Algorithm Predicts Risk of Death from Heart Disease Using Information from Lung Cancer Screening CT Exam

A deep learning (DL) algorithm that uses information from low-dose CT exams performed for lung cancer screening accurately predicts the risk of death from cardiovascular disease. For the study published in Radiology: Cardiothoracic Imaging, researchers tested a faster, automated method using DL that can predict five-year cardiovascular disease mortality with only minimal extra workload.

Data from 4,451 participants, median age 61 years, who underwent low-dose CT over a two-year period in the National Lung Screening Trial, the researchers trained the method to quantify six types of vascular calcification. They then tested the method on data from 1,113 participants.

The prediction model using calcium scores outperformed the baseline model that used only self-reported participant characteristics, such as age, history of smoking, and history of illness.

The method works in two stages, according to study lead author Bob D. de Vos, PhD, from Amsterdam University Medical Center in Amsterdam and the Image Sciences Institute, University Medical Center Utrecht, in Utrecht, the Netherlands. The first stage determines the amount and location of arterial calcification in the coronary arteries and the aorta using DL. The second stage uses a more conventional statistical approach for mortality prediction. The second stage also indicates which features are most predictive for five-year mortality.

“The analysis shows we found predictors that are typically not described in a literature, possibly because we performed analysis in lung cancer screening participants who are already at high risk of cardiovascular disease from a history of heavy smoking and the presence of extensive arterial calcification,” Dr. de Vos said.

The method could easily be integrated into lung cancer screening, Dr. de Vos said. It does not require any special equipment and would not add time to the exam.

“The method uses only image information, it is fully automatic, and it is fast,” Dr. de Vos said. “The method obtains calcium scores in a complete chest CT in less than half a second. This means that the method should be easy to implement in routine patient work ups and screening.”

Most importantly, the method could help identify people in a population of heavy smokers who might be at increased risk of death from cardiovascular disease-related causes.

“Lung screening studies show that heavy smokers die from cardiovascular disease as much as from lung cancer,” Dr. de Vos said. “But we also see that some people with very high calcium scores survive, while others with low scores do suffer from major cardiac events. The work offers a direction for future research to precisely pinpoint which calcifications are dangerous.”

The researchers have developed a number of methods for automatic calcium scoring that can be applied to a wide variety of data. They are now working toward a calcium scoring method that accurately detects arterial calcification in low-quality data, like data affected by cardiac motion, low image resolution or high noise levels.

“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. “This way, we hope to increase the reproducibility of calcium scoring and enable more accurate prediction.”