Project Outline
Obstructive sleep apnoea (OSA) is associated with major health conditions including cardiovascular disease and dementia, making earlier and more accurate diagnosis important. However, OSA is a heterogeneous condition: people with similar conventional measures can show different patterns of airway obstruction, oxygen desaturation and physiological response. This is particularly relevant in women, where OSA may present differently and can be under-recognised
Every snore is shaped by airflow, the upper airway and the sleeping body. This PhD will use acoustic analysis and AI to investigate whether snore sounds can support more personalised detection and characterisation of OSA, with a particular focus on women’s sleep health. Working with PFL Healthcare and clinical collaborators at Hywel Dda University Health Board, you will analyse real-world sleep audio alongside physiological signals collected across multiple nights.
The longer-term goal is to develop more personalised, unobtrusive and scalable approaches to support earlier and more accurate diagnosis of OSA.
Further project information
OSA is strongly associated with a range of conditions, including cardiovascular disease and dementia, but it is also highly heterogeneous: people with similar conventional measures can show different patterns of airway obstruction, oxygen desaturation and physiological response. These differences may be particularly important in women, who may more commonly show features such as airflow limitation and lower arousal thresholds. Gaps in current diagnostic pathways may therefore contribute to delayed or missed diagnosis.
This project asks: Can personalised acoustic AI capture individual- and sex-specific patterns of snoring to improve the detection and characterisation of OSA?
What will you do?
You will develop new methods to identify and model acoustic ‘phenotypes’ of snoring – recurring sound patterns that may reflect different forms of airway behaviour and physiological response. The project will combine digital signal processing, machine learning and multimodal time-series analysis to:
identify acoustic characteristics and phenotypes associated with different patterns of sleep-disordered breathing; investigate whether snore patterns differ between women and men; link acoustic patterns to oxygen desaturation, respiratory effort and cardiovascular signals; and develop personalised models that capture individual variability and within-person changes over time.
A major strength of the project is access to existing paired acoustic and physiological sleep datasets, allowing you to begin substantive research early in the PhD. Repeated home recordings will also provide an opportunity to investigate night-to-night variability and longitudinal acoustic biomarkers.
You will have scope to explore modern audio-AI approaches, including time-frequency representations, learned audio embeddings, deep sequence models and multimodal representation learning. The precise methods will evolve with the research questions and your interests, while maintaining afocus on robust, interpretable and clinically meaningful modelling.
Industry, clinical and translational collaboration
This project brings together academic, industry, clinical and translational expertise. PFL Healthcare Ltd, a UK digital-health company developing technologies for home sleep monitoring, will provide access to real-world acoustic and physiological datasets, technical expertise and an industrial placement. Building on an existing Sheffield–PFL collaboration involving SoundSleep technology, the placement will provide experience of data acquisition and quality assurance, product development, validation and the translation of acoustic AI into deployable health technology.
A specialised Sleep and Respiratory Physiologist at Hywel Dda University Health Board will contribute clinical expertise in sleep-disordered breathing and women’s sleep health, including the evaluation of new approaches to diagnosing OSA in women.
The NIHR HealthTech Research Centre in Long Term Conditions will support the project through its Women’s Health theme. This will include expertise in identifying unmet clinical needs, advising on patient and public involvement, and supporting translation of emerging Health Tech into clinical practice. The Centre can also help engage women with lived experience of sleep-disordered breathing to inform research priorities, support consideration of sex-specific diagnostic inequalities, and provide guidance on clinical validation, regulatory requirements and potential routes to NHS adoption.
Together, these partnerships will help ensure that the research is shaped by clinically relevant questions, patient priorities and real-world implementation needs, with a clear pathway from acoustic-AI methodology to evaluation and translation.
Subject Areas
- Women’s Health;
- Speech and audio technology;
- Acoustics;
- Artificial Intelligence and Machine Learning;
- Digital Health;
- Biomedical Signal Processing;
- Sleep and Sleep-Disordered Breathing;
- Wearable and Physiological Sensing;
- Personalised Health Monitoring
Required technical skills/academic qualifications/experience
We welcome applicants from a wide range of quantitative backgrounds, including Computer Science, Artificial Intelligence, Electronic/Electrical Engineering, Acoustics, Biomedical Engineering, Mathematics, Physics, Data Science or related disciplines.
You should have good programming skills, ideally in Python, and an interest in at least one of the following: machine learning, digital signal processing, audio/acoustics or time-series analysis.
Experience with deep learning, physiological signals, wearable sensing or biomedical data would be useful but is not essential. You do not need prior experience in sleep medicine or every technique described in the project; training will be provided in acoustic analysis, machine learning, physiological signal processing and sleep health.
We value curiosity, collaboration and a willingness to learn. We particularly encourage applications from women and other groups under-represented in research, while all appointments will be made on merit.
Mode of attendance
This project is only available on a full-time basis.
Supervisors
Project Partners
This project will be hosted at the University of Sheffield with support from PFL Healthcare Ltd,
NIHR Health Tech Research Centre in Long Term Conditions and Hywel Dda University Health Board, NHS Wales
