PhD, The Ohio State University
Edward W. Boyer, MD, PhD, The Ohio State University
Hilary Tindle, MD, MPH, associate professor of medicine, Vanderbilt University
Dr Hilary Tindle, MD, MPH, associate professor of medicine, Vanderbilt University
Hilary Tindle, MD, MPH, Vanderbilt University
Hilary Tindle, MD, MPH, Vanderbilt University
Julia Balmaceda
Julia Balmaceda
Related Content
© 2024 MJH Life Sciences
AJMC®
All rights reserved.