pDenoiser: A Personalized Speech Enhancement Neural Network for Pre-hospital Emergency Medical Services.

Pre-hospital emergency medical service (EMS) tasks often come with complex and diverse noise interferences, posing challenges in implementing ASR-based medical technologies and hindering efficient and accurate telephonic communication. Among the different types of noise distortion, interfering speech is especially annoying. To address these issues, our aim is to develop a technology capable of extracting the intended speech content of the target physician from noisy and mixed audio during EMS tasks. In this work, we propose a monoaural personalized speech enhancement (PSE) method called pDenoiser, which is a real-time neural network that operates in the time domain. By leveraging the prior vocalization cues of emergency physicians, pDenoiser selectively enhances target speech components while suppressing noise and nontarget speech components, thereby improving speech quality and speech recognition accuracy under noisy conditions. We demonstrate the potential value of our approach through evaluations on both public general-domain test sets and our self-collected real-world EMS test sets. The experimental results are promising, as our model effectively promotes both speech quality and ASR performance under various conditions and outperforms related methods across multiple evaluation metrics. Our methodology will hopefully elevate EMS efficiency and fortify security against nontarget speech during EMS tasks.

Towards an Environmentally Robust Speech Assistant System for Emergency Medical Services.

Automated speech recognition technology with robust performance in various environments is highly needed by emergency clinicians, but there are few successful cases. One main challenge is the wide variety of environmental interference involved during a typical prehospital care emergency service such as background noises and overlapping speech. To solve this problem, we try to establish an environmentally robust speech assistant system with the help of the proposed personalized speech enhancement (PSE) method, which utilizes the target physician's voiceprint feature to suppress non-target signal components. We demonstrate its potential value using both general public test set and our real EMS test set by evaluating the objective speech quality metrics, DNSMOS, and the recognition accuracy. Hopefully, the proposed method will raise EMS efficiency and security against non-target speech.

A retrospective study on epidemiological analysis of pre-hospital emergency care in Hangzhou, China.

Out-of-hospital cardiac arrest (OHCA) is a leading cause of global mortality, with numerous factors influencing the patient survival rate and prognosis. This study aimed to evaluate the OHCA epidemiology in China and elaborate on the current Hangzhou emergency system status. This retrospective analysis was based on the medical history system of the Hangzhou Emergency Center registered from 2015-2021. We provided a detailed description of OHCA characteristics and investigated the factors affecting the success rate of emergency treatment in terms of epidemiology, causes of onset, bystander rescue, and outcome factors. We included 9585 out-of-hospital cardiac arrest cases, of which 5442 (56.8%) had evidence of resuscitation. Patients with underlying diseases constituted the vast majority (80.1%); trauma and physicochemical factors accounted for 16.5% and 3.4%, respectively. Only 30.4% of patients (about 80.0% of bystanders witnessed) received bystander first aid. The outcome rate of emergency doctors dispatched by emergency centres was significantly higher than doctors dispatched by hospitals. Additionally, physician's first-aid experience, emergency response time, emergency telephone availability, initial heart rhythm, out-of-hospital defibrillation, out-of-hospital intubation, and using of epinephrine significantly can significantly improve the out-of-hospital return of spontaneous circulation in patients. All steps in pre-hospital care are important for patients, especially for bystander first aid and physician's first-aid experience. The popularity of first-aid training and the public emergency medical system are not potent enough. We should take those key factors into consideration when developing a pre-hospital care system for OHCA.

Multilayer perceptron-based prediction of stroke mimics in prehospital triage.

The identification of stroke mimics (SMs) in patients with stroke could lead to delayed diagnosis and waste of medical resources. Multilayer perceptron (MLP) was proved to be an accurate tool for clinical applications. However, MLP haven't been applied in patients with suspected stroke onset within 24 h. Here, we aimed to develop a MLP model to predict SM in patients. We retrospectively reviewed the data of patients with a prehospital diagnosis of suspected stroke between July 2017 and June 2021. SMs were confirmed during hospitalization. We included demographic information, clinical manifestations, medical history, and systolic and diastolic pressure on admission. First, the cohort was randomly divided into a training set (70%) and an external testing set (30%). Then, the least absolute shrinkage and selection operator (LASSO) method was used in feature selection and an MLP model was trained based on the selected items. Then, we evaluated the performance of the model using the ten-fold cross validation method. Finally, we used the external testing set to compare the MLP model with FABS scoring system (FABS) and TeleStroke Mimic Score (TM-Score) using a receiver operator characteristic (ROC) curve. In total, 402 patients were included. Of these, 82 (20.5%) were classified as SMs. During the ten-fold cross validation, the mean area under the ROC curve (AUC) of 10 training sets and 10 validation sets were 0.92 and 0.87, respectively. In the external testing set, the AUC of the MLP model was significantly higher than that of the FABS (0.855 vs. 0.715, P = 0.038) and TM-Score (0.855 vs. 0.646, P = 0.006). The MLP model had significantly better performance in predicting SMs than FABS and TM-Score.

