Deep learning enabled analysis of medical images identifies cardiac sphericity as an early marker of cardiomyopathy and related outcomes

Summary

Background:

Quantification of chamber size and systolic function is a fundamental component of cardiac imaging. However, the human heart is a complex structure with significant uncharacterized phenotypic variation beyond traditional metrics of size and function. Examining variation in cardiac shape can add to our ability to understand cardiovascular risk and pathophysiology.

Methods:

We measured the left ventricle (LV) sphericity index (short axis length / long axis length) using deep learning enabled image segmentation of cardiac magnetic resonance imaging data from the UK Biobank. Subjects with abnormal LV size or systolic function were excluded. The relationship between LV sphericity and cardiomyopathy was assessed using Cox analyses, genome-wide association studies, and two-sample Mendelian randomization.

Findings:

In a cohort of 38,897 subjects, we show that a one standard deviation increase in sphericity index is associated with a 47% increased incidence of cardiomyopathy (hazard ratio [HR] 1.47, 95% CI 1.10-1.98, P 0.01) and a 20% increased incidence of atrial fibrillation (HR 1.20, 95% CI 1.11-1.28, P <0.001), independent of clinical factors and traditional MRI measurements. We identify four loci associated with sphericity at genome-wide significance, and Mendelian randomization supports non-ischemic cardiomyopathy as causal for LV sphericity.

Conclusions:

Variation in LV sphericity in otherwise normal hearts predicts risk for cardiomyopathy and related outcomes and is caused by non-ischemic cardiomyopathy.

Graphical Abstract

eTOC

Vukadinovic et al. study cardiac sphericity in a population of over 30,000 healthy adults using deep learning-enabled automated image segmentation. They show that sphericity predicts risk for cardiomyopathy in otherwise normal hearts independent of traditional clinical and imaging factors.

Introduction

The heart is a complex three-dimensional structure exposed to continuous dynamic stress. Genetic risk factors, environmental stressors, and age-associated changes impact cardiac morphology and function over the lifespan in a multitude of ways. Dilation of cardiac chambers and/or decline in systolic function are key indicators of disease, and conventional imaging assessments aim to quantify such changes. In turn, there has been significant interest in understanding the genetic basis for variation in such phenotypes^1-3.

The study of cardiac morphology may provide additional clinical and genetic insights. Specifically, large biobanks with cardiac imaging data now offer an opportunity to define and analyze variation in cardiac morphology that is incompletely quantified by traditional measurements but present in normal populations⁴. However, manual phenotyping is time-consuming and not feasible for testing novel phenotypes at larger scale. These challenges can be addressed by recent advances in deep learning and computer vision which allow for high throughput automated measurements of cardiac structures^5-8, and may identify previously uncharacterized clinically-relevant phenotypic variation^4,9,10.

Cardiomyopathies of different etiologies often result in a similar end-stage phenotype of a more round, spherical ventricle. Among individuals with known cardiac disease, increased sphericity of the left ventricle (LV) has been associated with adverse outcomes, including ventricular arrhythmias and death^11-13. The sphericity index (defined as the ratio of LV short axis length divided by LV long axis length) is a simple metric of LV sphericity that has been previously applied to ventriculography¹¹ and to echocardiography ^14-16. In a small study of subjects without cardiovascular disease at baseline, a volumetric measurement of LV sphericity predicted incident heart failure and atrial fibrillation (AF)¹⁷. We hypothesized that within the spectrum of normal LV chamber size and systolic function, there exists variation in LV sphericity, and this variation may be a marker of cardiac risk with genetic underpinnings. To test this hypothesis, we used automated deep-learning segmentation of cardiac magnetic resonance imaging (MRI) data to estimate and analyze the sphericity index in the UK Biobank. We show that among subjects with normal LV chamber size and systolic function, increased sphericity predicts incident cardiovascular diseases, including cardiomyopathy, AF, and heart failure. Genetic analysis suggests a shared architecture between sphericity and non-ischemic cardiomyopathy (NICM), with NICM as a possible causal factor for LV sphericity among individuals with normal LV size and function.

