Use of pre-TAVR CT left ventricle myocardial radiomic phenotypes to predict long-term outcomes

Abstract Introduction Life-expectancy and quality of life after transcatheter aortic valve replacement (TAVR) procedure vary substantially among individuals. However, prognostic factors contributing to favorable outcomes remain poorly understood. Severe aortic stenosis can cause left ventricular hypertrophy due to increased afterload, which may become maladaptive as the disease progresses and can lead to myocardial fibrosis. TAVR promotes left ventricle myocardium (LVM) remodeling by reducing the afterload. However, prior maladaptive structural changes of the myocardium may not be reversible in some individuals and lead to poor outcomes. Currently, there are no established methods to evaluate the remodeling capacity of the LVM which may correlate with long-term survival. Purpose Radiomic analysis allows morphological and textural analyses of structures on radiological images. Our aim was to explore whether the radiomic phenotype of the LVM on pre-TAVR CT scans is associated with adverse outcomes following TAVR in a matched case-control study. Methods We identified 89 patients who underwent TAVR at our institution between died more than 30 days post-procedure. These cases were matched to control patients without an event based on age, sex, aortic valve gradient, ejection fraction, diabetes status and prior coronary intervention. Pre-TAVR LVM was segmented into the 17 American Heart Association regions and the segments were exported for radiomic analysis. The association between radiomic features and all-cause mortality was evaluated using conditional logistic regression models correcting for LVM volume index. Multiple comparisons were adjusted for using the false discovery rate (FDR) method. Results A total of 178 individuals were evaluated (age of 81.3 ± 3.3 years; 44% female, 71% high-gradient aortic stenosis). The median follow-up period was 2.57 years (interquartile range: 1.22–3.76 years). Across 3,026 LVM segments, 258 radiomic features were assessed, comprising of 42 histogram-based, 158 texture-based, and 58 volume-based features. Utilizing conditional logistic regression models to account for the matched study design and adjust for the LVM volume index, 22.5% of the radiomic features (n=58) were found to have a significant correlation with all-cause mortality (Figure 1). Among these, 7 were histogram-based, 50 texture-based, and 1 volume-based. The odds ratios for protective features ranged from 0.84 to 0.95, whereas those for adverse features spanned from 1.05 to 1.18 per population standard deviation. Conclusions Morphological and textural patterns of the LVM on pre-TAVR CT are associated with all-cause mortality. Precision phenotyping of LVM could aid in identifying individuals at higher risk of adverse outcomes, potentially guiding the use of additional therapies following the TAVR procedure.