Use of Opportunistic CT Biomarkers: Population Health Opportunities in Radiology

Given the substantial growth in imaging volumes and advancements in imaging quality, there exists the potential to harness extensive amounts of previously uncollected health data from cross-sectional imaging. In Canada, provincial programs and local integrated health networks exist for health care provision. Population-level information harnessed from these has significant potential. The aggregate benefits of opportunistic imaging will need to be evaluated from ethical, population-level, and patient-level perspectives. The recent article by Shah et al (2023) entitled: “Opportunistic Extraction of Quantitative CT Biomarkers: Turning the Incidental Into Prognostic Information” outlined several ways in which opportunistic extraction of computed tomography (CT) biomarkers could provide clinically meaningful data. ¹

In Canada, lung cancer screening programs are becoming more common: this CT imaging creates the potential to offer insights into other aspects of health, beyond the lungs. For example, coronary artery calcification detected through CT scans can serve as a predictor for future cardiovascular adverse events, which are elevated in this patient population. Also, a larger volume of epicardial fat has been associated with a higher risk of developing atrial fibrillation. ¹ A landmark publication by Pickhardt et al in 2020 demonstrated that automated CT biomarkers outperformed established clinical parameters in risk stratification for serious cerebrovascular adverse events in currently asymptomatic patients. ² Considering the number of patients who will be undergoing CT chest scans for lung cancer screening, there is an opportunity to collect additional data contributing to a more comprehensive understanding of population health in individuals who smoke or have a smoking history. This could include risks of osteoporosis and even risks of metabolic syndrome, since fatty liver quantification can be performed if the upper abdomen is included in the field of view, as it commonly is in CT lung cancer screening.

Liver fat measurements, as part of both chest CT and abdominal CT scans, can identify patients with asymptomatic non-alcoholic fatty liver disease. If identified early, it can potentially be managed and reversed before chronic sequela including NASH (non-alcoholic steatohepatitis) and cirrhosis have time to develop. ¹

Osteoporosis, an underdiagnosed and undertreated condition, is associated with patient morbidity, and mortality and a resultant large economic burden. ³ There have been several articles published in recent years investigating the use of artificial intelligence and machine learning for diagnosing osteoporosis. ⁴ CT scans can also measure degrees of sarcopenia as part of opportunistic screening and can have an important role in prognostication for geriatric populations as well as for patients with various cancers. ¹ In Canada, as our population continues to age, this type of information can help with health care resource planning as well as to spur on research to prevent or mitigate development of sarcopenia.

As Dr Shah and colleagues indicate in their manuscript, CT scans are requested for more benign presentations as compared to previous decades, and as a result, increasingly more healthy patients are exposed to CT radiation. Taking advantage of opportunistic data could potentially make this radiation exposure more “worthwhile.” Furthermore, the use of automated segmentation and artificial intelligence can retrospectively amalgamate data from numerous scans and clinical encounters, possibly enhancing future diagnostic accuracy, identifying patterns, and improving imaging assessment beyond what could be achieved manually. Additionally, with the growing field of personalized medicine, there may soon be a time where certain imaging biomarkers could identify a patient’s specific disease subtype or inform an individualized therapy option. Prior to incorporating these biomarkers into routine clinical practice, it will be important to carefully evaluate the potential for false positives, the risks of overdiagnosis, and the influence of lead-time bias.

From a health equity and inclusion lens, implementation of opportunistic screening in Canada has the potential to enable inclusion of individuals who might otherwise not participate in standard screening programs. If an individual presents for an acute condition requiring cross-sectional imaging, the additional opportunistic data could provide individual-level risk stratification for various additional disease entities, allowing for either preventative actions, earlier intervention, or for potential enrollment into programs that these individuals may not otherwise be able to access. Having a centralized data repository for opportunistic imaging and creation of associated research institutes would allow more efficient use of population health data both at a meta and individual level. Of course, informed consent alongside privacy and data protection policies will be essential in the development of these types of programs.

Diagnostic imaging accounts for a significant portion of total health care expenditure within Canada, and the Canadian Association of Radiologists (CAR) recently petitioned the government for an additional $1.5 billion to address wait times post-pandemic. ⁵ A 10-year simulated Markov model which showed that AI-assisted CT-based screening could be a highly cost-effective and clinically efficacious strategy. ⁶ Furthermore, preventative health screening using opportunistic CT biomarkers could have significant health care savings, as has been shown in other preventative health care strategies.^6,7

Taking advantage of opportunistic data will become increasingly relevant not only in diagnostic imaging, but across all of medicine. Data sources such as electronic medical records, wearable devices, census data, and disease registries can all be utilized to gain insight into an individual’s health profile. The early adoption and integration of multimodal data across medicine can improve the power of these insights. There remains a risk of inequitable access, particularly if a financial barrier exists. For example, wearable devices such as smartwatches are costly.

This article by Shah et al (2023) brings to light how opportunistic extraction of CT biomarkers can help in earlier disease detection, management planning, prognostication, and future disease risk stratification. ¹ Perhaps, not too far in the future, a patient undergoing a screening CT for lung cancer could also simultaneously result in clinically meaningful measurements for breast density, bone density, and aortic/coronary artery calcification. Compounded by the power of artificial intelligence, the potential breadth of information to be gained appears almost infinite. However, careful consideration of the relevant ethics, risks, and benefits, must take place before incorporating the use of opportunistic biomarkers in routine clinical practice.