Results

Measurement of the sphericity index using a convolutional neural network

In order to examine the clinical relevance and genetic influences of LV shape in the setting of normal size and function, we defined our main study cohort (n = 38,897) as UK Biobank participants with cardiac MRIs demonstrating normal LV end-diastolic volume (LVEDV), normal LV end-systolic volume (LVESV), and normal LV ejection fraction (LVEF) (Supplementary Figure 1). The baseline characteristics are shown in Table 1. We applied a fully convolutional neural network for semantic segmentation of cardiac chambers to cardiac MRIs from the UK Biobank. During end-diastole in each segmented 4-chamber image, we defined the smallest rectangle that fully encompassed the LV blood pool and measured the LV long axis length and short axis length. We defined the sphericity index as the ratio of the short axis length to the long axis length (Figure 1a). The correspondence between automated and human-read images (R² 0.73-0.79, mean absolute error [MAE] 0.04-0.06) was similar to human-human variation (R² 0.79, MAE 0.09) in the assessment of sphericity index on 100 random images. Figure 1b provides representative examples of the observed sphericity index in our study cohort. Sphericity index was similarly normally distributed in both men and women (Figure 1c) and showed a slight trend of increasing with age (Figure 1d). Within our study population, the sphericity index was not correlated with traditional MRI measurements of the LV (Supplementary Figure 2).

Table 1.

Cohort baseline characteristics

Characteristic	N (%) or mean ± SD
N	38897
Age at MRI	54.9 ± 7.6
Male	17983 (46.2%)
Self-identified White British	32965 (84.7%)
Body mass index (kg/m2)	26.5 ± 4.2
Pulse rate	67.7 ± 10.7
Hypertension	2388 (6.1%)
Diabetes	947 (2.5%)
Prior myocardial infarction	361 (0.01%)
LVEF (%)	57.0 (4.47)
LVEDV (mL)	135 ± 26.4
LVESV (mL)	58.4 ± 13.4
LVEDVi (mL/m2)	73.0 ± 11.6
LVESVi (mL/m2)	31.5 ± 6.2

Figure 1. Deep learning enabled measurement of left ventricular sphericity index.

a, Automated image segmentation of cardiac chambers was performed by applying a fully convolutional neural net to 4-chamber view cardiac MRI images. Sphericity index was defined by the ratio of the short and long axis lengths of a bounding box of the left ventricle. b, Distribution of the sphericity index in the study population and representative examples. c, The sphericity index is similarly normally distributed in male and female subjects. d, The sphericity index showed a small positive correlation with age. Concentric contour lines reflect increasing density of data point.

Phenome-wide association study of left ventricular sphericity

We broadly assessed for clinical associations with LV sphericity using a phenome-wide association study (PheWAS) approach (Figure 2), and we identified AF as the top association by P value (Pearson r 0.05, P 9.5x10⁻¹⁴). In comparison, LV long axis length had no association with AF (r −0.007, P 0.3), while LV short axis length showed a similar association with AF (r 0.05, P 2.6x10⁻¹⁴). Notably however, AF only ranked 8^th by P value for short axis length. Other associations, including pulse rate (r −0.15, P 2.4x10⁻¹²⁹), systolic blood pressure (r 0.09, P 2.3x10⁻⁴³), and obesity (r 0.08, P 3.2x10⁻³⁸) were substantially stronger for PheWAS of LV short axis length, demonstrating how measurements of organ size may associate with traits related to body size (Supplementary Figure 3). Conversely, metrics of shape may better highlight those associations independent of anthropometrics.

Figure 2. Phenome-wide association study of left ventricular sphericity index.

Upward pointing triangles represent positive correlations with sphericity index, and downward pointing triangles represent negative correlations. The horizontal red line reflects the threshold of Bonferroni significance for 966 tests (5x10⁻⁵). Associations meeting Bonferroni significance are labeled.

Left ventricular sphericity predicts incident disease

We next assessed whether the sphericity index predicts incident disease. We hypothesized that LV sphericity is a marker for cardiomyopathy. We therefore examined incident cardiomyopathy as well as clinical manifestations of cardiomyopathy, including AF, heart failure, and cardiac arrest. Examination of these outcomes is also supported by prior studies of sphericity^11-13,17 and by the strong association with AF that we observed in our PheWAS. In a model adjusted for age at MRI and sex (Model 1), increased sphericity index was associated with increased risk for cardiomyopathy (hazard ratio [HR] 1.57, 95% CI 1.27-1.94, P <0.001), AF (HR 1.35, 95% CI 1.28-1.42, P <0.001), and heart failure (HR 1.37, 95% CI 1.24-1.53, P <0.001), with HR being reported per 1 standard deviation increase in sphericity index. There was no significant association with cardiac arrest. Stratifying the cohort by quintiles of sphericity index and comparing the top 20%, middle 60%, and bottom 20% shows that the relationship between sphericity index and risk extends across the distribution. Just as individuals with higher than median sphericity have increased disease incidence, individuals with lower than median sphericity have decreased disease incidence (Figure 3).

