PhD project – Deep learning for auto segmentation and dose prediction in the brain
Call for applications for a fully financed PhD fellowship
High precision radiotherapy techniques exploit steep dose gradients; therefore, it is of paramount importance that there is a clear and concise definition of what is to be treated (target) and what should be sparred (critical normal tissue organs). Any errors in this step will either lead to inadequate dose coverage, reducing the chance of tumor control, or cause unnecessary toxicity. This is an actual clinical problem, as shown in a recent European study revealing that around one-third of delineation had errors that would impact patients’ treatment.
Variation in the treatment planning quality is also a challenge that results in suboptimal treatment of patients. Identifying potentially subpar plans that could be further optimized would again improve target coverage and reduce toxicity.
Therefore, there is a clear need to ensure the quality of the definition of target and normal tissues and the treatment plans.
Deep learning semantic segmentation can automatically define the target and normal tissue structures through a learning process from prior expert delineations. This tool, along with a centralized server infrastructure (dcmCollab), will then automatically be able to assess the doctor-defined delineations and potentially identify subpar contours from the different referring treating centers in Denmark. If a poor delineation is flagged, the physician will be notified, and feedback will be used to improve the deep learning model in a prospective setting continuously.
Dose prediction is also a method that is trained on prior expert radiotherapy treatment plans to “guess” the delivered dose. Large deviations from the predicted dose distribution can then be flagged, and the planner can be notified. Again based on the feedback of the planner, this will be used to improve the predictions.
Successful implementation of the proposed solution will be an invaluable tool to ensure that the number of patients treated according to the national guidelines with the best possible quality is maximized.
In this position, you will implement and test deep learning semantic segmentation and dose prediction in a national setting through the radiotherapy database dcmCollab. You will have access to retrospective data for training the auto segmentation and dose prediction models. Collecting, curating, and analysing this data will also be an integral part of your job. The main goal will be to take these methods to the next level and make them ready, so we can confidently integrate them into clinical workflow and improve the treatment for brain cancer patients.
Work and collaborators
This multidisciplinary project will take place at Aarhus University Hospital, with strong ties to Odense University hospital (Ass. Prof. Christian Rønn Hansen); you will be housed at the Danish Center for Particle Therapy.
The project will be solidly anchored in the Danish Neuro Oncology Group (http://www.dnog.dk/forside), The DCCC Brain tumor Centre (https://dcccbraintumor.dk/) and you will be embedded in the research group of associate professor Jesper Kallehauge (https://www.linkedin.com/in/jekallehauge/). The group is part of the collaborative oncology research environment, where over 30 Ph.D. students, postdocs, and more than ten senior full-time researchers are working together.
If you want to be part of a large national project that aims to change the current practice of treating brain cancer patients and you like to work with people that are ambitious and driven in a vibrant and fun environment, we encourage you to apply for this position.
For further questions, don’t hesitate to get in touch with Jesper Kallehauge (firstname.lastname@example.org). Qualified applicants will be interviewed in late September.
- Master degree in physics, medical physics, biomedical engineering, computer sci-ence, or equivalent
- Prior experience in programming and deep learning is highly desired (Python, Matlab, or C++)
- Fluency in English (oral and written)
- Prior experience in image processing and radiotherapy will be an advantage
How to apply
Please submit your application via this link. Application deadline is 16 September 2022 23:59 CET. Preferred starting date is 1 October 2022.
For information about application requirements and mandatory attachments, please see our application guide.
Please contact Associate Professor Jesper Kallehauge, email@example.com, for further information.
All interested candidates are encouraged to apply, regardless of their personal background. Salary and terms of employment are in accordance with applicable collective agreement.