CV Risk Assessment Using Deep Learning-Based Tissue Characterization

The elevated risk for cardiovascular disease (CVD) in patients with rheumatoid arthritis (RA) has been known for decades. Compared with individuals without RA, those with RA are twice as likely to experience a silent myocardial infarction, carry a higher burden of coronary plaques even in the absence of a clinical history of coronary artery disease, and after a new CVD event, to have a higher risk for CVD mortality.^1,2 The prevailing theory is that genetic, environmental factors, and medications used to treat RA have a direct influence on RA pathophysiologic characteristics and a modifying influence on the classical CVD risk factors.³ Indeed RA and atherosclerosis are closely linked in regard to the genetic and environmental factors that contribute to activation of endothelial cells to increase permeability, and production of proinflammatory cytokines that collectively contribute to physiologic changes, including arterial stiffness, changes in lipid salvage, and destabilization of plaque, that contribute to rupture and infarction.^3,4 The tight control of RA that advocates a treat-to-target strategy has contributed significantly to slowing, and in some patients achieving remission of articular symptoms of RA.⁵ However, appropriate management of the elevated CVD risk is less clear, despite a growing understanding of these mechanisms and their complex interplay with conventional cardiovascular risk factors. Managing CVD risk in patients with RA starts with a comprehensive individual risk assessment and stratification of risk level, and herein lies the challenge.

The CVD risk assessment algorithms used in the general population, such as the Framingham Risk Score, categorize individuals based on traditional risk factors.⁶ The recent review by Khanna and colleagues summarizes the paradoxical relationship between RA and some of the known risk factors for CVD such as body mass index, physical inactivity, and lipid levels; the less than clear influence of hypertension and smoking; and the known elevated risk associated with insulin resistance, diabetes, and RA-specific risk factors such as erythrocyte sedimentation rate, C-reactive protein, rheumatoid factor, and anti-citrullinated protein antibodies.⁴ Consequently, traditional risk assessment algorithms such as the Framingham Risk Score or the systematic coronary risk evaluation that do not take into account the influence of RA on the traditional risk factors fall short of accurately assessing CVD risk, and tend to result in underestimation or overestimation of these risks.⁷

Although various modifications of the conventional risk algorithms have been introduced, such as the modified systematic coronary risk evaluation,⁸ QRISK®2 calculator,⁹ and the American College of Cardiology/American Heart Association Pooled Cohort Equations Risk Calculator,¹⁰ accurate CVD risk stratification in high-risk patients continues to be a problem. Underestimating risk represents missed opportunity for intensive therapy and prevention of CVD events, while overestimating the risk in low-risk populations results in unnecessary therapy.^4,11 A significant limitation and challenge associated with the current risk prediction algorithm is their inability to capture cases of premature plaque development that are known to occur in patients with RA.¹² Although imaging studies have been used to augment CVD risk assessment algorithms, noninvasive imaging techniques cannot provide detailed visualization or characterization of the plaque microstructure, with the added risk for exposure to ionizing radiation,⁴ and invasive imaging techniques may not be suitable for all patients.  

The limitations of conventional CVD risk prediction algorithms and imaging techniques expose unmet needs that are being addressed by artificial neural networks, an intelligent learning-based algorithm. The concept of artificial intelligence and machine learning is not new, however, its application in several areas of medicine is growing rapidly such as in kidney dialysis, physiotherapy, and assessment of liver function. ^13-16 Machine-learning techniques have been applied to several domains of cardiology for tissue characterization in carotid arteries using carotid ultrasound images and for improving medical imaging as a result of the ability of deep learning-based tissue characterization to extract high-level image feature compared to conventional imaging.^4,17 Machine learning was, in fact, found to significantly improve the accuracy of CVD risk prediction, increasing the number of patients identified who could benefit from preventive treatment while avoiding unnecessary treatment to others,¹⁸ outperforming the American College of Cardiology/American Heart Association Pooled Cohort Equations Risk Calculator by recommending less drug therapy, yet missing fewer cardiovascular events.¹¹

