Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Development of Deep Learning Methods for Medical Imaging Synthesis and Image Characterisation
The Department of Radiation Oncology at the University Hospital Zurich is currently offering various Master's and Bachelor's thesis projects in clinical applications of deep learning algorithms.
Keywords: Keywords: Deep Learning, Machine Learning, Artificial Intelligence, Radiation Oncology, Big Data, Radiomics, Computer Vision, Cancer
Deep Learning (DL) algorithms have shown compelling results for segmentation tasks in recent years. In the oncological field, tumor segmentation is pivotal for analyzing tissue characteristics and data mining, such as for radiomics, which might improve survival by determining the best therapeutic interventions for each individual (personalized medicine).
In a large multi-center study on Glioblastoma multiforme, currently, the segmentations of the tumors are precisely computed with a DL algorithm that requires four different Magnetic Resonance (MR) imaging contrasts. For some patients, only three imaging contrasts are available; consequently, the tumor segmentations cannot be computed and the patient has to be excluded from analysis. We aim to solve this issue by generating the missing fourth image contrast from the three available ones by using deep learning-based medical imaging synthesis. Additionally, this synthetic image will be used as input for quantitative image analysis.
Deep Learning (DL) algorithms have shown compelling results for segmentation tasks in recent years. In the oncological field, tumor segmentation is pivotal for analyzing tissue characteristics and data mining, such as for radiomics, which might improve survival by determining the best therapeutic interventions for each individual (personalized medicine). In a large multi-center study on Glioblastoma multiforme, currently, the segmentations of the tumors are precisely computed with a DL algorithm that requires four different Magnetic Resonance (MR) imaging contrasts. For some patients, only three imaging contrasts are available; consequently, the tumor segmentations cannot be computed and the patient has to be excluded from analysis. We aim to solve this issue by generating the missing fourth image contrast from the three available ones by using deep learning-based medical imaging synthesis. Additionally, this synthetic image will be used as input for quantitative image analysis.
In this project, the candidate will participate in investigating, optimizing, and developing current DL algorithms for imaging synthesis and image analysis. Moreover, the candidate will contribute to curating the real-world dataset continuously acquired at several clinics.
_Your profile_ - You have a technical background (computer science, physics, or engineering) and ideally previous experience with machine learning algorithms. Basic knowledge of Python and/or other scripting languages is a must. Interest and motivation to work with many clinical images in a multi-disciplinary team are essential.
_What we offer_ - Interesting topics in the machine learning domain with real-world applications in a young group with many master's and doctoral students.
In this project, the candidate will participate in investigating, optimizing, and developing current DL algorithms for imaging synthesis and image analysis. Moreover, the candidate will contribute to curating the real-world dataset continuously acquired at several clinics.
_Your profile_ - You have a technical background (computer science, physics, or engineering) and ideally previous experience with machine learning algorithms. Basic knowledge of Python and/or other scripting languages is a must. Interest and motivation to work with many clinical images in a multi-disciplinary team are essential.
_What we offer_ - Interesting topics in the machine learning domain with real-world applications in a young group with many master's and doctoral students.
_Contact_ - Dr. sc. nat. Orso Pusterla, orso(dot)pusterla(at)usz(dot)ch
_Contact_ - Dr. sc. nat. Orso Pusterla, orso(dot)pusterla(at)usz(dot)ch