Figure 3. Risk of incident disease stratified by quintiles of sphericity index.

Time-to-event analysis was performed from time of MRI to first occurrence of (a) cardiomyopathy, (b) atrial fibrillation, (c) heart failure, and (d) cardiac arrest. Plots show cumulative incidence of disease stratified by sphericity index quintiles, comparing the top 20% (red), middle 60% (green), and bottom 20% (blue). Models were adjusted for age at MRI and sex.

To determine if the risk associated with increased sphericity can be accounted for by clinical factors, we assessed a second model (Model 2) with additional adjustments for BMI, pulse rate, hypertension, diabetes, and prior myocardial infarction. With these adjustments, the associations with cardiomyopathy, AF, and heart failure remained significant, and there was a modest attenuation in the hazard ratio for heart failure (Table 2). We then asked whether the sphericity index remains predictive once accounting for traditional MRI measurements of size and function. In models with additional adjustment for LV ejection fraction (Model 3a), LVEDV indexed on body surface area (LVEDVi) (Model 3b), or LVESV indexed on body surface area (LVESVi) (Model 3c), the associations with cardiomyopathy and AF remained significant, while the association with heart failure was lost. The same pattern was observed in a model adjusting for both LVEF and LVESVi (Table 2). In this fully adjusted model (Model 4), the association with cardiomyopathy remained robust (HR 1.47, 95% CI 1.10-1.98, P 0.01), while the association with AF partially attenuated (HR 1.20, 95% CI 1.11-1.28, P <0.001).

Table 2.

Risk of incident cardiomyopathy and related outcomes associated with increased sphericity index.

	Cardiomyopathy		Atrial fibrillation		Heart failure		Cardiac arrest
	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
Model 1 Age, sex	1.57 (1.27-1.94)	<0.001	1.35 (1.28-1.42)	<0.001	1.37 (1.24-1.53)	<0.001	1.17 (0.88-1.54)	0.3
Model 2 Model 1 + BMI, pulse rate, HTN, DM, prior MI	1.62 (1.30-2.03)	<0.001	1.31 (1.23-1.38)	<0.001	1.23 (1.10-1.38)	<0.001	1.08 (0.81-1.44)	0.6
Model 3a Model 2 + LVEF	1.51 (1.13-2.02)	0.005	1.23 (1.15-1.32)	<0.001	1.11 (0.96-1.28)	0.2	1.01 (0.73-1.41)	0.9
Model 3b Model 2 + LVEDVi	1.46 (1.09-1.97)	0.01	1.19 (1.11-1.28)	<0.001	1.13 (0.97-1.31)	0.1	0.94 (0.67-1.34)	0.7
Model 3c Model 2 + LVESVi	1.47 (1.09-1.97)	0.01	1.19 (1.11-1.28)	<0.001	1.09 (0.94-1.26)	0.3	0.92 (0.65-1.31)	0.7
Model 4 Model 2 + LVEF, LVESVi	1.47 (1.10-1.98)	0.01	1.20 (1.11-1.28)	<0.001	1.09 (0.94-1.26)	0.3	0.92 (0.65-1.31)	0.7

Hazard ratios (HR) reported per 1 standard deviation increase in sphericity index.

BMI = body mass index, HTN = hypertension, DM = diabetes mellitus, MI = myocardial infarction, LVEF = left ventricular ejection fraction, LVEDVi = left ventricular end-diastolic volume indexed on body surface area, LVESVi = left ventricular end-systolic volume indexed on body surface area

In summary, sphericity index is a predictor of incident cardiomyopathy, AF, and heart failure in a minimally adjusted model. Adjusting for clinical factors partially attenuates the association with heart failure. Additionally adjusting for traditional MRI measurements fully attenuates the association with heart failure and partially attenuates the association with AF. In all adjusted models, the association with cardiomyopathy shows little attenuation.

