In a recent study, NCI-funded researchers applied an artificial intelligence (AI) model to lung screening computed tomography (CT) images to assess body composition. Specifically, the team looked at two key indicators of health—skeletal muscle and fat (adipose) tissue.

Using their model, the researchers improved their ability to predict death from lung cancer, cardiovascular disease, and other causes. The model wasn’t as helpful in predicting the incidence of lung cancer.

Earlier studies show that assessing a patient’s body composition is a useful way of predicting health outcomes for cancer and other diseases. Here, the researchers used a secondary analysis of non-contrast, low-dose CT images from NCI’s National Lung Screening Trial to see if body composition measurements could help predict lung cancer incidence and death, as well as cardiovascular death and all-cause mortality.

In designing their model, the researchers were able to compensate for a “field of view” limitation, which has been one of the biggest obstacles when assessing body composition from routine chest CT scans.

The authors acknowledge that although their model looks promising, it’s not quite ready for clinical practice. Future studies will help confirm the findings of this exploratory analysis and establish cutoff values for the model to further define what’s normal and what’s not. Once refined, this technology could help clinicians make more informed decisions on cancer treatment and care.

As noted by the corresponding author, Mr. Kaiwen Xu, Vanderbilt University, “Incorporating fully automated CT body composition analysis into a patient’s routine exam may someday help identify people at particularly high risk for negative outcomes, allowing us to target interventions around physical conditioning, lifestyle modifications, and other medical interventions.”