ABSTRACT
Speech inversion is a well-known ill-posed problem and addition of speaker differences typically makes it even harder. Normalizing the speaker differences is essential to effectively using multi-speaker articulatory data for training a speaker independent speech inversion system. This paper explores a vocal tract length normalization (VTLN) technique to transform the acoustic features of different speakers to a target speaker acoustic space such that speaker specific details are minimized. The speaker normalized features are then used to train a deep feed-forward neural network based speech inversion system. The acoustic features are parameterized as time-contextualized mel-frequency cepstral coefficients. The articulatory features are represented by six tract-variable (TV) trajectories, which are relatively speaker invariant compared to flesh point data. Experiments are performed with ten speakers from the University of Wisconsin X-ray microbeam database. Results show that the proposed speaker normalization approach provides an 8.15% relative improvement in correlation between actual and estimated TVs as compared to the system where speaker normalization was not performed. To determine the efficacy of the method across datasets, cross speaker evaluations were performed across speakers from the Multichannel Articulatory-TIMIT and EMA-IEEE datasets. Results prove that the VTLN approach provides improvement in performance even across datasets.
ABSTRACT
BACKGROUND AND STUDY AIMS: This report describes the use of a novel, fully covered, self-expanding metal stent (FCSEMS) for endoscopic ultrasound (EUS)-guided drainage of walled-off pancreatic necrosis (WON). PATIENTS AND METHODS: Patients with WON, as defined by the revised Atlanta Criteria, were included in this open-lable, two-center, observational study. The WON was punctured using a cystotome, and the FCSEMS was inserted under fluoroscopic guidance. Necrosectomy procedures were performed as necessary. RESULTS: A total of 19 patients were included. The median maximum collection size was 15âcm with a median of 50â% necrosis. A total of 14/19 patients underwent necrosectomy, requiring a median of 4 procedures. Resolution or reduction in the size of collection by at least 80â% was achieved in all patients. Percutaneous or surgical drainage was required in three patients. Five stents migrated or dislodged. One patient had abdominal pain post-procedure. Five patients died during follow-up (three from multi-organ failure, and two unrelated to pancreatitis). CONCLUSIONS: Use of this stent is feasible and safe for drainage of WON. However, stent displacement rates were high, and improvements to the stent design are required before it can be advocated for routine use in WON.
Subject(s)
Drainage/methods , Endoscopy, Digestive System/methods , Pancreatectomy/methods , Pancreatitis, Acute Necrotizing/surgery , Self Expandable Metallic Stents , Adolescent , Adult , Aged , Aged, 80 and over , Child , Endosonography , Equipment Design , Female , Follow-Up Studies , Humans , Magnetic Resonance Imaging , Male , Middle Aged , Pancreatitis, Acute Necrotizing/diagnosis , Retrospective Studies , Tomography, X-Ray Computed , Treatment Outcome , Young AdultABSTRACT
OBJECTIVE: To determine the yield of endoscopic ultrasound (EUS) in the investigation of patients with normal liver function tests (LFTs) and unexplained dilatation of common bile duct (CBD) and/or pancreatic duct (PD), following CT and/or magnetic resonance cholangiopancreatography. MATERIALS AND METHODS: Consecutive patients undergoing linear EUS between January 2007 and August 2011 for the indication of dilated CBD and/or PD, normal LFT, and nondiagnostic cross-sectional imaging formed the study group. The study was performed as a retrospective analysis of prospectively collected data. RESULTS: During the study period, 83 patients (CBD and PD dilatation n = 38, PD dilatation n = 5, CBD dilatation n = 40) met the inclusion criteria and underwent EUS. Five (13.1%) of the CBD and PD groups had a new finding, which in one (2.6%) case was causal. In this group, men were significantly more likely to have a new finding (p = 0.012). Eight (20%) of the CBD group had a new finding, which in seven (17.5%) cases was causal. In the CBD group, cholecystectomy was significantly (p = 0.005) more common in those without a finding. Three (60%) of the PD group had a finding on EUS, all of which were causal, including a case of pancreatic malignancy. CONCLUSION: There is a significant yield from EUS in individuals with isolated PD dilatation and isolated CBD dilatation. Previous cholecystectomy is significantly associated with a negative EUS in the group with isolated CBD dilatation. The yield in those with CBD and PD dilatation was low and a finding was more likely in males.