Design and validation of a new scale for prehospital evaluation of stroke and large vessel occlusion.

Background: Rapid recognition of acute stroke and large vessel occlusion (LVO) is essential in prehospital triage for timely reperfusion treatment. Objective: This study aimed to develop and validate a new screening tool for both stroke and LVO in an urban Chinese population. Methods: This study included patients with suspected stroke who were transferred to our hospital by emergency medical services between July 2017 and June 2021. The population was randomly partitioned into training (70%) and validation (30%) groups. The Staring-Hypertension-atrIal fibrillation-sPeech-weakneSs (SHIPS) scale, consisting of both clinical and medical history information, was generated based on multivariate logistic models. The predictive ability of the SHIPS scale was evaluated and compared with other scales using receiver operating characteristic (ROC) curve comparison analysis. Results: A total of 400 patients were included in this analysis. In the training group (n = 280), the SHIPS scale showed a sensitivity of 90.4% and specificity of 60.8% in predicting stroke and a sensitivity of 75% and specificity of 61.5% in predicting LVO. In the validation group (n = 120), the SHIPS scale was not inferior to Stroke 1-2-0 (p = 0.301) in predicting stroke and was significantly better than the Cincinnati Stroke Triage Assessment Tool (C-STAT; formerly CPSSS) and the Prehospital Acute Stroke Severity scale (PASS) (all p < 0.05) in predicting LVO. In addition, including medical history in the scale was significantly better than using symptoms alone in detecting stroke (training group, 0.853 versus 0.818; validation group, 0.814 versus 0.764) and LVO (training group, 0.748 versus 0.722; validation group, 0.825 versus 0.778). Conclusion: The SHIPS scale may serve as a superior screening tool for stroke and LVO identification in prehospital triage. Including medical history in the SHIPS scale improves the predictive value compared with clinical symptoms alone.

Prediction of large vessel occlusion for ischaemic stroke by using the machine learning model random forests.

BACKGROUNDS: The timely identification of large vessel occlusion (LVO) in the prehospital stage is extremely important given the disease morbidity and narrow time window for intervention. The current evaluation strategies still remain challenging. The goal of this study was to develop a machine learning (ML) model to predict LVO using prehospital accessible data. METHODS: Consecutive acute ischaemic stroke patients who underwent CT or MR angiography and received reperfusion therapy within 8 hours from symptom onset in the Computer-based Online Database of Acute Stroke Patients for Stroke Management Quality Evaluation-II dataset from January 2016 to August 2021 were included. We developed eight ML models to integrate National Institutes of Health Stroke Scale (NIHSS) items with demographics, medical history and vascular risk factors to identify LVO and validate its efficiency. RESULTS: Finally, 15 365 patients were included in the training set and 4215 patients were included in the test set. On the test set, random forests (RF), gradient boosting machine and extreme gradient boosting presented area under the curve (AUC) of 0.831 (95% CI 0.819 to 0.843), which were higher than other models, and RF presented the highest specificity (0.827). In addition, the AUC of RF was higher than other scales, and the accuracy of the model was improved by 6.4% compared with NIHSS. We also found the top five items of identifying LVO were total NIHSS score, gaze deviation, level of consciousness (LOC), LOC commands and motor left leg. CONCLUSIONS: Our proposed model could be a useful screening tool to predict LVO based on the prehospital accessible medical data. TRIAL REGISTRATION NUMBER: NCT04487340.

Automated pupillometry helps monitor the efficacy of cardiopulmonary resuscitation and predict return of spontaneous circulation.

BACKGROUND: We investigated the effectiveness of automated pupillometry on monitoring cardiopulmonary resuscitation (CPR) and predicting return of spontaneous circulation (ROSC) in a swine model of cardiac arrest (CA). METHODS: Sixteen male domestic pigs were included. Traditional indices including coronary perfusion pressure (CPP), end-tidal carbon dioxide (ETCO2), regional cerebral tissue oxygen saturation (rSO2) and carotid blood flow (CBF) were continuously monitored throughout the experiment. In addition, the pupillary parameters including the initial pupil size before constriction (Init, maximum diameter), the end pupil size at peak constriction (End, minimum diameter), and percentage of change (%PLR) were measured by an automated quantitative pupillometer at baseline, at 1, 4, 7 min during CA, and at 1, 4, 7 min during CPR. RESULTS: ROSC was achieved in 11/16 animals. The levels of CPP, ETCO2, rSO2 and CBF were significantly greater during CPR in resuscitated animals than those non-resuscitated ones. Init and End were decreased and %PLR was increased during CPR in resuscitated animals when compared with those non-resuscitated ones. There were moderate to good significant correlations between traditional indices and Init, End, and %PLR (|r| = 0.46-0.78, all P < 0.001). Furthermore, comparable performance was also achieved by automated pupillometry (AUCs of Init, End and %PLR were 0.821, 0.873 and 0.821, respectively, all P < 0.05) compared with the traditional indices (AUCs = 0.809-0.946). CONCLUSION: The automated pupillometry may serve as an effective surrogate method to monitor cardiopulmonary resuscitation efficacy and predict ROSC in a swine model of cardiac arrest.

