Novel Method to Predict Cardiovascular Mortality Risk at Lung Cancer Screening
Posted: Tuesday, June 29, 2021
A deep learning network successfully used lung images from CT scans to accurately predict the 5-year mortality risk from cardiovascular disease, according to research published in Radiology: Cardiothoracic Imaging. The study emerged from a retrospective analysis of 4,451 participants who underwent low-dose CT exams over 2 years through the National Lung Screening Trial.
The two-stage method first used deep learning to determine the amount and location of arterial calcification in the aorta and coronary arteries, according to lead study author Bob D. de Vos, PhD, of Amsterdam University Medical Center, and colleagues. Then 5-year mortality was predicted using conventional statistics while also indicating which features may best predict mortality.
To do so, the researchers trained the prediction model to quantify six types of vascular calcification. The algorithm was then tested on data from 1,113 participants, with a median age of 61 years. The prediction model using calcium scores (C = 0.74; 95% confidence interval) outperformed the baseline model that used only self-reported participant characteristics (C = .69; P = .049), such as age, history of illness, and history of smoking. When all variables were combined, the algorithm produced the best results (C = .76; P < .001), the study investigators noted.
“We developed a method, for example, that can detect coronary calcifications even when the lesions are below the clinically used threshold,” Dr. de Vos said in a Radiological Society of North America press release. “This way, we hope to increase the reproducibility of calcium scoring and enable more accurate prediction.”
Disclosure: For full disclosures of the study authors, visit pubs.rsna.org.