Genome-wide association study of left ventricular sphericity

We performed a genome-wide association study (GWAS) of common genetic variants and LV sphericity index and compared the results to the GWAS of the LV long and short axis lengths (Figure 4). We identified four loci associated with LV sphericity index at genome-wide significance (P ≤ 5x10⁻⁸): PLN, ANGPT1, PDZRN3, and HLA DR/DQ (Table 3). PLN is a gene well known for its causal role in dilated and arrhythmogenic cardiomyopathies^18,19. ANGPT1 has not been associated with cardiomyopathy to date, but mouse experiments have demonstrated a critical role in mammalian heart development^20,21. The PDZRN3 locus has been previously associated with electrocardiographic characteristics²² and LV fractal structure in prior GWAS⁴, and regulation of post-natal expression has been implicated in murine cardiac maturation and geometry²³. The HLA DR/DQ locus has long been implicated in dilated cardiomyopathy^24,25, and recent GWAS that included both normal and abnormal cardiac MRIs have identified this locus for both left and right chamber size metrics^3,26.

Figure 4. Manhattan plots for genome-wide associations studies of left ventricular sphericity index, short axis length, and long axis length.

The red horizontal line represents a genome-wide significant P value of 5x10⁻⁸. Regions that reach genome-wide significant are labeled with candidate loci.

Table 3.

Lead variants and candidate genes for genome-wide significant loci of each left ventricle (LV) trait.

LV trait	RSID	Chr	hg19 position	Ref	Alt	Beta	P	Candidate gene
Sphericity index	rs11756438	6	118993632	C	A	0.004	1.16x10−15	PLN
Sphericity index	rs1461990	8	108087628	C	G	0.003	4.90x10−10	ANGPT1
Sphericity index	none	3	73578036	AACACAC	A	0.003	2.28x10−8	PDZRN3
Sphericity index	rs199682224	6	32486364	G	T	0.006	3.58x10−8	HLA-DRB5
Short axis length	rs57912492	6	118702621	C	T	0.24	2.43x10−10	PLN
Short axis length	rs462797	16	22832777	G	A	0.46	2.93x10−8	HS3ST2
Short axis length	rs111692972	2	179669931	C	T	−0.44	4.48x10−8	TTN
Long axis length	none	3	73578036	AACACAC	A	−0.37	4.31x10−9	PDZRN3
Long axis length	rs72007904	8	108085680	A	AACTATTC	−0.36	7.33x10−9	ANGPT1

Comparing the GWAS of sphericity to LV short and long axis lengths highlights the value of studying shape-related phenotypes. Given the inherent interplay between sphericity and axis lengths, we expected to see overlap in genetic associations. Indeed, the genetic correlations with sphericity index were 0.64 (P 2.7x10⁻²⁵) and −0.63 (P 6.8x10⁻²⁷) for short axis and long axis lengths respectively. In terms of genome-wide significant loci, PLN was identified by both sphericity and short axis length but not long axis length. Notably, the significance of the association was substantially stronger for sphericity. Conversely, ANGPT1 and PDRZRN3 were identified by long axis length but not short axis length. Interestingly, the TTN gene was significant for short axis length (4.5x10⁻⁸) and suggestive for long axis length (P 1.5x10⁻⁷), but the association with sphericity was much weaker (P 2.1x10⁻⁵). This pattern might suggest that TTN impacts overall heart size more so than shape. To further investigate if the genetic architecture of sphericity is distinct from traditional measurements of size and function, we compared our results to a recent large GWAS of cardiac MRI phenotypes, including LVEF, LVEDVi, and LVESVi³. The genetic correlation between sphericity index and LVEF, LVEDVi, and LVESVi was −0.22 (P 0.07), 0.32 (P 5.8x10⁻⁷), and 0.32 (P 4.6x10⁻⁵) respectively. Of our lead variants for sphericity index, only one (rs11756438) was identified by the prior GWAS of these phenotypes (Supplementary Table 1). These results indicate that there are genetic drivers of LV sphericity that are distinct from those of LV size and function.

Relationship between sphericity and cardiomyopathy

We explored the relationship between LV sphericity and non-ischemic cardiomyopathy (NICM) by first measuring the genetic correlation. Given the findings from our Cox analysis, we also assessed AF as a comparison. For each outcome, we used recent large GWAS^27,28, and correlations were assessed by LD score regression, a technique that is robust to potential sample overlap. We found a significant correlation between sphericity index and cardiomyopathy (rg 0.38, P 0.02) but no genetic correlation between sphericity and AF (rg 0.04, P 0.5).

