Blood pressure monitoring during anesthesia induction using PPG morphology features and machine learning.

Blood pressure (BP) is a crucial biomarker giving valuable information regarding cardiovascular diseases but requires accurate continuous monitoring to maximize its value. In the effort of developing non-invasive, non-occlusive and continuous BP monitoring devices, photoplethysmography (PPG) has recently gained interest. Researchers have attempted to estimate BP based on the analysis of PPG waveform morphology, with promising results, yet often validated on a small number of subjects with moderate BP variations. This work presents an accurate BP estimator based on PPG morphology features. The method first uses a clinically-validated algorithm (oBPM®) to perform signal preprocessing and extraction of physiological features. A subset of features that best reflects BP changes is automatically identified by Lasso regression, and a feature relevance analysis is conducted. Three machine learning (ML) methods are then investigated to translate this subset of features into systolic BP (SBP) and diastolic BP (DBP) estimates; namely Lasso regression, support vector regression and Gaussian process regression. The accuracy of absolute BP estimates and trending ability are evaluated. Such an approach considerably improves the performance for SBP estimation over previous oBPM® technology, with a reduction in the standard deviation of the error of over 20%. Furthermore, rapid BP changes assessed by the PPG-based approach demonstrates concordance rate over 99% with the invasive reference. Altogether, the results confirm that PPG morphology features can be combined with ML methods to accurately track BP variations generated during anesthesia induction. They also reinforce the importance of adding a calibration measure to obtain an absolute BP estimate.

Feature Learning for Blood Pressure Estimation from Photoplethysmography.

Blood pressure (BP) is an important indicator for prevention and management of cardiovascular diseases. Alongside the improvement in sensors and wearables, photoplethysmography (PPG) appears to be a promising technology for continuous, non-invasive and cuffless BP monitoring. Previous attempts mainly focused on features extracted from the pulse morphology. In this paper, we propose to remove the feature engineering step and automatically generate features from an ensemble average (EA) PPG pulse and its derivatives, using convolutional neural network and a calibration measurement. We used the large VitalDB dataset to accurately evaluate the generalization capability of the proposed model. The model achieved mean errors of -0.24 ± 11.56 mmHg for SBP and -0.5 ± 6.52 mmHg for DBP. We observed a considerable reduction in error standard deviation of above 40% compared to the control case, which assumes no BP variation. Altogether, these results highlight the capability to model the dependency between PPG and BP.

Machine Learning Approaches For Improved Continuous, Non-occlusive Arterial Pressure Monitoring Using Photoplethysmography.

Arterial pressure (AP) is a crucial biomarker for cardiovascular disease prevention and management. Photoplethysmography (PPG) could provide a novel, paradigm-shifting approach for continuous, non-obtrusive AP monitoring, comfortably integrated in wearable and mobile devices; yet, it still faces challenges in accuracy and robustness. In this work, we sought to integrate machine learning (ML) techniques into a previously established, clinically-validated classical approach (oBPM®) to develop new accurate AP estimation tools based on PPG, and at the same time improve our understanding of the underlying physiological parameters. In this novel approach, oBPM® was used to pre-process PPG signals and robustly extract physiological features, and ML models were trained on these features to estimate systolic AP (SAP). A feature relevance analysis showed that reference (calibration) information, followed by various morphological parameters of the PPG pulse wave, comprised the most important features for SAP estimation. A performance analysis then revealed that LASSO-regularized linear regression, Gaussian process regression and support vector regression are effective for SAP estimation, particularly when operating on reduced feature sets previously obtained with e.g. LASSO. These approaches yielded substantial reductions in error standard deviation of 9-15% relative to conventional oBPM®. Altogether, these results indicate that ML approaches are well-suited, and promising tools to help overcoming the challenges of ubiquitous AP monitoring.

Ultrastructural basis of strong unitary inhibition in a binaural neuron.

KEY POINTS: Neurons of the lateral superior olive (LSO) in the brainstem receive powerful glycinergic inhibition that originates from the contralateral ear, and that plays an important role in sound localization. We investigated the ultrastructural basis for strong inhibition of LSO neurons using serial block face scanning electron microscopy. The soma and the proximal dendrite of an LSO neuron are surrounded by a high density of inhibitory axons, whereas excitatory axons are much sparser. A given inhibitory axon establishes contacts via several large axonal thickenings, called varicosities, which typically elaborate several active zones (range 1-11). The number of active zones across inhibitory axon segments is variable. These data thus provide an ultrastructural correlate for the strong and multiquantal, but overall variable, unitary IPSC amplitude observed for inhibitory inputs to LSO neuron. ABSTRACT: Binaural neurons in the lateral superior olive (LSO) integrate sound information arriving from each ear, and powerful glycinergic inhibition of these neurons plays an important role in this process. In the present study, we investigated the ultrastructural basis for strong inhibitory inputs onto LSO neurons using serial block face scanning electron microscopy. We reconstructed axon segments that make contact with the partially reconstructed soma and proximal dendrite of a mouse LSO neuron at postnatal day 18. Using functional measurements and the Sr2+ method, we find a constant quantal size but a variable quantal content between 'weak' and 'strong' unitary IPSCs. A 3-D reconstruction of a LSO neuron and its somatic synaptic afferents reveals how a large number of inhibitory axons intermingle in a complex fashion on the soma and proximal dendrite of an LSO neuron; a smaller number of excitatory axons was also observed. A given inhibitory axon typically contacts an LSO neuron via several large varicosities (average diameter 3.7 µm), which contain several active zones (range 1-11). The number of active zones across individual axon segments was highly variable. These data suggest that the variable unitary IPSC amplitude is caused by a variable number of active zones between inhibitory axons that innervate a given LSO neuron. The results of the present study show that relatively large multi-active zone varicosities, which can be repeated many times in a given presynaptic axon, provide the ultrastructural basis for the strong multiquantal inhibition received by LSO neurons.