Subject(s)
Common Bile Duct/diagnostic imaging , Endosonography/methods , Pancreatic Ducts/diagnostic imaging , Adult , Aged , Aged, 80 and over , Cholangiopancreatography, Magnetic Resonance , Common Bile Duct/pathology , Dilatation, Pathologic , Female , Humans , Liver Function Tests , Male , Middle Aged , Pancreatic Ducts/pathology , Retrospective Studies , Tomography, X-Ray ComputedABSTRACT
Objectives: This analysis assessed current endoscopic retrograde cholangiopancreatography (ERCP) practice within the UK, including use of sedation and patient comfort. Methods: ERCPs conducted over 1 year (1 July 2021-30 June 2022) and uploaded to the National Endoscopy Database (NED) were analysed. The endoscopist workforce was classified by gender and specialty, use of sedation was analysed. Logistic regression was used to assess associations between patient age, gender and procedure indications on moderate to severe discomfort risk. Results: 27 812 ERCPs were performed by 491 endoscopists in 175 sites and uploaded to NED, an estimated 50% of total UK activity. 13% were training procedures, 94% of the endoscopists were male, with 72% being gastroenterologists. Most ERCPs were performed under conscious sedation (89%). The discomfort rate among patients aged 60-90 undergoing ERCP under conscious sedation was 4.2% (95% CI 3.9% to 4.5%), with only 5.5% (95% CI 5.2% to 5.9%) receiving greater than 5 mg midazolam or 100 µg fentanyl.Younger patients (<30 years) had a higher discomfort risk during conscious sedation ERCPs than those aged 70-79 (OR 3.0, 95% CI 2.2 to 4.3, p<0.05), while male patients had a lower discomfort risk compared with females (OR 0.9, 95% CI 0.8 to 1.0, p=0.05).Enhanced sedation (propofol or general anaesthetic) was associated with lower frequency of discomfort (0.3%, 95% CI 0.1 to 0.6) compared with conscious sedation (5.1%, 95% CI 4.9% to 5.4%, p<0.05). Conclusions: Conscious sedation is well tolerated for most patients and prescribing practices have improved. However, triage of more patients, particularly young females, to enhanced sedation lists should be considered to reduce discomfort rates in future.
ABSTRACT
Mind-body interventions such as mindfulness-based stress reduction (MBSR) may improve well-being by increasing awareness and regulation of physiological and cognitive states. However, it is unclear how practice may alter long-term, baseline physiological processes, and whether these changes reflect improved well-being. Using respiration rate (RR), which can be sensitive to effects of meditation, and 3 aspects of self-reported well-being (psychological well-being [PWB], distress, and medical symptoms), we tested pre-registered hypotheses that: (1) Lower baseline RR (in a resting, non-meditative state) would be a physiological marker associated with well-being, (2) MBSR would decrease RR, and (3) Training-related decreases in RR would be associated with improved well-being. We recruited 245 adults (age range = 18-65, M = 42.4): experienced meditators (n = 42), and meditation-naïve participants randomized to MBSR (n = 72), active control (n = 41), or waitlist control (n = 66). Data were collected at pre-randomization, post-intervention (or waiting), and long-term follow-up. Lower baseline RR was associated with lower psychological distress among long-term meditators (p* = 0.03, b = 0.02, 95% CI [0.01, 0.03]), though not in non-meditators prior to training. MBSR decreased RR compared to waitlist (p = 0.02, Cohen's d = - 0.41, 95% CI [- 0.78, - 0.06]), but not the active control. Decreased RR related to decreased medical symptoms, across all participants (p* = 0.02, b = 0.57, 95% CI [0.15, 0.98]). Post-training, lower RR was associated with higher PWB across training groups compared to waitlist (p* = 0.01, b = 0.06, 95% CI [0.02, 0.10]), though there were no significant differences in change in PWB between groups. This physiological marker may indicate higher physical and/or psychological well-being in those who engage in wellness practices.
