This paper describes an integrated chest wearable apparatus for continuous blood pressure (BP) measurements exploiting the heart-rate (HR) and the pulse arrival time (PAT) in a modified Moens-Korteweg model. The device embeds a miniaturized gas pressure sensor to record the phonocardiogram (PCG) of the heart sounds, a LED-photodiode pair to detect the photoplethysmogram (PPG) of the blood flow wave, a μcontroller, a wireless communication module and the power supply. With the proposed device no active participation would be required from human subjects for BP measurements, since the HR and the PAT are continuously extracted from the PCG and PPG signals. Dedicated signal processing algorithms were developed and implemented off-line to extract both HR and PAT. A subject-specific calibration protocol of the BP model was designed and implemented. The calibration and validation of the apparatus were performed on a cohort of 20 healthy subjects. A GIMA ABPM pressure Holter was chosen as reference device, and 8 measurements points, evenly distributed over a 10-minute interval, were used for model calibration for each subject. The range of DBP and SBP measurements were 52-85 mmHg and 90-141 mmHg, respectively. The results from Bl- and-Altman analysis showed that the mean±1.96SD for the estimated diastolic, systolic, and mean BP with the proposed method against reference were 0.01±7.55, 1.47±3.76, 0.74±4.38 mmHg, respectively. The corresponding mean absolute differences (MAD) were 3.06, 1.83, and 1.80 mmHg. These results demonstrates that the acquisition apparatus is able to continuously estimate the BP with an accuracy comparable to traditional cuff-based devices.

The aim of our study was to study regional lung function by standard computed tomography (CT) and characterise regional variations of density and specific gas volume (SVg) between different lung volumes.We studied 10 healthy and 10 severely emphysematous subjects. Corresponding CT images taken at high and low lung volumes were registered by optical flow to obtain two-dimensional maps of pixel-by-pixel differences of density (ΔHU) and SVg (ΔSVg) at slice levels near the aortic arch, carina and top diaphragm.In healthy subjects, ΔHU was higher at all levels (p<0.001) with higher variability expressed as interquartile range (p<0.001), largely due to its differences between dorsal and ventral regions. In patients, median ΔSVg values were 3.2 times lower than healthy volunteers (p<0.001), while heterogeneity of ΔSVg maps, expressed as quartile coefficient of variation, was 5.4 times higher (p<0.001). In all patients, there were areas with negative values of ΔSVg.In conclusion, ΔSVg is uniform in healthy lungs and minimally influenced by gravity. The significant ΔSVg heterogeneity observed in emphysema allows identification of areas of alveolar destruction and gas trapping and suggests that ΔSVg maps provide useful information for evaluation and planning of emerging treatments that target trapped gas for removal.

Evidence is rapidly accumulating that multifactorial nocturnal monitoring, through the coupling of wearable devices and deep learning, may be disruptive for early diagnosis and assessment of sleep disorders. In this work, optical, differential air-pressure and acceleration signals, acquired by a chest-worn sensor, are elaborated into five somnographic-like signals, which are then used to feed a deep network. This addresses a three-fold classification problem to predict the overall signal quality (normal, corrupted), three breathing-related patterns (normal, apnea, irregular) and three sleep-related patterns (normal, snoring, noise). In order to promote explainability, the developed architecture generates additional information in the form of qualitative (saliency maps) and quantitative (confidence indices) data, which helps to improve the interpretation of the predictions. Twenty healthy subjects enrolled in this study were monitored overnight for approximately ten hours during sleep. Somnographic-like signals were manually labeled according to the three class sets to build the training dataset. Both record- and subject-wise analyses were performed to evaluate the prediction performance and the coherence of the results. The network was accurate (0.96) in distinguishing normal from corrupted signals. Breathing patterns were predicted with higher accuracy (0.93) than sleep patterns (0.76). The prediction of irregular breathing was less accurate (0.88) than that of apnea (0.97). In the sleep pattern set, the distinction between snoring (0.73) and noise events (0.61) was less effective. The confidence index associated with the prediction allowed us to elucidate ambiguous predictions better. The saliency map analysis provided useful insights to relate predictions to the input signal content. While preliminary, this work supported the recent perspective on the use of deep learning to detect particular sleep events in multiple somnographic signals, thus representing a step towards bringing the use of AI-based tools for sleep disorder detection incrementally closer to clinical translation.

The accurate detection of physiologically-related events in photopletismographic (PPG) and phonocardiographic (PCG) signals, recorded by wearable sensors, is mandatory to perform the estimation of relevant cardiovascular parameters like the heart rate and the blood pressure. However, the measurement performed in uncontrolled conditions without clinical supervision leaves the detection quality particularly susceptible to noise and motion artifacts. This work proposes a new fully-automatic computational framework, based on convolutional networks, to identify and localize fiducial points in time as the foot, maximum slope and peak in PPG signal and the S1 sound in the PCG signal, both acquired by a custom chest sensor, described recently in the literature by our group. The event detection problem was reframed as a single hybrid regression-classification problem entailing a custom neural architecture to process sequentially the PPG and PCG signals. Tests were performed analysing four different acquisition conditions (rest, cycling, rest recovery and walking). Cross-validation results for the three PPG fiducial points showed identification accuracy greater than 93 % and localization error (RMSE) less than 10 ms. As expected, cycling and walking conditions provided worse results than rest and recovery, however reaching an accuracy greater than 90 % and a localization error less than 15 ms. Likewise, the identification and localization error for S1 sound were greater than 90 % and less than 25 ms. Overall, this study showcased the ability of the proposed technique to detect events with high accuracy not only for steady acquisitions but also during subject movements. We also showed that the proposed network outperformed traditional Shannon-energy-envelope method in the detection of S1 sound, reaching detection performance comparable to state of the art algorithms. Therefore, we argue that coupling chest sensors and deep learning processing techniques may disclose wearable devices to unobtrusively acquire health information, being less affected by noise and motion artifacts.

Purpose To quantify regional gas trapping in the lung by using computed tomographic (CT)–determined specific gas volume and hyperpolarized helium 3 (³He) magnetic resonance (MR) imaging in a porcine model of airway obstruction. Materials and Methods Four porcine lungs were removed after sacrifice for unrelated cardiac experiments, for which animal studies approval was obtained. Dynamic expiratory thin-section CT and ³He MR imaging were performed during passive deflation from total lung capacity after obstructions were created with inverted one-way endobronchial exit valves in segmental or lobar bronchi to produce identifiable regions of trapped gas. Changes in specific gas volume were assessed from CT data for defined regions of interest within and outside of obstructed segments and for entire lobes. Helium 3 data were analyzed according to the corresponding regional signal reduction during expiration, compared with the total magnetic moment at each time point. Results In 4.5 seconds of free collapse, volume decreased by 6% ± 2 (standard error) and 53% ± 3, respectively, in trapped-gas lobes and in unobstructed regions (P < .0001). Specific gas volume changed by 6% ± 2 in areas of trapped gas and decreased by 56% ± 3 in unobstructed regions, from 3.4 mL/g ± 0.2 to 1.5 mL/g ± 0.1 (P < .0001). The ³He signal intensity decreased by 25% ± 6 and 71% ± 3, respectively, in trapped-gas and normal regions (P = .0008). In unobstructed regions, the percentage decreases in specific gas volume and ³He signal intensity were not statistically different from one another (P = .89). Conclusion The results obtained from the model of gas trapping demonstrate that CT-determined specific gas volume and ³He MR imaging can help identify and quantify the extent of regional trapped gas in explanted porcine lungs. © RSNA, 2009