June 2024
Ekin Ermis is a Senior physician at the Department of Radiation Oncology of the Inselspital, Bern University Hospital, where she validates AI-based automation for the radiotherapy pipeline developed by the University of Bern. Over the last seven years, she has dedicated research and interdisciplinary dialogue to improving workflows and assuring sooner and high-quality treatment for patients with brain tumors.
Dr. Ermis, why would you involve AI in the treatment of patients?
Our main aim is improving the oncological outcome for patients with brain tumors while maintaining or even increasing their quality of life. I am responsible for patients with brain tumors, both benign and malignant. I have long years of relationship with those facing benign tumors. After receiving definitive treatments, we perform several years of follow-up care for those patients, where we care maintaining good quality of life.
Unfortunately, patients who have aggressive brain tumors mostly have a short life expectancy. In this case, we aim to significantly reduce the time from diagnosis to treatment, currently spanning two weeks, to just two days. Accelerating the start of treatment is of enormous importance to patients. Especially given the psychological impact of the diagnosis, it may even have a positive impact on the prognosis. By automating the more repetitive and numerical aspects of my work, particularly in radiation oncology, I can reclaim valuable time to dedicate to patients and their families.
We want to shorten the time to treatment for the patient from currently 2 weeks to 2 days.
How exactly would I have to imagine this?
In 80 percent of the time radiation therapy planning is repeating the same six steps which generally take two weeks: image acquisition, manual delineation of the tumor, definition of the therapy target and sensitive areas to spare nearby (organs-at-risk), generation of the radiation plans by medical planers/physicists, quality assurances of the generated plans, application of treatment and further follow-up. As most of this work is done on the computer, our field is very open to involve AI-based automatization. This could not only save time but also support more standardized decisions in radiation therapy planning without inter-physician variability.
We have already implemented AI tools into our clinical practice today. These mainly involve auto-segmentation of organs at risk that require sparing from adverse radiation effects. While we have not quite reached the point of utilizing AI for defining therapy targets, I am confident this will become a reality within the next years. Our department is generating a huge amount of data – especially in neuro-oncology. So far, this data has not been exploited to develop technology. But we can now translate it into patient care.
What is your personal motivation?
In Switzerland we work in a dynamic environment with many opportunities. After spending 7 years here, I see that there is a lot of financial support for cancer research. On the other hand, there is a great freedom of research. If you have an idea – any idea – you can go for it and reach your goals. I have experienced a lot of support from interdisciplinary working groups and flat hierarchies that allow you to transfer ideas into concrete tools – as a resident, as a consultant and as a professor. I know that if we give enough energy, time, and effort, we can make it real.
I also feel very involved into the research with our technical partners. They are remarkably open to receive feedback from someone without a technical background. Students from the ARTORG Center frequently visit our department, where I give them insights into our clinical workflow, our equipment, and our daily practice. We have years of experience spending time to understand each other. This kind of relation needs building up. Now, we have reached a point were communication flows smoothly and effectively.
If you have an idea – any idea – you can go for it and reach your goals.
Where do you see AI in the clinic in 20 years?
It is likely that AI could reduce the number of needed clinical staff. However, that might actually give us more time to focus on improving our environment. We as doctors in clinical routine are quite submerged in the workload. Should AI interventions reduce this burden, it could mean we get to gather and brainstorm ideas to innovate and refine the practice of medicine, thereby enriching patient care.
An ongoing challenge is how to embed AI alongside with more “traditional skills” in medical education. For instance, trainees in radiation oncology would still require proficiency in manually segmentation, even when automated results are available. Otherwise, there is a risk that residents might rely too heavily on AI, potentially overlooking errors and blindly endorsing its analysis.
