Preliminary results of a single, easy-to-use device for multimodal detection of intervals associated with heart failure decompensation

Figures

• 

  Figure 1. Schematic representation of the device used. The device incorporates sensors on both faces. The face contacting the patient’s chest houses the electrodes and the membrane. On the opposite face, within a groove designed to accommodate the patient’s finger while holding the device, is the photoplethysmography sensor. Example of use: the patient holds the device in the left hand, inserts a finger into the groove, and presses the device against the precordial area, ensuring contact of the sensors (electrodes and membrane) with the chest.

• 

  Figure 2. Schematic of the main steps in the signal-processing pipeline for interval extraction. The common step of grouping detections by patient and acquisition date, used in all analyses, is shown in orange. The phonocardiography signal and its processing are shown in black, the electrocardiography in blue, and the pulse-wave photoplethysmography in red. Note that the robustness of phonocardiography-to-R-wave coupling analyses depends on reliable detection of valid beats only; therefore, very restrictive criteria were applied. Extraction of all intervals relies on integration of the three channels, but intervals can still be obtained when one channel is missing: ejection period is derived from the photoplethysmography; pre-ejection period from photoplethysmography and electrocardiography; electromechanical activation time from electrocardiography and phonocardiography; and isovolumetric contraction time from calculations combining electromechanical activation time and ejection period. The figure presents a multi-stage signal-processing pipeline for extracting cardiac timing intervals from concurrent phonocardiography, electrocardiography, and photoplethysmography recordings. A common…

• 

  Figure 3. Cohort analysis flowchart. Forty-one patients were recruited. Thirty-six were included in the computational analysis. Twenty-six patients comprised the acute heart failure decompensation cohort, with measurements available at admission (acute heart failure decompensation) and at discharge (non-acute heart failure decompensation: euvolaemic state). Ten patients admitted with a diagnosis of acute coronary syndrome and without heart failure were used as controls. Five patients were excluded: two cases presented with severe aortic regurgitation on admission and underwent aortic valve replacement during the same hospitalization; consequently, only a single measurement under acute decompensated heart failure (acute heart failure decompensation) conditions at admission was available for these patients. Three cases were excluded because their clinical characteristics or clinical course rendered them ineligible for inclusion in the analysis: one case of myocarditis met the N-terminal pro-B-type natriuretic peptide threshold for inclusion in the acute heart failure decompensation group but did not exhibit signs of congestion; a second case was reclassified after admission as an…

• 

  Figure 4. Analysis of clinically relevant intervals in the cohort. (A) Paired comparison of electromechanical activation time in the heart-failure admission group (0.148 ± 0.051 vs. 0.108 ± 0.039; n = 20; Padj = 0.0005). (B) Paired comparison of ejection period in the heart-failure admission group (0.257 ± 0.078 vs. 0.344 ± 0.082; n = 21; Padj = 0.00003). (C) Paired comparison of isovolumetric contraction time in the heart-failure admission subgroup (0.152 ± 0.098 vs. 0.071 ± 0.023; n = 17; Padj = 0.0018). (D) Comparison of electromechanical activation time distributions in the aggregated analysis (0.163 ± 0.055 vs. 0.088 ± 0.037; n = 29; Padj = 0.0006). (E) Comparison of ejection period distributions in the aggregated analysis (0.284 ± 0.084 vs. 0.343 ± 0.078; n = 32; Padj = 0.019). (F) Comparison of isovolumetric contraction time distributions in the aggregated analysis (0.111 ± 0.021 vs. 0.051 ± 0.011; n = 22; Padj = 0.047). The aggregated analysis-entire study cohort comparison group included, in the acute heart failure decompensation group, admission measurements from patients admitted for heart failure and in the non-acute heart failure decompensation group comprised the second…

• 

  Figure 5. Contextualization of electromechanical activation time and ejection period with the healthy reference cohort. (A) Visual comparison of distributions. Because sample sizes differ, distributions are shown as estimated probability density functions (pdf) to facilitate visual comparison. A prolongation of electromechanical activation time is evident in hospitalized patients relative to the healthy cohort, most pronounced in the acute heart failure decompensation group. Sample sizes and mean values are indicated in the plot. (B) Hypothesis testing shows a statistically significant prolongation of electromechanical activation time in the acute heart failure decompensation group vs. both the non-acute heart failure decompensation patients and the healthy cohort (102 ± 26 ms vs. 77 ± 14 ms, Padj = 0.0006, n = 20 and 23; 102 ± 26 ms vs. 69 ± 14 ms, Padj = 0.000000001, n = 20 and 314). (C) Comparison of ejection period distributions expressed as a fraction of the RR interval. A subset of 30 healthy individuals was selected from the original publicly available healthy reference cohort and measured semi-automatically using the same method applied in the study cohort. Ejection period is…

• 

  Figure 6. Receiver operating characteristic curves for each interval: electromechanical activation time (A), ejection period/RR (B), and isovolumetric contraction time/RR (C). Each panel displays the estimated area under the curve, the Youden-index threshold with corresponding sensitivity and specificity, and the negative predictive value and positive predictive value obtained using that threshold.