Subject(s)
Meditation , Psychological Distress , Adult , Humans , Adolescent , Young Adult , Middle Aged , Aged , Self Report , Respiratory Rate , Physical ExaminationABSTRACT
Studies have shown that supplementary articulatory information can help to improve the recognition rate of automatic speech recognition systems. Unfortunately, articulatory information is not directly observable, necessitating its estimation from the speech signal. This study describes a system that recognizes articulatory gestures from speech, and uses the recognized gestures in a speech recognition system. Recognizing gestures for a given utterance involves recovering the set of underlying gestural activations and their associated dynamic parameters. This paper proposes a neural network architecture for recognizing articulatory gestures from speech and presents ways to incorporate articulatory gestures for a digit recognition task. The lack of natural speech database containing gestural information prompted us to use three stages of evaluation. First, the proposed gestural annotation architecture was tested on a synthetic speech dataset, which showed that the use of estimated tract-variable-time-functions improved gesture recognition performance. In the second stage, gesture-recognition models were applied to natural speech waveforms and word recognition experiments revealed that the recognized gestures can improve the noise-robustness of a word recognition system. In the final stage, a gesture-based Dynamic Bayesian Network was trained and the results indicate that incorporating gestural information can improve word recognition performance compared to acoustic-only systems.
Subject(s)
Gestures , Speech Perception/physiology , Speech Recognition Software , Speech/physiology , Bayes Theorem , Humans , Phonetics , Speech Acoustics , VocabularyABSTRACT
Speech can be represented as a constellation of constricting vocal tract actions called gestures, whose temporal patterning with respect to one another is expressed in a gestural score. Current speech datasets do not come with gestural annotation and no formal gestural annotation procedure exists at present. This paper describes an iterative analysis-by-synthesis landmark-based time-warping architecture to perform gestural annotation of natural speech. For a given utterance, the Haskins Laboratories Task Dynamics and Application (TADA) model is employed to generate a corresponding prototype gestural score. The gestural score is temporally optimized through an iterative timing-warping process such that the acoustic distance between the original and TADA-synthesized speech is minimized. This paper demonstrates that the proposed iterative approach is superior to conventional acoustically-referenced dynamic timing-warping procedures and provides reliable gestural annotation for speech datasets.
Subject(s)
Acoustics , Gestures , Glottis/physiology , Mouth/physiology , Speech Acoustics , Voice Quality , Biomechanical Phenomena , Female , Humans , Male , Models, Theoretical , Signal Processing, Computer-Assisted , Sound Spectrography , Speech Production Measurement/methods , Time FactorsABSTRACT
BACKGROUND: The COVID-19 pandemic has profoundly affected endoscopy services including pancreatobiliary (PB) endoscopy across the UK. The British Society of Gastroenterology and Joint Advisory Group have issued guidance for managing endoscopy services safely throughout this period. There have been perceived concerns among the PB endoscopists that wearing full personal protective equipment might have an adverse impact on key performance indicators (KPIs) in endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic ultrasound (EUS) procedures leading to non-compliance with the national guidelines. The aim of the study was to assess the impact of COVID-19 pandemic on KPIs in ERCP and EUS and ascertain the risk of procedure-related complications. METHODS: A retrospective audit of a prospectively maintained endoscopy database was carried out between 18 March and 31 July 2020. RESULTS: 146 ERCP procedures (common bile duct (CBD) cannulation rate of naïve papilla 89.2%, complete CBD stone extraction rate at first ERCP 88.2%, biliary stricture decompression rate 91%) and 87 EUS procedures (diagnostic accuracy of EUS-fine needle aspiration 92%) were carried out during this period. ERCP-related complications included pancreatitis (4.8%), bleeding (0.68%) and cholangitis (0.68%). 30-day ERCP procedure-related mortality was 0.68%. There were no complications or procedure-related mortality in the EUS group. CONCLUSION: This is the first study looking at the impact of COVID-19 on KPIs and procedure-related complications in ERCP and EUS in the literature. Our study confirms that a high-quality PB endoscopy service can be delivered safely and effectively during the COVID-19 pandemic.