We then assessed for evidence of causal relationships using bidirectional two-sample Mendelian randomization by inverse variance weighted meta-analysis. To do so, we used summary statistics for NICM and AF from the FinnGen project²⁹, a cohort independent of the UK Biobank. When assessing sphericity index as an exposure, we found no evidence for causality with either NICM or AF. We then assessed sphericity index as an outcome. When using AF as an exposure, we found no evidence for causality with sphericity (Supplementary Table 2). However, when using NICM as an exposure, we found a significant causal relationship for sphericity (beta 0.002, P 0.03). In sensitivity analyses, the relationship remained significant by weighted medians (beta 0.001, P 0.049) but not by MR Egger (beta 0.0007, P 0.6). Together, these findings suggest that cardiomyopathy may be causal for increased sphericity even among individuals with normal measurements of LV systolic function and volumes.

Discussion

We have shown that the LV sphericity index is a predictor of incident cardiomyopathy and cardiomyopathy-related outcomes in healthy adults, independent of clinical factors and other imaging features. We were motivated to study LV sphericity by clinical experience and prior studies that suggested cardiac sphericity may be a marker of pathogenic remodeling. However, previous studies have focused primarily on sphericity after the onset of clinical disease^11-16. We have added to these prior works by using deep learning to conduct the first large-scale study of sphericity in a healthy population, and we have shown that sphericity has genetic underpinnings relevant to the pathophysiology of cardiomyopathy.

Our study found that LV sphericity has overlapping genetic architecture with clinical NICM. Mendelian randomization indicates that NICM is causal for sphericity, but sphericity is not causal for NICM. This finding is consistent with other imaging phenotypes. For example, increased chamber size or decreased systolic function are the result of cardiomyopathy rather than the cause of it. Sphericity thus predicts clinical outcomes because it is a marker of subclinical disease. Under this model, genetic and environmental factors lead to molecular/cellular dysfunction. These micro-level changes may eventually manifest as macro-level changes impacting cardiac size, morphology, and function. The presence of increased sphericity may therefore identify individuals with underlying molecular/cellular abnormalities that place them at heightened risk for developing overt cardiomyopathy or related diseases such as atrial fibrillation. Additional work is needed to determine if sphericity index provides clinical utility in current practice settings where patients are being screened for the development of cardiomyopathies or where diagnosed patients are being monitored for disease progression. An important first step towards such efforts will be the validation of sphericity index as a predictive tool when applied to echocardiograms of healthy subjects.

In conclusion, this study demonstrates the utility of using deep learning and advanced imaging analysis to define and study non-traditional cardiac imaging risk biomarkers using the large-scale data that is increasingly available in biobanks. While conventional imaging metrics have significant diagnostic and prognostic value, some of these measurements have been adopted out of convenience or tradition. By representing a specific multi-dimensional remodeling phenotype, sphericity has emerged as a distinct morphologic trait with features not adequately captured by conventional measurements. We expect that the search space of potential imaging measurements is vast, and we have only begun to scratch at the surface of disease associations.

Limitations of the Study

The first limitation of this study is the use of a single cohort. The UK Biobank is known to have a healthy cohort bias and does not represent a random sample of the UK population. Moreover, the subjects who underwent MRI may reflect an even healthier subset of UK Biobank participants (Supplementary Table 3). Given our goal was to study the relevance of sphericity in a healthy population, we do not expect this bias to significantly impact our results. Still, replication in additional cohorts will be necessary to validate the use of sphericity for clinical risk prediction. A second limitation is that in order to limit inflation due to population substructure, our GWAS included only a subset of unrelated participants of similar European ancestry. A third limitation is the reliance on diagnostic codes to identify cases for our study outcomes. Although we used a curated set of outcomes provided by the UK Biobank for the research community, some misclassification is likely to occur. We expect misclassification to weaken the observed associations in our outcomes analyses. A fourth limitation is that we used uniform thresholds to identify subjects with normal LV size and function measurements, and we did not account for sex differences in these measurements. The result is the inclusion of a small number of subjects with measurements just outside the normal range when using sex-specific ranges. This limitation is at least partially addressed by our models that adjust for measurements of LV size and function and for sex. Fifth, sphericity index may not fully capture phenotypic variation within the left ventricle, as additional features of variable trabeculation, angle, and thickness simultaneously influence cardiac function. We hope this work motivates more studies of novel imaging phenotypes that might capture such variability.

STAR Methods

Resource Availability

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Shoa Clarke (shoa@stanford.edu)

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

• This paper analyzes UK Biobank data, which is available after the approval of an application at https://www.ukbiobank.ac.uk.

