Training Clinician Communication Skills Using Generative AI-Driven Virtual Patients
A generative AI platform that lets clinicians interact with lifelike virtual patients to build communication skills through realistic, real-time conversations.
A generative AI platform that lets clinicians interact with lifelike virtual patients to build communication skills through realistic, real-time conversations.
In this work, we propose a machine learning framework for hearing aid development based on a physiological model of the inner ear (CARFAC) and neural network.
We used a computational modeling framework to simulate auditory evoked brain responses to transient and periodic stimuli at a range of intensities in healthy ears and in ears with different profiles of cochlear impairment, based on audiograms, human aging studies and hypothetical cases of isolated cellular damage.
An AI-based acoustic scene recognition system was developed to classify environmental sounds in real time and enable adaptive, context-aware hearing aid adjustments for individuals with hearing loss.
HASPI tracks directionality, spatial noise and reverberation effects on hearing aid output intelligibility at the group level, yet supra-threshold processing—not electro-acoustic SNR gain—best explains individuals’ benefits.
Taking advantage of information theory, this study investigated the functional connectivity in the auditory cortex and how its temporal dynamics is affected by stimulus inputs.
This study identifies the urban noise types that trigger auditory hypersensitivity in children and adolescents with autism spectrum disorder, highlighting noise pollution as a factor of social exclusion and distress.
Data-driven clustering of 70,557 audiograms reveals distinct audiometric configurations that better reflect variability in speech perception, especially in noise, than pure-tone averages, advancing precision audiology and individualized rehabilitation strategies.
The effect of masker type and cochlear implant speech processing strategy on speech intelligibility is investigated with typical-hearing listeners using vocoding as a simulation of cochlear implant processing.
Examining the neural mechanisms underlying misophonia and hyperacusis to understand the sensory and emotional processing differences.
An in depth qualitative study and patient and public engagement exercise to establish stake-holder priorities for the development of earable devices to support balance and falls assessments.
Individuals with T2DM show reduced VEMP response prevalence and amplitude, indicating potential vestibulocochlear nerve involvement despite preserved auditory neural processing.
This study shows that Transcranial Magnetic Stimulation, a neuromodulation technique, can alter brain-to-ear feedback by influencing efferent activity, highlighting its potential for investigating auditory processing and informing treatments like tinnitus therapy.
In the Clarity prediction challenges, we are encouraging the development of novel approaches to the intelligibility prediction of speech processed by hearing aids.
This project assessed a novel method of covertly measuring adaptive directionality in hearing aids using simultaneous, inaudible, direction-dependent intensity modulations applied to diffuse background noise.
This study demonstrates the real-world feasibility of using digital technologies to enable community healthcare workers in low- and middle-income settings to provide ear assessments, hearing tests, and telehealth-supported hearing care.
Bayesian Framework Allows Trading between Time and Accuracy for 2AFC Tasks such as PTA.
This study reveals that normal hearing individuals present with different profiles on vertical sound source localization wherein discrimination remains robust, identification abilities—especially in up-down and front-back localization, underscoring the need for comprehensive virtual spatial hearing assessments.
ChatGPT can support hearing self-monitoring by providing expert-level recommendations based on mobile app data.
ChatGPT can support hearing self-monitoring by providing expert-level recommendations based on mobile app data.
AICAS facilitates pure tone audiometry procedure in audiological practical training based on virtual reality and artificial intelligence technologies.
Reference values were established for ABR-BIC in adults with normal hearing, including latency and amplitude of the Binaural Interaction Component (BIC). The measurements of the latencies and amplitudes of the ABR-BI showed similarities between the right and left ears of normal listeners.
Monitoring cortical responses via fNIRS in a child with bimodal hearing transitioning to bilateral cochlear implants revealed significant adaptation to the second implant and highlighted the potential of this technology to monitor auditory brain development in this population, including the importance of measuring rehabilitation outcomes.
In this study, we describe our tuning of an inner ear model (CARFAC) that incorporates three different classes of auditory nerve fiber. We propose its use as the basis for novel, personalized models of individuals' hearing losses.
The extended pressure tympanometry (extended to -600 daPa pressure) in Children with Middle Ear Pathology were investigated and it is seen that extended tympanometry provided more sensitive results.
Machine learning models trained on simulated auditory nerve representations demonstrate the limits of attentional selection of speech in cocktail party scenario when heard through a cochlear implant.
Improving the Epley Manoeuvre - an anatomical study of the semicircular canals that guided development of an earable and app for the particle repositioning manoeuvre in BPPV.
Improving the Epley Manoeuvre - an anatomical study of the semicircular canals that guided development of an earable and app for the particle repositioning manoeuvre in BPPV.
This study used functional near-infrared spectroscopy to show that a deep neural network-based noise management program in hearing aids reduced listening effort and brain activation in the left prefrontal cortex over a traditional quiet-listening program while in a lab-controlled noisy environment.
Study explored differential advantages of signal-to-noise ratio (SNR), reverberation time, and distractor location for speech perception in noise in school-age children.
A computational model of the stimulated peripheral auditory system shows that the focused partial‑tripolar versus monopolar threshold difference could be a possible indicator of neural health status. Additionally, the model incorporates a novel intracranial auditory prosthesis, revealing that it elicits lower stimulation thresholds and distinct activation patterns compared to conventional cochlear implants.
This study explored how hearing-impaired listeners define lyrics intelligibility in music, providing a working definition for evaluating improvements and developing new signal processing strategies.
This study demonstrates that televideofeedback, combining asynchronous video and synchronous guidance, significantly improves parent-child communication and parental self-efficacy in families of children with hearing loss.
In this study, we investigated early auditory processing at the fundamental frequency of natural speech in terms of audiovisual integration and whether these processes are associated with behavioural measures of speech comprehension.
Pupil-linked arousal modulates auditory perceptual belief updating during implicit online perceptual decision-making generalized across the spatial and temporal domain, similar to explicit belief updating processes known from statistical learning.
Central auditory processing disorder is prevalent in school-going children with academic difficulties
This project explores the use of Automatic Speech Recognition (ASR) models to transcribe and score open-set speech audiometry responses from cochlear implant users, demonstrating that ASR models can approximate benchmark scores while significantly reducing transcription time.
Intermediate data generated in pure-tone audiometry, by our patient response simulator (Hughson-Westlake protocol), help to understand the probabilistic and continuous representation of a patient's hearing ability derived from Gaussian process inference used in computational audiometry.
We developed and validated of an easy, language-agnostic speech test using audiovisual VCV stimuli and AI-driven thresholding, demonstrating its potential for equitable, efficient, and personalized hearing assessment across diverse populations.
This study evaluated six AI chatbots on their accuracy and consistency in addressing unproven audiological methods, finding generally high performance.