How neuromonitoring can prevent post-operative brain damage

Analysing signal data in intraoperative neuromonitoring using machine learning – Simon Koller won over the jury at the DMEA in Berlin with this topic for his Bachelor’s thesis. For the sixth time in eight years, a graduate of the Bachelor of Science in Medical Informatics at Bern University of Applied Sciences BFH was honoured with the International Young Talent Award for the best Bachelor’s thesis. He explains the topic in an interview.

Societybyte: You analysed signal data in intraoperative neuromonitoring using machine learning in your final thesis for the Bern University Hospital for Neurosurgery. What motivated you to write this thesis?

Simon Koller: During my training as an orthopaedic surgeon, I had to go to intensive care units immediately after patients’ brain operations to make a mass helmet to protect the brain. These were lasting experiences that left a lasting impression on me, as brain damage in some cases meant that the patient was no longer the same as before the operation. I was therefore highly motivated to carry out research to make brain surgery safer during the operation and machine learning is a subject area that interests me greatly and where I see great potential.

Simon Koller (winner of the DMEA Young Talent Award 2024 and graduate of the Bachelor of Medical Informatics BFH) in the centre with representatives of the Department of Medical Informatics at BFH Engineering and Computer Science

How did you approach your analysis in terms of study design and methodology?

In tumour operations, it is sometimes a challenge for surgeons to recognise the difference between healthy nerve tissue and a tumour that is to be removed. That’s why tools are needed to support surgeons during the operation, and Inselspital is conducting intensive research into this. I developed a machine learning model that uses muscle signals after brain stimulation during surgery to recognise subtle patterns that indicate nerve damage. The model is designed to warn of potential damage intraoperatively. It was trained using real data from patients at Inselspital Bern and tested on new patient data from past operations for which the model had not yet been trained. The whole process took one semester as part of my bachelor’s thesis.

Which patients can the procedure be used on?

Basically for all operations with intraoperative neuromonitoring, i.e. on the brain and spine. However, there are still hurdles to overcome before it can be used in practice. My focus was on developing the best possible model. The next step is to research how it should be used in practice and how more structured data can be collected.

To what extent do our brain and machine learning go together?

Put simply, surgeons stimulate the nerve tissue during the operation and transmit electrical signals from the brain via the nerves to the muscles. If the muscle moves, this indicates that the nerves are intact; if it remains still, there may be damage. The resulting complex muscle signals are recorded and fed into a neural network as training and test data. This then learns to recognise the subtle patterns that correlate with nerve damage. Fascinatingly, these artificial neural networks have been used to digitally replicate the functioning of the human brain in order to protect the biological brain.

How can your work help patients and healthcare professionals – now and in the future?

In future, the model should be able to predict with a comprehensible probability what damage could occur later during the operation. In some cases, this would allow adjustments to be made during the operation to reverse the damage or at least not make it worse. As a “by-product”, I also developed tools that could already help doctors to interpret signals and recognise errors when wiring patients.

What subsequent investigations will arise from the results for the future?

The Inselspital is working on a scientific publication on the work, for which more training data will also be used to increase the prediction accuracy. I am happy to support them in this, but I am no longer involved full-time.

About the prize

Every year, the best Bachelor’s and Master’s theses in the fields of medical informatics, e-health, health IT, health management, health economics and healthcare management and other degree programmes are awarded prizes at the DMEA (Digital Medical Expertise & Applications) in Berlin. The young talent prize is awarded for theses that sustainably improve healthcare using IT. The first prize for the best Bachelor’s thesis is endowed with 1,500 euros.