AI Analysis of Body Composition Predicts Survival After Aortic Valve Implantation

By Dr Silpaja Chandrasekar, PhDApr 26 2024Reviewed by Susha Cheriyedath, M.Sc.

In a paper published in the journal Scientific Reports, researchers explored the potential of artificial intelligence (AI) in analyzing body composition from computed tomography (CT) scans to predict all-cause mortality in patients undergoing transcatheter aortic valve implantation (TAVI) for severe aortic stenosis. They utilized a neural network known as automated muscle and tissue analysis in CT angiography (AutoMATiCA) to extract parameters such as skeletal muscle index (SMI) and adipose tissue density from CT scans of 866 patients.

Higher SMI correlated with decreased mortality risk, while elevated visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) density indicated heightened mortality risk following TAVI. This automated CT-based body composition assessment holds promise for bolstering preoperative risk evaluation and clinical decision-making in TAVI patients.

Background

Previous works have highlighted the significance of TAVI as a minimally invasive solution for replacing dysfunctional aortic valves, which is particularly beneficial for frail elderly patients. Recent studies have delved into the predictive value of body composition parameters extracted from pre-procedural CT scans in this patient cohort. These parameters encompass skeletal muscle mass and adipose tissue metrics, including SMI, VAT, and SAT indexes and their corresponding X-radiation (X-ray) attenuations expressed in Hounsfield units (HU). Sarcopenia is excessive body fat, both separately and together as "sarcopenic obesity," which has a big impact on what happens to patients after TAVI.

TAVI Registry Analysis

This retrospective study utilized data from the TAVI Registry, comprising patients who underwent TAVI. The Registry collects demographic characteristics, imaging data, and clinical information from electronic health records. All personal data handling adheres to the general data protection regulation (GDPR), with participants providing written informed consent. Ethical approval was obtained in line with the declaration of Helsinki. The study, registered under trial number NCT05672160, involved subjects with available demographic characteristics and pre-procedural torso CT scans.

Patient characteristics, including sex, body weight, height, date of birth, and date of TAVI procedure, were extracted from the Registry. Additionally, personal history data, such as primary hypertension, diabetes mellitus, myocardial infarction, and respiratory disease, were collected. Body mass index (BMI) was standardized using a standard formula. The primary outcome measure assessed in the study was overall survival (OS), defined as the duration from the TAVI procedure to the occurrence of death from any cause, with data censored as of December 31, 2022.

TAVI procedures followed current guidelines, mainly using the transfemoral approach under fluoroscopy control. The analysts provided detailed information on TAVI indications, techniques, methods, and complications. Imaging data, particularly CT scans performed pre-TAVI, were retrieved from the picture archiving and communication system (PACS). Siemens Somatom Definition and Siemens Somatom Drive CT scanners were employed for image acquisition. These images were then segmented at the third lumbar vertebra (L3) level using AutoMATiCA without manual editing.

AutoMATiCA segmented CT scans into skeletal muscle and adipose tissue sections, including intramuscular, visceral, and subcutaneous adipose tissue. These sections' areas and mean radiation attenuation values were calculated. SMI, intramuscular adipose tissue index, visceral adipose tissue index, and subcutaneous adipose tissue index were determined, along with their corresponding densities.

Statistical analysis involved Spearman correlation, univariate and multivariable Cox proportional hazards models, adjusting for age, sex, comorbidities, and procedural complications. Researchers determined optimal cutoff values for risk classification and generated Kaplan-Meier survival curves to visualize and compare patients categorized as low risk versus high risk.

TAVI Cohort Analysis

The study cohort comprised 866 TAVI Registry patients enrolled from 2010 until December 31, 2022, with available pre-procedural CT scans at the L3 level. Among these patients, 49.5% were men and 50.5% were women, with a high prevalence of comorbidities such as hypertension (89.8%), type 2 diabetes mellitus (43.3%), coronary heart disease (20.6%), and pulmonary disease (30.8%). The median follow-up time was 5.89 years, with no dropouts recorded. CT scans were segmented at the L3 level to obtain CTL3 parameters for each patient.

Informative CTL3 parameters associated with overall survival were identified through Spearman correlation and univariate Cox proportional hazard models. Strong correlations were observed between VAT density and VAT index, as well as VAT density and SAT density. The multivariable analysis confirmed SMI, VAT density, and SAT density as significant predictors of survival, with hazard ratios (HRs) of 0.986 and 1.014, respectively.

Optimal threshold values for stratifying patients into low- and high-risk subgroups based on SMI, VAT, and SAT densities were determined using the maximization of the log-rank technique. Subsequent comparison of overall survival between these subgroups, adjusted for age, comorbidities, and procedural complications, revealed increased mortality risk in the high-risk groups, particularly among males.

VAT density emerged as a potential predictor suitable for estimating overall survival risk in both genders. Therefore, the team performed a Kaplan-Meier survival analysis to demonstrate mortality risk in individual years after TAVI for low- and high-risk groups. These findings provide valuable insights for clinical decision-making and patient discussion, offering survival estimates for the first five years post-TAVI based on gender and risk group. Standard Kaplan-Meier graphs illustrate the impact of SMI and SAT density parameters on survival for both men and women.

Conclusion

In conclusion, the analysis of the TAVI Registry data highlighted the importance of CT-based body composition parameters in predicting overall survival post-TAVI. Strong associations between SMI, VAT, and SAT density with mortality risk were identified in a cohort of 866 patients. Stratification based on these parameters revealed higher mortality risks in patients with lower SMI and higher VAT and SAT densities, particularly among males.

Notably, VAT density emerged as a robust predictor of survival in both genders. These findings offer valuable insights for clinical decision-making and patient counseling, providing tailored survival estimates for the initial five years following TAVI based on gender and risk group.