Towards Personalized Closed-Loop Mechanical CPR: A Model Relating Carotid Blood Flow to Chest Compression Rate and Duration.

OBJECTIVE: There is a growing interest in the personalization of chest compressions to increase blood flow during cardiopulmonary resuscitation (CPR), but there has been very little systematic work to test the feasibility of a closed loop mechanical CPR system. The purpose of this study is to determine if it is possible to model the response of the carotid blood flow to different chest compression waveforms as a function of time during resuscitation from cardiac arrest. This work tests several approaches to predict the carotid blood flow generated by the next chest compression based on knowledge of the duration of resuscitation, the chest compression rate, and the last compression's carotid blood flow. METHODS: Using an existing physiological database from swine cardiac arrest studies, we computed the features of CPR epoch, compression index, compression rate, and the previous carotid blood flow and used them as the inputs to our model in order to predict carotid blood flow using a Random Forest algorithm. We tested animal specific (estimated with data from a single animal) and global (estimated with data from all but one animals) models for effectiveness. RESULTS: Animal specific models did not generalize when applied to the rest of the animals. The global model performed reasonably well when trained on six animals and tested on the 7th, resulting in errors of 40-160 µL per compression, compared to an average of approximately 400 µL net carotid blood flow per compression in early compressions. In addition, the global model highlighted the inter-animal variability in carotid blood flow generated by identical chest compression waveforms. Generation of probability distribution functions of carotid blood flows suggested at least three different distribution profiles in seven animals. CONCLUSION: A single physiological metric, carotid blood flow, combined with information about the duration of resuscitation and the compression rate was sufficient to model and predict carotid blood flow in the next compression. SIGNIFICANCE: This demonstrates that the physiological response to chest compression can be predicted from a relatively modest data set. This suggests that closed loop mechanical CPR is a viable medical device target.

Closed-loop neuromuscular electrical stimulation using feedforward-feedback control and textile electrodes to regulate grasp force in quadriplegia.

BACKGROUND: Transcutaneous neuromuscular electrical stimulation is routinely used in physical rehabilitation and more recently in brain-computer interface applications for restoring movement in paralyzed limbs. Due to variable muscle responses to repeated or sustained stimulation, grasp force levels can change significantly over time. Here we develop and assess closed-loop methods to regulate individual finger forces to facilitate functional movement. We combined this approach with custom textile-based electrodes to form a light-weight, wearable device and evaluated in paralyzed study participants. METHODS: A textile-based electrode sleeve was developed by the study team and Myant, Corp. (Toronto, ON, Canada) and evaluated in a study involving three able-body participants and two participants with quadriplegia. A feedforward-feedback control structure was designed and implemented to accurately maintain finger force levels in a quadriplegic study participant. RESULTS: Individual finger flexion and extension movements, along with functional grasping, were evoked during neuromuscular electrical stimulation. Closed-loop control methods allowed accurate steady state performance (< 15% error) with a settling time of 0.67 s (SD = 0.42 s) for individual finger contact force in a participant with quadriplegia. CONCLUSIONS: Textile-based electrodes were identified to be a feasible alternative to conventional electrodes and facilitated individual finger movement and functional grasping. Furthermore, closed-loop methods demonstrated accurate control of individual finger flexion force. This approach may be a viable solution for enabling grasp force regulation in quadriplegia. TRIAL REGISTRATION: NCT, NCT03385005. Registered Dec. 28, 2017.

Corrigendum: Feature Selection Methods for Robust Decoding of Finger Movements in a Non-human Primate.

[This corrects the article on p. 22 in vol. 12, PMID: 29467602.].

Feature Selection Methods for Robust Decoding of Finger Movements in a Non-human Primate.

Objective: The performance of machine learning algorithms used for neural decoding of dexterous tasks may be impeded due to problems arising when dealing with high-dimensional data. The objective of feature selection algorithms is to choose a near-optimal subset of features from the original feature space to improve the performance of the decoding algorithm. The aim of our study was to compare the effects of four feature selection techniques, Wilcoxon signed-rank test, Relative Importance, Principal Component Analysis (PCA), and Mutual Information Maximization on SVM classification performance for a dexterous decoding task. Approach: A nonhuman primate (NHP) was trained to perform small coordinated movements-similar to typing. An array of microelectrodes was implanted in the hand area of the motor cortex of the NHP and used to record action potentials (AP) during finger movements. A Support Vector Machine (SVM) was used to classify which finger movement the NHP was making based upon AP firing rates. We used the SVM classification to examine the functional parameters of (i) robustness to simulated failure and (ii) longevity of classification. We also compared the effect of using isolated-neuron and multi-unit firing rates as the feature vector supplied to the SVM. Main results: The average decoding accuracy for multi-unit features and single-unit features using Mutual Information Maximization (MIM) across 47 sessions was 96.74 ± 3.5% and 97.65 ± 3.36% respectively. The reduction in decoding accuracy between using 100% of the features and 10% of features based on MIM was 45.56% (from 93.7 to 51.09%) and 4.75% (from 95.32 to 90.79%) for multi-unit and single-unit features respectively. MIM had best performance compared to other feature selection methods. Significance: These results suggest improved decoding performance can be achieved by using optimally selected features. The results based on clinically relevant performance metrics also suggest that the decoding algorithm can be made robust by using optimal features and feature selection algorithms. We believe that even a few percent increase in performance is important and improves the decoding accuracy of the machine learning algorithm potentially increasing the ease of use of a brain machine interface.