Lung cancer is the most deadly cancer in men and women. Early detection is the best strategy to reduce lung cancer mortality. Lung cancer is visible on a CT scan as a pulmonary nodule. If these are detected and confirmed to be cancer when they are still small, the disease is still local and lung cancer is curable.
The Diagnostic Image Analysis Group has developed state-of-the-art algorithms and software to find pulmonary nodules in CT scans, assess the type, measure their growth and estimate how likely it is that a nodule is malignant (cancerous). This software has been commercialized and is used worldwide in CT lung cancer screening programs. Recent guidelines state that also for CT scans obtained in routine clinical practice, nodules should be identified, measured and followed according to strict guidelines. The goal of this project is to further develop automated software to do this. It will be challenging to adapt the algorithms to deal with the much more diverse CT scans from clinical practice, where many abnormalities are visible in the lungs, compared to a screening scenario where the participants do not have complaints and have largely normal looking lungs. You will use deep learning and convolutional neural networks to analyze the scans. A focus will be on temporal analysis, as in most cases we will have multiple CT scans available for a patient, obtained at different time points. We will also focus on standardizing CT quality, in order to be able to process a wide variety of CT scanning protocols. Generative models and adversarial training should be investigated to achieve the standardized CT quality.
Another important part of the project will be the implementation and validation of the software in both Radboudumc and Jeroen Bosch Hospital. Together, both hospitals obtain over 10,000 of chest CT scans every year.
We are looking for an ambitious Ph.D. student to work on this four-year project that is funded by both hospitals who are strongly collaborating to implement various artificial intelligence solutions in the radiological routine workflow. You will be embedded in the Diagnostic Image Analysis Group, a large group with around 50 researchers all working on deep learning based medical image analysis.
You are a creative and ambitious researcher with an MSc/Ph.D. degree in Computer Science, Data Science, Physics, Engineering, Biomedical Sciences, Technical Medicine or similar, with a clear interest in machine learning, deep learning, medical image analysis. Good communication skills and expertise in software development, preferably in Python or C++, are essential. You should have an affinity with applied medical research. Experience with deep learning should be evident from the (online) courses you've followed, your publications, GitHub account, etc.
The candidate will be employed at Jeroen Bosch Ziekenhuis in Den Bosch and will be stationed both there, at the Department of Radiology, and at the Department of Radiology and Nuclear Medicine at Radboudumc in Nijmegen. The research is expected to result in a Ph.D. thesis defended at Radboud University Nijmegen.
Send applications by following this link. The following should be included: CV, list of followed courses and grades, a letter of motivation, and reprints or, preferably, links to your Master thesis or any scientific publications or reports in English that you have written. Applications will be processed immediately and the position will be open until a suitable candidate has been hired.