• All original code has been deposited at https://github.com/vukadinovic936/UKB and is publicly available as of the date of publication. DOIs are listed in the key resources table.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Sphericity index GWAS summary statistics	This paper	https://my.locuszoom.org/gwas/834868/
LV short axis GWAS summary statistics	This paper	https://my.locuszoom.org/gwas/983285/
LV long axis GWAS summary statistics	This paper	https://my.locuszoom.org/gwas/865649/
Software and algorithms
Sphericity Index Extractor	This paper	https://github.com/vukadinovic936/sphericity
Survival 3.5	Therneau et al.	https://github.com/therneau/survival
Hail 0.2	Neale Lab	https://github.com/hail-is/hail
LDSC 1.01	Bulik-Sullivan et al.	https://github.com/bulik/ldsc
TwoSampleMR 0.5.6	Hemani et al.	https://github.com/MRCIEU/TwoSampleMR

Experimental Model and Subject Details

The UK Biobank (UKB) is a population-based cohort that links genetic and phenotypic data for approximately 500,000 adult participants from the United Kingdom^{30 31}. Age, gender, and race were self-reported. Our analyses focused on ~48,000 participants who have undergone cardiac MRI³². A subset of this group that met quality control filters and that had normal MRI measurements (LV end-diastolic volume 88-218 mL, LV end-systolic volume 31-97 mL, and LV ejection fraction 48-70%)³³ was used as the main cohort (Supplementary Figure 1). For the genome-wide association study (GWAS), we selected subjects from the main cohort who passed additional GWAS filters.

Method Details

Extraction of Sphericity Index Trait using Machine Learning

We downloaded cardiac MRI containing steady-state free precession image sequences of horizontal long-axis view. Each sequence is given as 210 × 208 × 50 matrix in DICOM format, with in-plane resolution 1.8 × 1.8 mm². In the pre-processing phase, we converted DICOM images and manual annotations (.XML) to NIfTI format, which saves the sequence as a single 3d image for the sake of better file management.

We used a fully convolutional neural network^5,34 for automated LV segmentation of MRI images. Trained weights were obtained from a prior work and inference was run on each video of each patient with cardiac MRIs in the UKB. For each individual, the semantic segmentation was manually evaluated and extreme outliers were excluded, as well as outliers in summary statistics provided from the UKB. We performed a segmentation quality check in two phases. We both heuristically checked (identifying when the segmentation was not continuous or implausibly small) and visually inspected segmentations to excluded patients with suboptimal segmentations.

For all images that passed quality check we created an automated bounding box around the segmentation of left ventricular blood pool (Figure 1a). The longer side of the rectangle was taken as LV long axis length (LVL), and the shorter side was taken as LV short axis length (LVS). Sphericity index is calculated as quotient of LVS and LVL $(SI=\frac{LVS}{LVL})$. Higher value of sphericity index indicates more spherical appearance, and lower value indicates an elongated appearance of left ventricular chamber.

The automated measurement of sphericity index was validated by comparison to human measurement in a set of 100 images randomly selected from our study. Two cardiologists independently measured sphericity index on these images in a blinded fashion. The human-human and human-machine similarities in estimates were evaluated by the coefficient of determination (R²) and the mean absolute error (MAE).

Quantification and Statistical Analyses

PheWAS

Diagnostic data was downloaded as ICD10 codes, processed, and saved as diagnostic groups organized by phecodes using a mapping file provided by PheWAS Catalog³⁵. Phenome-wide association studies (PheWAS)³⁶ were performed by measuring the Pearson r correlation between sphericity index and each phecode. In addition, three quantitative measurements were assessed: pulse rate, systolic blood pressure, and diastolic blood pressure. A total of 966 correlations were assessed. Significant associations were identified as those with a P value below the Bonferroni-corrected threshold of 5x10⁻⁵.

