Detecting Aortic Valve-Induced Abnormal Flow with Seismocardiography and Cardiac MRI

Abstract

Cardiac MRI (CMR) techniques offer non-invasive visualizations of cardiac morphology and function. However, imaging can be time-consuming and complex. Seismocardiography (SCG) measures physical vibrations transmitted through the chest from the beating heart and pulsatile blood flow. SCG signals can be acquired quickly and easily, with inexpensive electronics. This study investigates relationships between CMR metrics of function and SCG signal features. Same-day CMR and SCG data were collected from 28 healthy adults and 6 subjects with aortic valve disease history. Correlation testing and statistical median/decile calculations were performed with data from the healthy cohort. MR-quantified flow and function parameters in the healthy cohort correlated with particular SCG energy levels, such as peak aortic velocity with low-frequency SCG (coefficient 0.43, significance 0.02) and peak flow with high-frequency SCG (coefficient 0.40, significance 0.03). Valve disease-induced flow abnormalities in patients were visualized with MRI, and corresponding abnormalities in SCG signals were identified. This investigation found significant cross-modality correlations in cardiac function metrics and SCG signals features from healthy subjects. Additionally, through comparison to normative ranges from healthy subjects, it observed correspondences between pathological flow and abnormal SCG. This may support development of an easy clinical test used to identify potential aortic flow abnormalities.

Keywords: 4D flow MRI; Aortic valve disease; Valve disease seismocardiography.

Figure 1.

Seismocardiogram signals (left) are formed from ECG and chest-external acceleration measurements recorded by a wearable cardiac sensor. Complementary 4D flow cardiac MR imaging with 2D CINE SSFP (middle) and 4D phase contrast (right) is used to quantify left ventricular function, valve movements and thoracic aorta blood flow dynamics. alve movements annotated included the aortic valve (right, ‘aort’) and mitral valve (right, ‘mitr’) times of opening (right, ‘op’) and close (right, ‘cl’).

Figure 2.

In all plots, data for each subject are colored an arbitrary shade between magenta and cyan, with the shades used consistently across plots. The ages and sexes of subjects are shown (a); the heart rates recorded during each type of measurement were compared (‘SAX’: short-a is CMR, ‘4DF’: 4D flow MR), with heart rate during SAX vs. heart rate during SCG colored blue, 4DF vs SCG colored red, and the SAX and 4DF rates connected by the subject-particular colored line; and the left-ventricular volumes, blood velocities and flow rates were aggregated for comparison to SCG. An example of the time-scaling and time-shifting operations with two subjects is shown (b, left, middle), and the resulting dynamic distribution of flow rates calculated from the entire healthy cohort is also shown (b, right).

Figure 3.

The effects of using or not using cardiac valve-event shifts are also compared (bottom and top, respectively). Median valve times in the cohort are marked (red) within their corresponding windows.

Figure 4.

Metrics used for comparison of SCG and CMR across subjects were calculated from the data collected in the cohort. The beat-synchronized acceleration measurements (left, i) were decomposed by STFT for each beat (left, ii) and averaged across beats to find each a characteristic SCG spectrum (left, iii). The average energies in several time-frequency regions were used as features for comparison. Two example reference time markings (0.025s, 0.030s) are shown to illustrate correspondences between recorded acceleration and decomposed short-time spectra. CMR metrics of cardiac and flow function, such as stroke volume, maximal ejection rate, maximal blood velocity, extremal blood accelerations, and maximal flow (right) are calculated from the processed CMR images.

Figure 5.

Normative median and decile ranges for SCG energies (middle) were aggregated from the beat-averaged SCG spectra of the healthy cohort (top). Individual subjects’ SCG energies (right) were compared to the quantiles to classify each frequency and time point according to deciles of energy levels. Decile rankings were coded into heat maps (left) indicating distance from the median, so that for example, a frequency and time point with SCG power exceeding that of 90% of the normal cohorts’ powers is shown light pink; and, conversely, a point at which the power is less than 90% of the normal cohorts’ powers (10% decile) is shown black.

Figure 6.

(a) Distributions for a selection of CMR and SCG metrics are displayed (left) with the correlation coefficient values (left, red: significant, grey: not significant) and linear regression fits (left, dark red lines). Some SCG and CMR metrics that covary with underlying demographic trends are also shown (right). Several metrics are abbreviated in their plot labels from the names described in the Methods section: max(V) is Max(V_avg), max(A) is Max(A_avg), H is height, LDHF is latedia hi-f, ESLF is earlysys lo-f, LSLF is latesys lo-f. (b) Aligning SCG energy features to cardiac valve events (aortic open or close) before comparing to CMR metrics causes some correlations to gain or lose significance as compared to a global comparison.

Figure 7.

Analyses from 4D flow MRI (left 3 columns) and SCG spectral heat maps (right column) from one randomly-selected healthy control (top row), three BAV subjects (middle 3 rows), and two mechanical valve subjects (bottom 2 rows) with different implants (brand names listed for each). The third mechanical valve subject had the same valve brand as the first, and the subjects had similar flow and energy deviations. Velocity and flow quantifications are from a cutplane in the ascending aorta.

Figure 8.

Distributions of SCG energy in the normal subject cohort (blue histograms) are shown for the global IJdev feature and the local time-frequency-averaged early/end systole hi-f metrics when aligned to either aortic open or aortic close. Means (μ) and standard deviations (σ) for the normative distributions are annotated on each histogram. Measurements from BAV and mechanical valve patients are also overlaid; when a subject’s measured value saturates on the normative scale, the actual value is shown parenthetically.