ABSTRACT
Respiratory rate (RR) is a clinical metric used to assess overall health and physical fitness. An individual's RR can change from their baseline due to chronic illness symptoms (e.g., asthma, congestive heart failure), acute illness (e.g., breathlessness due to infection), and over the course of the day due to physical exhaustion during heightened exertion. Remote estimation of RR can offer a cost-effective method to track disease progression and cardio-respiratory fitness over time. This work investigates a model-driven approach to estimate RR from short audio segments obtained after physical exertion in healthy adults. Data was collected from 21 individuals using microphone-enabled, near-field headphones before, during, and after strenuous exercise. RR was manually annotated by counting perceived inhalations and exhalations. A multi-task Long-Short Term Memory (LSTM) network with convolutional layers was implemented to process mel-filterbank energies, estimate RR in varying background noise conditions, and predict heavy breathing, indicated by an RR of more than 25 breaths per minute. The multi-task model performs both classification and regression tasks and leverages a mixture of loss functions. It was observed that RR can be estimated with a concordance correlation coefficient (CCC) of 0.76 and a mean squared error (MSE) of 0.2, demonstrating that audio can be a viable signal for approximating RR.Clinical relevance-The subject technology facilitates the use of accessible, aesthetically acceptable wearable headphones to provide a technologically efficient and cost-effective method to estimate respiratory rate and track cardio-respiratory fitness over time.
Subject(s)
Asthma , Wearable Electronic Devices , Exercise , Humans , Respiration , Respiratory RateABSTRACT
This paper presents a neural network architecture using a support vector machine (SVM) as an inference engine (IE) for classification of light detection and ranging (Lidar) data. Lidar data gives a sequence of laser backscatter intensities obtained from laser shots generated from an airborne object at various altitudes above the earth surface. Lidar data is pre-filtered to remove high frequency noise. As the Lidar shots are taken from above the earth surface, it has some air backscatter information, which is of no importance for detecting underwater objects. Because of these, the air backscatter information is eliminated from the data and a segment of this data is subsequently selected to extract features for classification. This is then encoded using linear predictive coding (LPC) and polynomial approximation. The coefficients thus generated are used as inputs to the two branches of a parallel neural architecture. The decisions obtained from the two branches are vector multiplied and the result is fed to an SVM-based IE that presents the final inference. Two parallel neural architectures using multilayer perception (MLP) and hybrid radial basis function (HRBF) are considered in this paper. The proposed structure fits the Lidar data classification task well due to the inherent classification efficiency of neural networks and accurate decision-making capability of SVM. A Bayesian classifier and a quadratic classifier were considered for the Lidar data classification task but they failed to offer high prediction accuracy. Furthermore, a single-layered artificial neural network (ANN) classifier was also considered and it failed to offer good accuracy. The parallel ANN architecture proposed in this paper offers high prediction accuracy (98.9%) and is found to be the most suitable architecture for the proposed task of Lidar data classification.
Subject(s)
Artificial Intelligence , Image Interpretation, Computer-Assisted/methods , Imaging, Three-Dimensional/methods , Information Storage and Retrieval/methods , Lasers , Pattern Recognition, Automated/methods , Radar , Algorithms , Image Enhancement/methods , Neural Networks, Computer , Reproducibility of Results , Sensitivity and SpecificityABSTRACT
BACKGROUND AND AIMS: Our study aimed to assess the sensitivity of EUS and EUS-FNA for pancreatic neuro-endocrine tumors (pNETs) and compare performance over two consecutive 4 year 2 month periods, to investigate the comparative performance between solid and cystic pNETs and determine the incremental yield of EUS +/- FNA in individuals with a mass not diagnosed as a pNET after cross-sectional imaging. METHODS: A retrospective review of a prospectively maintained database was carried out to identify all pNET patients who underwent EUS-FNA between April 2003 and September 2011. RESULTS: A final diagnosis of solid and cystic pNETs was made in 43 and 10 patients, respectively. Overall, the yield of combined EUS imaging and cytology was significantly higher than that of CT and/or MRI (p< 0.05) across all groups [solid (83.7% vs. 41.8%), cystic (70% vs. 10%) and combined solid-cystic (81.1% vs. 35.8%)]. The yield of combined EUS imaging and cytology was significantly better than EUS imaging alone (p<0.05) in the solid (83.7% vs. 58%) and combined pNET cohort (81.1% vs. 52.8%) of patients. After a non-diagnostic CT and or MRI, EUS/EUS-FNA confirmed pNET in 19 out of 25 patients (76.0%) with solid pNETs and 6 out of 9 patients (66.7%) with cystic pNETs. CONCLUSION: EUS and EUS-FNA had a significant clinical impact in the 25/34 of cases where pNET was not suspected after initial cross-sectional imaging.
