Computed tomography (CT) is a well-known technique in radiology, in which X-rays are emitted from different directions around a patient. These X-rays are attenuated by different tissue types on their path, and are finally measured by a detector. These projections are used in a reconstruction algorithm to calculate cross-section images, representing the local X-ray absorption in the patient. False assumptions on the physical process are necessary to solve this reconstruction problem but often result in artifacts which have a negative impact on the reconstructed image quality. These artifacts worsen with lower dose or in cheaper modalities such as cone beam CT and tomosynthesis systems. Standard approaches have been used for many years to (partly) tackle these artefacts. With this project proposal we want to introduce convolutional neural networks in the reconstruction to reduce these artifacts. It will result in better image quality, especially at low dose and for cheaper modalities. Convolutional neural networks recently gathered vast interest. This uptake motivates us to apply it within our imaging solutions. The general target of this project is to improve current tomographic reconstruction algorithms by introducing a machine learning approach, in particular convolutional neural networks. Very recent academic research has shown the potential of such an approach. You will implement, improve and investigate the usability in medical CT images, and develop novel neural network approaches. In this way, you will build a competitive CT reconstruction as a basic technology for future 3D modalities.
The Vision Lab is a research group of the physics department at the University of Antwerp. The Vision Lab has unique expertise in the development of algorithms for reconstruction, processing and analysis of tomographic imaging data. Application domains are x-ray computed tomography (CT), MRI and electron tomography. The working environment is strongly interdisciplinary, combining techniques and insights from physics, mathematics, engineering, and computer science. The Vision Lab has a broad range of national and international collaborations with both academic and industrial partners. Recent Vision Lab publications on computed tomography can be found on http://visielab.uantwerpen.be/research/tomography . This project is in collaboration with Agfa Healthcare.
What will you do
You will primarily work with developing computed tomography reconstruction and classification methods with industrial application. A core part of the project is the development of novel neural network approaches, and their implementation, improvement and investigation of the usability in medical CT images. Furthermore, is it required that the you participate with the rest of the team to achieve the overall goals of the projects to ensure a successful implementation of the obtained solutions with the end goal of integration in a commercial X-ray scanner.
What we do for you
- An exciting 2-year research project in a dynamic and international context
- Multidisciplinary research: cooperation with strong academic research groups (collaboration with Agfa Healthcare)
- A world-class research environment with state-of-the-art instrumentation
Who you are
Candidates should have a PhD degree or equivalent in computer science, physics, mathematics or engineering
Experience with image analysis preferably advanced feature based recognition, classification methods and machine learning for image analysis
Experience with tomographic reconstruction techniques is an advantage
Enthusiasm for working with industry related problems
Good programming skills and experience with GPU implementation is an advantage
Skills for working in a team
Additional information about the vacancy can be obtained from: Prof. Dr. J. Sijbers, (Jan.Sijbers@uantwerpen.be).
Interested candidates are invited to send a motivation letter, a detailed CV, and contact info of at least two references to Christine Waerebeek, (Christine.Waerebeek@uantwerpen.be) with subject HR_AGFA_CT_2018. The preferred starting date is between 1 September 2018 and 1 November 2018, but will be adapted to the selected candidate’s availability. The deadline for submissions is 30 September 2018.