Artificial intelligence and machine learning are often described as transformative forces in healthcare, capable of enhancing patient experiences, streamlining processes, and unlocking insights that were previously hidden in complex data.
When the NIHR’s Cochlear Research Team approached DSP, they had a clear challenge: use AI to assist with the fine-tuning of cochlear implants, thereby reducing the number of in-person appointments, increasing the capacity of the health service to facilitate cochlear implant tuning procedures, and ultimately improving the quality of life for patients. Using Oracle’s Data Science services, DSP set out to uncover new ways of interpreting patient data and predict optimal implant settings with greater efficiency.
The challenge
Cochlear implant patients typically attend face-to-face appointments every three months for the first two years after surgery. At each session, clinicians adjust the settings on 22 electrodes in the implant to achieve the optimum balance of comfort and hearing clarity.
This is a highly skilled task. The correct configuration depends on multiple, interrelated factors that are difficult to interpret without years of specialist training. The growing number of patients eligible for cochlear implants has put additional pressure on NHS resources. Because the work can only be carried out by experienced audiologists, demand has outstripped supply, leading to longer waits (a wider problem across the NHS), and for some patients, extended periods of discomfort or reduced hearing.
The NIHR, working with Nottingham Biomedical Research Centre wanted to see if machine learning could assist. Could it help predict the right settings in advance? Could it lower the skill threshold needed to administer tuning adjustments? And could it reduce the number of follow-up appointments while maintaining (or even improving) patient outcomes?
The project
The work began as a proof of concept, using a simulated, representative patient dataset. This allowed DSP to test accuracy and viability while protecting patient confidentiality.
The first step was to analyse how cochlear implant settings change over time. Findings were:
- Settings for the left ear tend to be strongly correlated, while the right ear is more complex.
- Comfort values (maximum volume levels) soon after surgery vary widely between patients but converge over time.
- Threshold values (minimum volume levels) follow a similar pattern.
The project goal was to assess if machine learning could anticipate these changes, enabling the system to suggest optimal settings before the patient arrived at the clinic and configurations by clinicians with a lower level of experience were possible.
Using Oracle Data Science Service, part of Oracle Cloud Infrastructure and built on Python, DSP’s team developed predictive models tailored to the dataset. We evaluated and cross-referenced multiple model types available in OCI, carefully avoiding over-fitting to ensure results would be robust in real-world use.
The results
From the initial dataset, the models predicted configuration settings with around 90% accuracy. Based on this performance, a 95% accuracy when applying the full dataset was deemed a realistic outcome.
With that level of precision, the NIHR Cochlear Team could:
- Reduce the number of in-person appointments required per patient.
- Lower the barrier to entry for clinicians performing adjustments, helping address staffing bottlenecks.
- Shorten patient wait times.
- Improve comfort and hearing quality sooner.
The impact extends beyond immediate patient care — these insights will also inform and accelerate future NIHR research into cochlear implant optimisation.
Wider potential
While this project focuses on a specific medical application, its technical principles are widely transferable. Any organisation working with large, complex datasets can use Oracle Data Science Service to extract actionable insights and make better-informed decisions.
DSP can deliver these solutions end-to-end or provide expert advisory support to help organisations develop their own capabilities. The combination of our data science expertise with Oracle’s powerful cloud infrastructure enables organisations to turn complex challenges into practical, high-impact solutions.
DSP is the UK’s primary provider of Oracle Cloud services and the sole provider in Ireland for the OCRE 2024 Framework. We’re proud to share our collaboration with the UK’s National Institute for Health Research (NIHR). On this project, we united our respective expertise and applied advanced machine learning, hoping to improve comfort and hearing quality for cochlear implant users and achieve efficiency gains for the UK National Health Service (NHS).
For more information, please email DSP’s Higher Education and Research sector specialist, Seb Nash. Or visit www.dsp.co.uk/higher-education
Read the full online magazine here
