Physiotherapy - Research Publications
Permanent URI for this collection
Search Results
1 - 4 of 4
-
ItemUncertainty-aware non-invasive patient–ventilator asynchrony detection using latent Gaussian mixture generative classifier with noisy label correction(Springer Science and Business Media LLC, 2024-01-01)Abstract Patient–ventilator asynchrony (PVA) refers to instances where a mechanical ventilator’s cycles are desynchronised from the patient’s breathing efforts, and may result in patient discomfort and potential ineffective ventilation. Typically, they are identified with constant monitoring by trained clinicians. Such expertise is often limited; therefore, it is desirable to automate PVA detection with machine learning methods. However, there are three major challenges to applying machine learning to the problem: data collected from non-invasive ventilation are often noisy, there exists high variability between patients or between setting changes, and manual annotations of PVA events are not always consistent. To produce meaningful inference from such noisy data, a model needs to not only provide a measure of uncertainty, but also take into account potential inconsistencies in the training signal it is based on. In this work, we propose a conditional latent Gaussian mixture generative classifier with noisy label correction, which is capable of capturing variations within and between classes, providing well-calibrated class probabilities, detecting unlikely input instances that deviates from training data, while also taking into account possible mislabelling of event classes. We show that our model is able to match the performance of a well-tuned gradient boosting classifier, but also produce better calibrated predictions and smaller performance variability between patients.
-
ItemRapidly and slowly progressive neuromuscular disease: differences in pulmonary function, respiratory tract infections and response to lung volume recruitment therapy (LVR)(BMJ PUBLISHING GROUP, 2022-12)INTRODUCTION: Reduced lung volumes are a hallmark of respiratory muscle weakness in neuromuscular disease (NMD). Low respiratory system compliance (Crs) may contribute to restriction and be amenable to lung volume recruitment (LVR) therapy. This study evaluated respiratory function and the immediate impact of LVR in rapidly progressive compared to slowly progressive NMD. METHODS: We compared vital capacity (VC), static lung volumes, maximal inspiratory and expiratory pressures (MIP, MEP), Crs and peak cough flow (PCF) in 80 adult participants with motor neuron disease ('MND'=27) and more slowly progressive NMDs ('other NMD'=53), pre and post a single session of LVR. Relationships between respiratory markers and a history of respiratory tract infections (RTI) were examined. RESULTS: Participants with other NMD had lower lung volumes and Crs but similar reduction in respiratory muscle strength compared with participants with MND (VC=1.30±0.77 vs 2.12±0.75 L, p<0.001; Crs=0.0331±0.0245 vs 0.0473±0.0241 L/cmH2O, p=0.024; MIP=39.8±21.3 vs 37.8±19.5 cmH2O). More participants with other NMD reported an RTI in the previous year (53% vs 22%, p=0.01). The likelihood of having a prior RTI was associated with baseline VC (%predicted) (OR=1.03 (95% CI 1.00 to 1.06), p=0.029). Published thresholds (VC<1.1 L or PCF<270 L/min) were, however, not associated with prior RTI.A single session of LVR improved Crs (mean (95% CI) increase = 0.0038 (0.0001 to 0.0075) L/cmH2O, p=0.047) but not VC. CONCLUSION: These findings corroborate the hypothesis that ventilatory restriction in NMD is related to weakness initially with respiratory system stiffness potentiating lung volume loss in slowly progressive disease. A single session of LVR can improve Crs. A randomised controlled trial of regular LVR is needed to assess longer-term effects.
-
ItemObjective measurement of lung volume recruitment therapy: laboratory and clinical validation(BMJ PUBLISHING GROUP, 2021)Lung volume recruitment manoeuvres are often prescribed to maintain respiratory health in neuromuscular disease. Unfortunately, no current system accurately records delivered dose. This study determined the performance characteristics of a novel, objective, manual lung volume recruitment bag counter ('the counter') with bench and healthy volunteer testing, as well as in individuals with neuromuscular disease. We undertook (1) bench test determination of activation threshold, (2) bench and healthy volunteer fidelity testing during simulated patient interface leak and different pressure compressions and (3) comparisons with self-report in individuals with neuromuscular disease. The data are reported as summary statistics, compression counts, percentage of recorded versus delivered compressions and concordance (Cohen's kappa (K) and absolute agreement). RESULTS: Minimum counter activation pressure under conditions of zero leak was 1.9±0.4 cm H2O. No difference was observed between the number of repetitions delivered and recorded during high airway pressure condition. Interface leak approximating 25% resulted in underestimation of repetition counts, and once the leak was at 50% or beyond, the counter recorded no activity. Faster sampling frequency collected data with more fidelity. Counter data agreed with diary self-report during community trials (16 participants, 960 participant days, 77% agreement, Cohen's Κ=0.66 and p<0.001). Disagreement typically favoured more diary reported (18%) than counter (5%) sessions. CONCLUSIONS: The performance characteristics of a new lung volume recruitment counter have been established in both laboratory and community settings. Objective usage and dosage data should accelerate new knowledge development and better translation of lung volume recruitment therapy into policy and practice.
-
ItemTypical within and between person variability in non-invasive ventilator derived variables among clinically stable, long-term users(BMJ PUBLISHING GROUP, 2021)BACKGROUND: Despite increasing capacity to remotely monitor non-invasive ventilation (NIV), how remote data varies from day to day and person to person is poorly described. METHODS: Single-centre, 2-month, prospective study of clinically stable adults on long-term NIV which aimed to document NIV-device variability. Participants were switched to a ventilator with tele-monitoring capabilities. Ventilation settings and masking were not altered. Raw, extensible markup language data files were provided directly from Philips Respironics (EncoreAnywhere). A nested analysis of variance was conducted on each ventilator variable to apportion the relative variation between and within participants. RESULTS: Twenty-nine people were recruited (four withdrew, one had insufficient data for analyses; 1364 days of data). Mean age was 54.0 years (SD 18.4), 58.3% male with body mass index of 37.0 kg/m2 (13.7). Mean adherence was 8.53 (2.23) hours/day and all participants had adherence >4 hours/day. Variance in ventilator-derived indices was predominantly driven by differences between participants; usage (61% between vs 39% within), Apnoea-Hypopnoea Index (71% vs 29%), unintentional (64% vs 36%) and total leak (83% vs 17%), tidal volume (93% vs 7%), minute ventilation (92% vs 8%), respiratory rate (92% vs 8%) and percentage of triggered breaths (93% vs 7%). INTERPRETATION: In this clinically stable cohort, all device-derived indices were more varied between users than the day-to-day variation within individuals. We speculate that normative ranges and thresholds for clinical intervention need to be individualised, and further research is necessary to determine the clinically important relationships between clinician targets for therapy and patient-reported outcomes.