Unsupervised Time-Series Clustering of Left Atrial Strain for Cardiovascular Risk Assessment.

BACKGROUND: Early identification of individuals at high risk for developing cardiovascular (CV) events is of paramount importance for efficient risk management. Here, the authors investigated whether using unsupervised machine learning methods on time-series data of left atrial (LA) strain could distinguish clinically meaningful phenogroups associated with the risk for developing adverse events. METHODS: In 929 community-dwelling individuals (mean age, 51.6 years; 52.9% women), clinical and echocardiographic data were acquired, including LA strain traces, at baseline, and cardiac events were collected on average 6.3 years later. Two unsupervised learning techniques were used: (1) an ensemble of a deep convolutional neural network autoencoder with k-medoids and (2) a self-organizing map to cluster spatiotemporal patterns within LA strain curves. Clinical characteristics and cardiac outcome were used to evaluate the validity of the k clusters using the original cohort, while an external population cohort (n = 378) was used to validate the trained models. RESULTS: In both approaches, the optimal number of clusters was five. The first three clusters had differences in sex distribution and heart rate but had a similar low CV risk profile. On the other hand, cluster 5 had the worst CV profile and a higher prevalence of left ventricular remodeling and diastolic dysfunction compared with the other clusters. The respective indexes of cluster 4 were between those of clusters 1 to 3 and 5. After adjustment for traditional risk factors, cluster 5 had the highest risk for cardiac events compared with clusters 1, 2, and 3 (hazard ratio, 1.36; 95% CI, 1.09-1.70; P = .0063). Similar LA strain patterns were obtained when the models were applied to the external validation cohort, and clinical characteristics revealed similar CV risk profiles across all clusters. CONCLUSION: Unsupervised machine learning algorithms used in time-series LA strain curves identified clinically meaningful clusters of LA deformation and provide incremental prognostic information over traditional risk factors.

Challenging obesity and sex based differences in resting energy expenditure using allometric modeling, a sub-study of the DIETFITS clinical trial.

BACKGROUND & AIMS: Resting energy expenditure (REE) is a major component of energy balance. While REE is usually indexed to total body weight (BW), this may introduce biases when assessing REE in obesity or during weight loss intervention. The main objective of the study was to quantify the bias introduced by ratiometric scaling of REE using BW both at baseline and following weight loss intervention. DESIGN: Participants in the DIETFITS Study (Diet Intervention Examining The Factors Interacting with Treatment Success) who completed indirect calorimetry and dual-energy X-ray absorptiometry (DXA) were included in the study. Data were available in 438 participants at baseline, 340 at 6 months and 323 at 12 months. We used multiplicative allometric modeling based on lean body mass (LBM) and fat mass (FM) to derive body size independent scaling of REE. Longitudinal changes in indexed REE were then assessed following weight loss intervention. RESULTS: A multiplicative model including LBM, FM, age, Black race and the double product (DP) of systolic blood pressure and heart rate explained 79% of variance in REE. REE indexed to [LBM0.66 × FM0.066] was body size and sex independent (p = 0.91 and p = 0.73, respectively) in contrast to BW based indexing which showed a significant inverse relationship to BW (r = -0.47 for female and r = -0.44 for male, both p < 0.001). When indexed to BW, significant baseline differences in REE were observed between male and female (p < 0.001) and between individuals who are overweight and obese (p < 0.001) while no significant differences were observed when indexed to REE/[LBM0.66 × FM0.066], p > 0.05). Percentage predicted REE adjusted for LBM, FM and DP remained stable following weight loss intervention (p = 0.614). CONCLUSION: Allometric scaling of REE based on LBM and FM removes body composition-associated biases and should be considered in obesity and weight-based intervention studies.

Incremental Value of Aortomitral Continuity Calcification for Risk Assessment after Transcatheter Aortic Valve Replacement.

PURPOSE: To investigate the association of aortomitral continuity calcification (AMCC) with all-cause mortality, postprocedural paravalvular leak (PVL), and prolonged hospital stay in patients undergoing transcatheter aortic valve replacement (TAVR). MATERIALS AND METHODS: The authors retrospectively evaluated 329 patients who underwent TAVR between March 2013 and March 2016. AMCC, aortic valve calcification (AVC), and coronary artery calcification (CAC) were quantified by using preprocedural CT. Pre-procedural Society of Thoracic Surgeons (STS) score was recorded. Associations between baseline AMCC, AVC, and CAC and 1-year mortality, PVL, and hospital stay longer than 7 days were analyzed. RESULTS: The median follow-up was 415 days (interquartiles, 344-727 days). After 1 year, 46 of the 329 patients (14%) died and 52 (16%) were hospitalized for more than 7 days. Of the 326 patients who underwent postprocedural echocardiography, 147 (45%) had postprocedural PVL. The CAC score (hazard ratio: 1.11 per 500 points) and AMCC mass (hazard ratio: 1.13 per 500 mg) were associated with 1-year mortality. AVC mass (odds ratio: 1.93 per 100 mg) was associated with postprocedural PVL. Only the STS score was associated with prolonged hospital stay (odds ratio: 1.19 per point). CONCLUSION: AMCC is associated with mortality within 1 year after TAVR and substantially improves individual risk classification when added to a model consisting of STS score and AVC mass only.Supplemental material is available for this article.© RSNA, 2019See also the commentary by Brown and Leipsic in this issue.