Abstract:Introduction The seismocardiography (SCG) signal is a measure of the audible and subaudible frequency chest vibration induced by heart valve, heart myocardial activity and blood flow. Early pilot data suggested that SCG changes may provide early indication of HF decompensation requiring hospitalization. The current study aimed to further investigate this approach with the prospective study of more patients, and an extended artificial intelligence signal processing approach. Methods SCG was acquired from 40 inpatient heart failure (HF) subjects (13 females, age 58.7± 9.8y) followed prospectively for 6 months, or until HF re-admission. The SCG waveforms were analyzed using a custom convolutional neural network (CNN) model to predict HF hospital readmission. SCG and thoracic impedance (Zoe, NMT, Las Vegas, NV) data were acquired during 98 sessions, 43% from 20 readmitted patients, with the remainder from those subjects not requiring HF hospital readmission. Signal pre-processing included filtering and clustering SCG beats into two clusters using unsupervised machine learning. The median waveforms of the two clusters were chosen to represent SCG of each session. Analysis was performed on the waveforms to yield time-frequency representations (TFR). The TFR data was randomly split into 49 (=50%) training, 19 (=20%) validation and 30 (=30%) test sets, then analyzed by a custom CNN model for classification. Results The results are shown in the table, where the actual outcomes of non-readmission compared to hospital HF readmission are compared to predictions based SCG using the CNN model. The numbers in the table represent the number of test data sessions in each class. The corresponding prediction sensitivity and specificity were 91% and 89%, respectively. For the same patient sets, the HF readmission predictive sensitivity and specificity utilizing thoracic impedance and was 57% and 50%, respectively. Similarly, NT-proBNP sensitivity and specificity for HF readmission, using a cutoff value of 900 pg/ml, was 75% and 54% respectively. Conclusions Prediction of HF hospital readmission based on SCG analysis using a custom CNN model in this preliminary data set appeared to be highly sensitive and specific; potentially significantly superior to predictions based on NT-proBNP or thoracic impendence. Further research is needed to validate these preliminary results.

Reconstruction of 3-Axis Seismocardiogram from Right-to-left and Head-to-foot Components Using A Long Short-Term Memory Network

Noninvasive Reconstruction of Cardiac Transmembrane Potentials Using A Kernelized Extreme Learning Method

Physiology and Geometry-Informed Deep Learning Framework for 12-Lead ECG Cardiac Transmembrane Potential Recovery

Cardiac Transmembrane Potential Imaging with GCN Based Iterative Soft Threshold Network

Analyzing SCG methodology for identification of ventricular depolarization events

Deep Learning for identifying systolic complexes in SCG traces: a cross-dataset analysis

A Residual U-Net Neural Network for Seismocardiogram Denoising and Analysis During Physical Activity

End-to-End Deep Learning for Reliable Cardiac Activity Monitoring using Seismocardiograms

Heart Failure Hospital Readmission Accurately Predicted By Seismocardiography Convolutional Neural Networks Modeling

Using deep learning to predict cardiovascular magnetic resonance findings from echocardiography videos

Multitask Learning for Improved Late Mechanical Activation Detection of Heart from Cine DENSE MRI

A Deep Learning Network for the Classification of Intracardiac Electrograms in Atrial Tachycardia

3-Lead to 12-Lead ECG Reconstruction: A Novel AI-based Spatio-Temporal Method

Automated segmentation of long and short axis DENSE cardiovascular magnetic resonance for myocardial strain analysis using spatio-temporal convolutional neural networks

Error correcting 2D-3D cascaded network for myocardial infarct scar segmentation on late gadolinium enhancement cardiac magnetic resonance images

Early prediction of sudden cardiac death using multimodal fusion of ECG Features extracted from Hilbert-Huang and wavelet transforms with explainable vision transformer and CNN models

Non-contact heart vibration measurement using computer vision-based seismocardiography

CS-based multi-task learning network for arrhythmia reconstruction and classification using ECG signals

A deep learning approach to using wearable seismocardiography (SCG) for diagnosing aortic valve stenosis and predicting aortic hemodynamics obtained by 4D flow MRI

Successive Subspace Learning for Cardiac Disease Classification with Two-phase Deformation Fields from Cine MRI

[Early classification and recognition algorithm for sudden cardiac arrest based on limited electrocardiogram data trained with a two-stages convolutional neural network]