World-first application of AI-assisted scans to help patients with rare neck condition

Real-time muscle analysis technology to improve treatment of cervical dystonia

AI-assisted scans to improve treatment of cervical dystonia

AI-assisted scans to improve treatment of cervical dystonia

Artificial intelligence-assisted ultrasound scans could significantly improve the assessment of the rare and sometimes painful disease cervical dystonia, new research shows.

Cervical dystonia is a disabling condition in which the neck muscles contract involuntarily, causing the head to twist, turn and pull into an abnormal posture.

The current solution sees doctors injecting botulinum toxin – of which Botox is a famous example – into the affected muscles to stop the involuntary spasms by blocking nerve signals.

The injections are administered after clinicians observe and assess the patient’s neck area manually, a process that can lack accuracy

But researchers at Manchester Metropolitan University have been testing a new system, using artificial intelligence (AI) to visualise and analyse the patterns of neck muscles associated with cervical dystonia to help guide doctors to the specific problem areas, with minimal training.

The researchers trained the AI system using thousands of ultrasound scan images of patients with and without cervical dystonia.

The research into the potential of the system, published in the Journal of Biomedical and Health Informatics, is a world first in muscle analysis technology.

Real-time analysis

The neck is formed of several layers of muscles and currently clinicians have to manually detect which muscles need treating based on observations of the patient’s posture, such as head position and shoulder elevation.

The new system uses a normal ultrasound probe to scan and produce images of the layers.

It then overlays in real time an augmented reality ‘map’ of identified muscles onto an ultrasound scan image. 

Doctors can then see the muscle size and shape to identify and target the precise problem area, and AI technology called neural networks, which works similarly to how the human brain learns, allows clinicians to do this in real time.

Neural networks

Lead researcher Ian Loram, Professor of Neuromuscular Control of Human Movement at Manchester Metropolitan University, said: “The benefit to using this AI technology is that we can compile the knowledge of many experts into one data-trained neural network.

“This new method is non-invasive, requires minimal training and expertise for clinicians, and will provide no burden on clinical time. 

“This new system could also have an impact outside of cervical dystonia and could have implications on other neurological diseases, which affect movement, such as Parkinson’s disease.

“When developed further it could also be beneficial for nerve injuries where the messages to the muscle are damaged, inform diagnosis of chronic pain where the muscles are altered and be used to improve sports performance.”

Dr Christopher Kobylecki, a Consultant Neurologist at Manchester Centre for Clinical Neurosciences at Salford Royal NHS Foundation Trust, who has been working alongside the researchers at Manchester Metropolitan, added: “Treatment of people with cervical dystonia with botulinum toxin does not always work fully. Studies have shown that being able to select the right muscles for injection is a key factor. Currently this is mainly done by examination of the patient, which is not always accurate.

“Ultrasound is used in botulinum toxin injections, but current methods rely on training of the person using the ultrasound machine.

“Being able to automatically identify neck muscles would potentially be of great help to clinicians treating cervical dystonia.”

Interdisciplinary

Researchers with specialisms in computer science and human movement science worked together to train a computer neural network to recognise abnormal patterns of muscles by observing thousands of ultrasound images from 35 cervical dystonia patients from Salford Royal NHS Foundation Trust in Salford, Greater Manchester, and 26 healthy, similarly aged patients.

The imagery and data collected by the system will provide detailed pictures of the neck muscles to allow clinicians to learn more about the condition and track changes over time.

Dr Ryan Cunningham, Lecturer in Data Science at Manchester Metropolitan and part of the research team, said: “There are a limited number of ways that we can extract data on what the muscles are doing deep inside the body.

“The current method of treatment essentially relies on an educated guess of clinicians and can often lead to treatment of the incorrect muscles. This can exacerbate the problems for patients.

“This new method provides an automated, objective visualisation of the neck muscles and so makes the treatment more reliable.

“It also allows us to look how the muscles develop or respond to treatment over time, which we have never been able to do before.”

The researchers will now continue to work alongside clinicians at Salford Royal NHS Foundation Trust to develop, deploy and test this system, and track the effect of the botulinum toxin injections on patients.

This project was funded by Dystonia UK.

Previous Story Dr David Lambrick joins Manchester-Wuhan City Circle Cooperation Summit