Fast hypothermia induced by extracorporeal circuit cooling alleviates renal and intestinal injury after cardiac arrest in swine.

BACKGROUND: Continuous renal replacement therapy (CRRT) was currently demonstrated to be an effective way to induce fast hypothermia and had proective effects on cardiac dysfunction and brain damage after cardiac pulmonary resuscitation (CPR). In the present study, we aimed to investigate the influence of extracorporeal circuit cooling using CRRT on renal and intestinal damage after CPR based on a porcine model. METHODS: 32 pigs were subjected to ventricular fibrillation for 8 min, followed by CPR for 5 min before defibrillation. All were randomized to receive extracorporeal circuit cooling using CRRT (CRRT, n = 9), surface cooling (SC, n = 9), normothermia (NT, n = 9) or sham control (n = 5) at 5 min post resuscitation. Pigs in the CRRT group were cooled by 8-h CRRT cooling with the infusion line initially submerged in 4 °C of ice water and 16-h SC, while in the SC group by a 24-h SC. Temperatures were maintained at a normal range in the other two groups. Biomarkers in serum were measured at baseline and 1, 3, 6, 12, 24 and 30 h post resuscitation to assess organ functions. Additionally, tissues of kidney and intestine were harvested, from which the degree of tissue inflammation, oxidative stress, and apoptosis levels were analyzed. RESULTS: The blood temperature decreased faster by extracorporeal circuit cooling using CRRT than SC (9.8 ± 1.6 vs. 1.5 ± 0.4 °C/h, P < 0.01). Post-resuscitation renal and intestinal injury were significantly improved in the 2 hypothermic groups compared to the NT group. And the improvement was significantly greater in animals received extracorporeal circuit cooling than those received surface cooling, from both the results of biomarkers in serum and pathological evidence. CONCLUSION: Fast hypothermia induced by extracorporeal circuit cooling was superior to. surface cooling in mitigating renal and intestinal injury post resuscitation.

Prehospital Notification Procedure Improves Stroke Outcome by Shortening Onset to Needle Time in Chinese Urban Area.

Intravenous thrombolysis (IVT) with recombinant tissue plasminogen activator (rt-PA) can improve clinical outcome in eligible patients with acute ischemic stroke (AIS). However, its efficacy is strongly time-dependent. This study was aimed to examine whether prehospital notification by emergency medical service (EMS) providers could reduce onset to needle time (ONT) and improve neurological outcome in AIS patients who received IVT. We prospectively collected the consecutive clinical and time data of AIS patients who received IVT during one year after the initiation of prehospital notification procedure (PNP). Patients were divided into three groups, including patients that transferred by EMS with and without PNP and other means of transportation (non-EMS). We then compared the effect of EMS with PNP and EMS use only on ONT, and the subsequent neurological outcome. Good outcome was defined as modified Rankin Scale score of 0-2 at 3-months. In 182 patients included in this study, 77 (42.3%) patients were transferred by EMS, of whom 41 (53.2%) patients entered PNP. Compared with non-EMS group, EMS without PNP group greatly shortened the onset to door time (ODT), but EMS with PNP group showed both a significantly shorter DNT (41.3 ± 10.7 min vs 51.9±23.8 min, t=2.583, p=0.012) and ODT (133.2 ± 90.2 min vs 174.8 ± 105.1 min, t=2.228, p=0.027) than non-EMS group. Multivariate analysis showed that the use of EMS with PNP (OR=2.613, p=0.036), but not EMS (OR=1.865, p=0.103), was independently associated with good outcome after adjusting for age and baseline NIHSS score. When adding ONT into the regression model, ONT (OR=0.994, p=0.001), but not EMS with PNP (OR=1.785, p=0.236), was independently associated with good outcome. EMS with PNP, rather than EMS only, improved stroke outcome by shortening ONT. PNP could be a feasible strategy for better stroke care in Chinese urban area.