Timely identification of the need for lifesaving intervention in battlefield conditions may be improved through automated monitoring of the injured warfighter. Technologies that combine maximal noninvasive insight with minimal equipment footprint give the greatest opportunity for deployment at scale with inexperienced providers in forward areas. Finger photoplethysmography (PPG) signatures are associated with impending hemorrhagic shock but may be insufficient alone. Transthoracic impedance (TTI) monitoring is a complementary modality to PPG and able to identify volume loss and estimate functional cardiovascular parameters. We sought to understand how PPG features correlate with volume loss estimation from TTI during lower body negative pressure (LBNP) challenge. We hypothesized that features of the PPG waveform would correlate with thoracic fluid content (TFC) as measured by TTI. We obtained physiologic monitoring data from healthy adult subjects in LBNP hemorrhagic shock models after local Institutional Review Board and DoD Human Research Protection Office approval. Subjects were excluded for pregnancy, age >45 years, and conditions prohibitive of LBNP exposure. Subjects were instrumented with noninvasive sensors, including a finger PPG sensor and a TTI monitor. Subjects underwent a stepwise LBNP exposure program of -10 mmHg every 10 minutes and notified laboratory staff at first sign of near syncope, terminating the sequential program. TTI data were continuously streamed to a custom program written in MATLAB and time synchronized. To calculate PPG measures, we downsampled data to 250 Hz, screened, and parsed each beat. We featurized each beat to include a systolic, diastolic, and dicrotic notch peak, beat length and area under the curve (AUC), peak-to-peak systolic/diastolic interval, and leading/trailing slopes, all normalized to instantaneous heart rate. Thoracic fluid content was normalized to subjects' pre-LBNP baselines. We summarized all PPG features and the TFC using means (SD) generated as a subject average for each step. We used generalized estimating equation models to examine the relationship between TFC and PPG features while controlling for LBNP stage and subject. Thirty-two subjects were enrolled; 4 participants were excluded because of sensor malfunction. Twenty-eight subjects had a mean (SD) age of 25.11 (6.66) years. A total of 35.7% of subjects were female. Photoplethysmography analysis demonstrated a decreased systolic-diastolic peak interval, diastolic peak height, and beat AUC with decreased LBNP pressure. End-stage baseline normalized TFC showed an average decrease of 14.68% (±4.98%) (range: 7.54% to 27.69%). The strongest average correlations between stage TFC and PPG occurred in beat length (0.68) and normalized AUC (0.69). In generalized estimating equation models incorporating all stages, beat length, normalized AUC, and the systolic-diastolic interval were all significantly associated with time as a function of LBNP level (P < .001). Thoracic fluid content began decreasing at 12.8 (4.7) minutes, the normalized AUC decreased at 20.7 (7.2) minutes, the beat length decreased at 20.9 (7.0) minutes, and the systolic-diastolic time interval decreased at 30.6 (16.7) minutes. While both PPG features and impedance-based TFC trend congruently in the perishock state following LBNP exposure, peripheral pulse wave signals lag redistribution of thoracic fluid volume. Photoplethysmography features of beat length and normalized AUC may serve as a surrogate for TFC when direct thoracic sensing is not available.
Read full abstract