Cox Survival Analysis

We performed a time-to-event analysis to assess the association of sphericity index with incident cardiomyopathy, AF, heart failure, and cardiac arrest, using the UKB first occurrence data. These outcomes are provided by the UKB as part of a curated set of outcomes derived from health record diagnostic codes. We defined days without incident as the time from the first MRI visit to the first occurrence of the given outcome, death, or last follow-up (March 31st, 2021). For outcomes of atrial fibrillation, heart failure, and cardiomyopathy, participants with the prevalent disease at the time of the first MRI visit were excluded. We measured the association between the sphericity index and each outcome using a multivariable Cox proportional hazards models. Clinical covariates (age, pulse rate, BMI, hypertension, diabetes, prior myocardial infarction) were determined at the time of MRI. For this analysis, the sphericity index was normalized to mean equaling 0 and standard deviation (SD) equaling 1, and the hazard ratio (HR) is reported as per 1-SD. We additionally stratified the cohort into the lower 20th, middle 60th, and upper 20th percent of the sphericity index and plotted adjusted cumulative incidence curves by strata.

Genome-wide association study

We used the UKB imputed genotype calls in BGEN v1.2 format. Samples were genotyped using the UK BiLEVE or UK Biobank Axiom arrays. Imputation was performed using the Haplotype Reference Consortium panel and the UK10K+1000 Genomes panel³⁰. We used the QC files provided by UKB to create a GWAS cohort consisting of subjects who did not withdraw, were of inferred European ancestry, and were unrelated. Subjects with a genotype call rate < 0.98 were also removed. We considered variants with a minor allele frequency (MAF) ≥ 0.01, and we required genotyped variant to have a call rate ≥ 0.95 and imputed variants to have an INFO score ≥ 0.3. Variants with a Hardy-Weinberg equilibrium P value < 1x10⁻²⁰ were excluded. GWAS was done on a Spark 3.1.1 cluster, using library Hail 0.2 with Python version 3.6. The GWAS was adjusted for age at MRI and sex. We used the conventional P value of 5x10⁻⁸ as the threshold for defining genome-wide significance.

Genetic correlation

Genetic correlations were estimated using ldsc version 1.01³⁷. For GWAS performed in this study (sphericity index, short axis length, long axis length), we used well imputed variants by filtering for INFO ≥ 0.9. For LVEF, LVEDVi, and LVESVi, we used previously published summary statistics³, and we filtered SNPs for INFO ≥ 0.9. For NICM and AF, we used summary statistics from previously published GWAS^27,28, and we limited to well imputed variants by intersecting with HapMap3 variants (given INFO score was not provided). For all datasets, only variants with MAF ≥ 0.01 were used.

Mendelian randomization

Two-sample Mendelian randomization was conducted using the TwoSampleMR version 0.5.6 R package. We used a bidirectional approach with LV sphericity index as the exposure for the outcomes of NICM and AF and with LV sphericity as an outcome with NICM and AF as exposures. We used FinnGen²⁹ release 6 GWAS summary statistics for NICM (finngen_R6_I9_NONISCHCARDMYOP_STRICT) and AF (finngen_R6_I9_AF) to assure non-overlapping samples. The FinnGen NICM strict phenotype excludes hypertrophic cardiomyopathy. For each exposure, the instrument variable was created by identifying independent significant or suggestive single nucleotide polymorphisms (SNPs) that could be harmonized to the outcome of interest unambiguously. Independent SNPs were defined using a clumping distance of 10 megabases and an R² threshold of 0.001. Both sphericity index and NICM had a limited number of genome-wide significant variants after excluding palindromic SNPs and indels, and we therefore included SNPs with a suggestive P value of <10⁻⁶. This resulted in an 11-SNP instrument variable for sphericity index and a 7-SNP instrument variable for NICM. For AF, we only considered genome-wide significant SNPs (≤5x10⁻⁸), resulting in a 34-SNP instrument variable. We used inverse-variance weighted meta-analysis as our primary approach, and we used median weighted meta-analysis and MR Egger as sensitivity analyses.

Highlights.

• Measurement of left ventricular sphericity can be automated using deep learning.

• Sphericity predicts incident cardiomyopathy independent of traditional factors.

• Genome-wide association study of sphericity identifies loci relevant to cardiomyopathy.

Context and Significance.

Imaging plays an important role in the diagnosis and management of heart disease. Modalities such as ultrasound and magnetic resonance imaging (MRI) allow clinicians to visualize manifestations of heart disease, including enlargement of the heart chambers or decrease in the pumping strength of the heart muscle. Here, researchers hypothesized that beyond size and function, shape may carry additional information about heart health. They use artificial intelligence to analyze >30,000 heart MRIs and show that even when the size and function is normal, the roundness (or “sphericity”) of the heart predicts risk for heart conditions. Additionally, heart roundness is impacted by genetic influences. Consideration of heart shape may help improve our ability to identify people at risk for developing heart disease.