Subject(s)
Carcinoma, Neuroendocrine/diagnosis , Endoscopic Ultrasound-Guided Fine Needle Aspiration , Endosonography , Neoplasms, Cystic, Mucinous, and Serous/diagnosis , Pancreatic Cyst/diagnosis , Pancreatic Neoplasms/diagnosis , Adult , Aged , Aged, 80 and over , Carcinoma, Neuroendocrine/diagnostic imaging , Carcinoma, Neuroendocrine/pathology , Databases, Factual , Female , Humans , Magnetic Resonance Imaging , Male , Middle Aged , Neoplasms, Cystic, Mucinous, and Serous/diagnostic imaging , Neoplasms, Cystic, Mucinous, and Serous/pathology , Pancreatic Cyst/diagnostic imaging , Pancreatic Cyst/pathology , Pancreatic Neoplasms/diagnostic imaging , Pancreatic Neoplasms/pathology , Predictive Value of Tests , Prognosis , Reproducibility of Results , Retrospective Studies , Time Factors , Tomography, X-Ray ComputedABSTRACT
BACKGROUND: Endoscopic retrograde cholangio-pancreatography (ERCP) is an important tool in the management of pancreato-biliary disease. OBJECTIVE: To compare the current practice of ERCP in a district general hospital with those reported in the 2007 British Society of Gastroenterology (BSG) ERCP audit and assess access to the service. DESIGN: This was a service evaluation study. Data were collected retrospectively for all people who underwent ERCP. Demographic, clinical and procedure related data were collected and analysed. SETTING: Sunderland Hospital. RESULTS: 236 patients (median age 70 years, 56% women) underwent ERCP. The median period from referral to patient review was 1.0 day. The median period from the decision to carry out an ERCP to the actual procedure date was 3 days. All patients had radiological imaging before their first procedure. 96% patients had their bloods checked within 1 week of the procedure. The most common indication was related to choledocholithiasis and its complications. The mean doses of midazolam and diazemul used were 4.4 mg and 11.1 mg, respectively. The selective biliary cannulation rate was 92%. Sphincterotomy, biliary stent insertion and complete stone extraction were achieved in 94%, 85% and 88% of patients before the procedure. Complications that occurred as a result of ERCPs were as follows: bleeding (1.7%), pancreatitis (3.8%), cholangitis (0.4%) and renal failure (0.4%). The 30-day death rate was 4.6%. However, none of these were procedure related. CONCLUSIONS: The structure of the ERCP services at Sunderland Royal Hospital provides patients with a high-quality and accessible service. The technical success rate and sedation rate were better than those reported in the BSG ERCP audit. The complication rate and procedure-related mortality were comparable to the BSG audit and much below the published figures.
ABSTRACT
Many different studies have claimed that articulatory information can be used to improve the performance of automatic speech recognition systems. Unfortunately, such articulatory information is not readily available in typical speaker-listener situations. Consequently, such information has to be estimated from the acoustic signal in a process which is usually termed "speech-inversion." This study aims to propose and compare various machine learning strategies for speech inversion: Trajectory mixture density networks (TMDNs), feedforward artificial neural networks (FF-ANN), support vector regression (SVR), autoregressive artificial neural network (AR-ANN), and distal supervised learning (DSL). Further, using a database generated by the Haskins Laboratories speech production model, we test the claim that information regarding constrictions produced by the distinct organs of the vocal tract (vocal tract variables) is superior to flesh-point information (articulatory pellet trajectories) for the inversion process.