Exploring Semi-supervised Learning for Audio-based COVID-19 Detection using FixMatch

While there has been recent success in audio-based COVID-19 detection, challenges still exist in developing more reliable and generalised models due to the limited amount of high quality labelled audio recordings. With a substantial amount of unlabelled audio recordings available, exploring semi-supervised learning (SSL) may benefit COVID-19 detection by incorporating this extra data. In this paper, we propose a SSL framework which adjusted FixMatch, one of the most advanced SSL approaches, to audio signals and explored its effectiveness in COVID-19 detection. The proposed framework is validated with a crowd-sourced audio database collected from our app, and showed superior performance over supervised models with a maximum of 7.2% relative improvement. Furthermore, we demonstrated that the proposed framework significantly benefits model development using imbalanced datasets, which is a common challenge in clinical data. It can also improve model generalisation. This potentially paves a new pathway of utilising unlabelled data effectively to build more accurate and reliable COVID-19 detection tools. Copyright © 2022 ISCA.

Positive role of continuous positive Airway pressure for intensive care unit Patients with severe hypoxaemic respiratory Failure due to covid-19 pneumonia: A single Centre experience

Objectives Continuous positive airway pressure (CPAP) may be a useful treatment strategy for patients with severe COVID-19 pneumonia but its effectiveness in preventing mechanical ventilation is unknown. We aimed to evaluate the outcomes of COVID-19 patients treated with CPAP and determine predictors of CPAP response. Methods This was a retrospective observational cohort study which took place in the intensive care unit at Royal Papworth Hospital (RPH) in Cambridge, UK. We included all consecutive patients with confirmed COVID-19 pneumonia who were transferred from neighbouring hospitals between 14th March and 6th May, 2020 for consideration of ventilatory support. We instituted the use of CPAP for all patients who arrived in RPH not intubated and were not making satisfactory progress on supplemental oxygen alone. Results Of 33 self-ventilating patients included in this study, 22 (66.7%) were male and the mean age was 54±13. 23 patients received CPAP. They were more hypoxaemic than those treated with oxygen alone (PaO2/FiO2 ratio;84.3±19.0 vs 170.0±46.0 mmHg, p= 0.001). There was a significant improvement in PaO2/FiO2 ratio 1-2 hours after CPAP initiation (167.4±49.0 from 84.3±19.0 mmHg, p= 0.001) with no indication that CPAP augmented minute ventilation (pH actually fell from 7.48±0.04 to 7.45±0.04, p=0.000 and PaCO2 increased from 4.55±0.78 mmHg to 4.88±0.83 mmHg, p=0.001). 14 (61%) patients responded to CPAP and 9 required intubation. There was no difference between these two groups in terms of the severity of baseline hypoxaemia but CPAP responders had significantly lower C-reactive protein, interleukin-6 and D-dimer (see table 1). CT pulmonary angiogram was performed in 6 out of 9 intubated patients and demonstrated pulmonary emboli in 5 of them. All patients were discharged from ICU and there were no fatalities. Conclusions In this cohort, CPAP was an effective treatment modality to improve hypoxaemia and prevent invasive ventilation in a substantial proportion of patients with severe respiratory failure. Accepting the small sample size, we also found that raised biomarkers of inflammation (CRP and IL-6) and coagulopathy (D-Dimer) to be more useful predictors of CPAP responsiveness than the severity of hypoxaemia, and could help to guide intubation decisions in this clinical setting.

P54 Positive role of continuous positive airway pressure for intensive care unit patients with severe hypoxaemic respiratory failure due to COVID-19 pneumonia: a single centre experience

P54 Table 1Comparison of respiratory parameters and laboratory biomarkers between continuous positive airway pressure responders and non-respondersCPAP responders (n= 14)CPAP non-responders (n= 9)P valueAge, mean±SD, years54±1254±180.89PaO2/FiO2 ratio prior to CPAP therapy, mean±SD, mmHg 84.5±16.0 83.9±23.00.94PaO2/FiO2 ratio change on CPAP therapy, mean±SD, mmHg +83.7±43.0 +82.4±40 0.95pH prior to CPAP therapy, mean±SD 7.47±0.037.49±0.040.39pH change on CPAP, mean±SD-0.02±0.02-0.04±0.030.11PaCO2 prior to CPAP therapy, mean±SD, mmHg 4.6±0.664.4±0.980.44PaCO2 change on CPAP, mean±SD, mmHg +0.23±0.4+0.51±0.420.13*Tidal Volume on CPAP, mean±SD, ml 475±179498±1860.80RR before CPAP therapy, mean±SD, minute-1 28±9 29±4 0.8 RR change on CPAP, mean±SD, minute-1 +1.6±7.0+0.9±9.10.84CRP, mean±SD, mg/L176±83274±630.01IL-6, median±IQR, pg/mL30±47139±1480.04D-dimer, median±IQR, ng/mL321±267 941±19900.001High sensitivity troponin, median±IQR, ng/L 11.0±4.29.7±34.00.57N/L ratio, median±IQR7.9±10.08.8±8.90.55Serum ferritin, mean±SD, ug/L1407±10791396±10560.9Abbreviations: CPAP= continuous positive airway pressure, SD= standard deviation, PaO2/FiO2 ratio= ratio of arterial oxygen partial pressure to fractional inspired oxygen, RR=respiratory rate, CRP= C-reactive protein, IL-6= interleukin-6, IQR= interquartile range, N/L= neutrophil/lymphocyte.*Tidal Volume was recorded following CPAP initiation in 13 CPAP responders and 6 non-responders.