Study: AI Assessment of Chest CT May Predict Multiple Mortality Risks

Emerging research suggests that artificial intelligence (AI) assessment of low-dose chest computed tomography (CT) scans can help predict lung cancer, cardiovascular disease (CVD) and all-cause mortality risks in patients being screened for lung cancer.

For the study, recently published in Radiology, researchers assessed the use of an AI algorithm, which provides automated body composition measurements from lung screening CT exams, for 20,768 participants (median age of 61) from the National Lung Screening Trial (NLST). The cohort was comprised of 12,317 men and 8,451 women, according to the study. The researchers noted that 4,180 study participants died during the 12.3-year follow-up period with 913 deaths attributed to lung cancer and 972 deaths attributed to CVD.

The study authors found that AI-enabled body composition measurements led to improved prediction of lung cancer mortality (_x2=23.09 for men and _x2=15.04 for women) and death due to CVD (_x2=69.94 for men and _x2=16.60 for women). The AI algorithm also demonstrated the capability to predict all-cause mortality (_x2=248.13 for men and _x2=94.54 for women), according to the study authors.

The researchers noted that higher subcutaneous adipose tissue (SAT) attenuation was associated with a 17 percent higher risk of CVD-related mortality and an 18 percent higher risk of all-cause mortality in men as well as a 17 percent higher risk of all-cause mortality in women. Higher SAT attenuation was also associated with a 27 percent higher risk of lung cancer mortality in women, according to the study.

“Opportunistic assessment of imaging data beyond the initial study indication has great potential to extend the value of established population-based CT screening, especially when combined with fully automated AI solutions,” noted study co-author Kim L. Sandler, M.D., an associate professor in the Department of Radiology at the Vanderbilt University Medical Center in Nashville, Tenn., and colleagues.

For a 57-year-old male study participant, one can see automated AI assessment of body composition via low-dose chest computed tomography (CT) that includes the field of view of each CT section (B) and segmentation of subcutaneous adipose tissue (SAT) and skeletal muscle (SM).

“Our analyses provide the evidence that fully automated body composition assessment in lung cancer screening can potentially improve the overall health of lung cancer screening population by identifying high-risk individuals for targeted interventions, such as medical optimization, physical conditioning, or lifestyle modification.”

(Editor’s note: For related content, see “AI-Based Denoising Technique Achieves 76 Percent Reduction in Radiation Dosing for CT Lung Cancer Screening,” “Nine Takeaways from Recent Meta-Analysis on Lung Cancer Screening with Low-Dose CT” and “Deep Learning Model May Predict Lung Cancer Risk from a Single CT Scan.”)

The researchers noted that higher skeletal muscle (SM) attenuation (per 7.6 Hounsfield unit (HU) increase) was associated with a 22 percent decrease in lung cancer mortality risk in men and a 24 percent decrease in women. Sandler and colleagues found that lower SM attenuation was associated with a 41 percent increased risk of lung cancer mortality in men and greater than twofold risk for CVD-related mortality in men and women.

“We observed that measurements associated with muscle adiposity, especially the attenuation of SM, play an important role in predicting mortality end points,” noted Sandler and colleagues. “ … Indeed, fat deposition within or around non-adipose tissues or organs has been associated with increased inflammation, metabolic disorders (e.g. insulin resistance, type 2 diabetes, CVD) and physical impairment.”

In regard to study limitations, the authors conceded a lack of analysis on additional incidental findings such as liver steatosis, bone mineral density and interstitial lung abnormalities that can affect mortality. They also cautioned that additional research is necessary to validate their findings given the exploratory nature of their study.