Non-Small Cell Lung Cancer Coverage from Every Angle
Advertisement
Advertisement

Can a Deep Learning Algorithm Predict the Risk of Pulmonary Nodules Becoming Cancerous?

By: Gavin Calabretta, BS
Posted: Wednesday, September 1, 2021

In a recent retrospective study published in Radiology, a novel deep learning algorithm showed the potential to accurately estimate malignancy risk in pulmonary nodules detected at low-dose screening CT. According to the study authors, the algorithm was developed to reduce the rate of false-positives and avoid the overtreatment of benign growths.

“As it does not require manual interpretation of nodule imaging characteristics, the proposed algorithm may reduce the substantial interobserver variability in CT interpretation,” commented senior author Colin Jacobs, PhD, of Radboud University Medical Center, Nijmegen, the Netherlands, in a press release from the Radiological Society of North America. “This may lead to fewer unnecessary diagnostic interventions, lower radiologists’ workload, and reduce costs of lung cancer screening,” he added.

To hone the algorithm’s detection ability, the researchers used its deep learning capability to train it on 16,077 CT images of lung nodules, of which 1,249 were malignant. These images were all collected from the National Lung Screening Trial. For external validation, three additional cohorts of 883, 177, and 175 nodules were also used from the Danish Lung Cancer Screening Trial. The algorithm was evaluated by comparing its performance with the Pan-Canadian Early Detection of Lung Cancer (PanCan) model and a group of 11 clinicians composed of thoracic radiologists, radiology residents, and pulmonologists.

In its testing, the deep learning algorithm outperformed the PanCan model (AUC = 0.93 [95% confidence interval = 0.89–0.96] vs. 0.90 [95% confidence interval = 0.86–0.93]; P = .046) and performed similarly to radiologists in cohorts with random benign nodules as well as size-matched benign nodules. As development of this artificial intelligence algorithm continues, the researchers plan to add parameters such as age and smoking history, as well as a function that can read multiple CT exams simultaneously.

Disclosure: For full disclosures of the study authors, visit pubs.rsna.org.



By continuing to browse this site you permit us and our partners to place identification cookies on your browser and agree to our use of cookies to identify you for marketing. Read our Privacy Policy to learn more.