The pathologic process in the development of atherosclerotic plaque is similar in the carotid and coronary blood vessels because of their same genetic makeup. Since machine learning-based and deep learning-based algorithm have been widely used in carotid artery characterization, the assumption is that these tools can be applied to patients with RA to better characterize the atherosclerotic plaque tissues and more accurately predict CVD risk. “The inflammation is characterized by the change in components of the plaque build over time, and this is characterized by the reflectance of the sound using low cost ultrasound scanners, unlike MRI or CT”, said Jasjit S Suri, PhD, MBA, Chairman of AtheroPoint™, in Roseville, California, adding “These characteristics can further help in estimating the accurate plaque burden in arterial walls.” Dr. Suri is one of the pioneers of medical machine learning systems, especially for stroke and cardiovascular risk assessment, and also designed features which combines conventional risk factors with image-based factors, so-called “integrated approaches” for machine learning. Currently, little attention has been paid to characterizing atherosclerotic plaque tissue in RA or to predicting CVD risk using automated intelligence-based techniques. However, Dr. Suri believes that automated intelligence-based techniques have clinical applications as part of early an RA screening tool to identify patients at risk for CVD, although more research is needed.

References

1. Liao KP. Cardiovascular disease in patients with rheumatoid arthritis. Trends Cardiovasc Med. 2017;27(2):136-140.

2. Jagpal A, Navarro-Millán I. Cardiovascular co-morbidity in patients with rheumatoid arthritis: a narrative review of risk factors, cardiovascular risk assessment and treatment. BMC Rheumatol. 2018;2:10.

3. Chodara AM, Wattiaux A, Bartels CM. Managing cardiovascular disease risk in rheumatoid arthritis: clinical updates and three strategic approaches. Curr Rheumatol Rep. 2017;19(4):16.

4. Khanna NN, Jamthikar AD, Gupta D, et al. Rheumatoid arthritis: atherosclerosis imaging and cardiovascular risk assessment using machine and deep learning-based tissue characterization. Curr Atheroscler Rep. 2019;21(2):7.

5. Ajeganova S, Huizinga T. Sustained remission in rheumatoid arthritis: latest evidence and clinical considerations. Ther Adv Musculoskelet Dis. 2017;9(10):249-262.

6. Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97(18):1837-1847.

7. Zegkos T, Kitas G, Dimitroulas T. Cardiovascular risk in rheumatoid arthritis: assessment, management and next steps. Ther Adv Musculoskelet Dis. 2016;8(3):86-101.

8. Peters MJ, Symmons D, McCarey D, et al. EULAR evidence based recommendations for cardiovascular risk management in patients with rheumatoid arthritis and other forms of inflammatory arthritis. Ann Rheum Dis. 2010;69(2):325-331.

9. Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2. BMJ. 2008;336(7659):1475-1482.

10. Stone NJ, Robinson JG, Lichtenstein AH, et al. 2013 ACC/AHA guideline on the treatment of blood cholesterol to reduce atherosclerotic cardiovascular risk in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. 2014. 2013;63(25 Part B):2889-2934.

11. Kakadiaris IA, Vrigkas M, Yen AA, Kuznetsova T, Budoff M, Naghavi M. Machine learning outperforms ACC/AHA CVD risk calculator in MESA. J Am Heart Assoc. 2018;7(22):e009476.

12. Kahlenberg JM, Kaplan MJ. Mechanisms of premature atherosclerosis in rheumatoid arthritis and lupus. Annu Rev Med. 2013;64:249-263.

13. Buch VH, Ahmed I, Maruthappu M. Artificial intelligence in medicine: current trends and future possibilities. Br J Gen Pract. 2018;68(668):143-144.

14. Hueso M, Vellido A, Montero N, et al. Artificial intelligence for the artificial kidney: pointers to the future of a personalized hemodialysis therapy. Kidney Dis (Basel). 2018;4(1):1-9.

15. Tack C. Artificial intelligence and machine learning | applications in musculoskeletal physiotherapy. Musculoskelet Sci Pract. 2019;39:164-169.

16. Zhou LQ, Wang JY, Yu SY, et al. Artificial intelligence in medical imaging of the liver. World J Gastroenterol. 2019;25(6):672-682.

17. Al’Aref SJ, Anchouche K, Singh G, et al. Clinical applications of machine learning in cardiovascular disease and its relevance to cardiac imaging [published online July 27, 2018]. Eur Heart J. doi: 10.1093/eurheartj/ehy404

18. Weng SF, Reps J, Kai J, Garibaldi JM, Qureshi N. Can machine-learning improve cardiovascular risk prediction using routine clinical data? PLoS One. 2017;12(4):e0174944.

This article originally appeared on Rheumatology Advisor