Home
The Department of Biomedicine
Computational and Personalized Medicine

Bergen Research Foundation awards funding to Professor Arvid Lundervold

Bergen Medical Research Foundation (BFS) has announced to award funding for 4-years to Professor Arvid Lundervold for his project „Computational Medical Imaging and Machine Learning».

Arvid og Alexander Lundervold på kontor med datamaskinen
Prof. Arvid Lundervold (left) with collaborator Alexander S. Lundervold in his office at the Department of Biomedicine. The computer equipped with NVIDIA GeForce 1080Ti graphic card will be duplicated and upgraded at the new Medical Imaging and Visualization Centre at the Department of Radiology.
Photo:
Arvid Lundervold

Main content

Lundervold is one of three beneficiaries of this call, that was specific for projects to be integrated with the novel Medical Imaging and Visualisation Centre. This centre is currently being established at the Department of Radiology, Haukeland University Hospital, and will provide state-of-the-art technology for medical research, diagnostics and education.

Machine Learning

Lundervold’s project will feed into the new centre by developing and deploying machine learning technologies that allow automated analysis of medical images. Machine learning, or more precisely “Deep Neural Networks” are computational techniques that allow computers to help analyze medical images in a similar way as a trained doctor would – partly faster, more reliably, and with much higher capacity. By employing this technique on the images acquired on high-tech medical imaging equipment, and integrating other data such as omics data, clinical findings and epidemiology, one claim is that the trained machines will be able to improve “precision medicine” for doctors and patients. The development of new methods in machine learning will play an important role in the project. This methodological research will be coordinated by Alexander S. Lundervold, a mathematician at Western Norway University of Applied Sciences.

Translational Research

Lundervold points out that “we don’t want to take the responsibility from the doctors – but we want to develop technology that can help and support them”. The project is therefore also highly translational in nature: many of the preclinical and clinical research groups in Bergen collaborate with Lundervold and his team. These have diverse biomedical interests, and usually provide images from patients, tissue, or cell samples that the computational group uses to develop algorithms, train machines, and then provide high-quality analysis of the samples. Research collaborations include aging and neurodegeneration (Laurence Bindoff, Haris Tzoulis), prostate cancer (CCBIO), endometrial carcinoma (Ingfrid Haldorsen), kidney (Jarle Rørvik, Olav Tenstad) and the brain-gut project (Trygve Hausken).

Promising results

“We just came back from a meeting in Berlin, where we presented our data on segmentation of the kidney”, explains Lundervold. He refers to his collaborators Alexander S. Lundervold and clinician Jarle Rørvik at Haukeland University Hospital. “We used images from dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), and applied deep learning technology. The computer was able to define where the kidney was in these images, just as a person would – analysing the gross morphology as well as the fine details of the left and right kidney – in less than five seconds.” Later, they want to use this technology to measure the filtration function of patient’s kidneys and assist in the diagnosis of kidney diseases.

e-Infrastructure

The generous funding from Bergen Research Foundation will be used to install and maintain state-of-the-art machine learning technology at the Centre.” We currently have some low-budget technology that we installed in collaboration with the Department of computing, mathematics and physics at the Western Norway University of Applied Sciences, and with funding from Helse Vest and UH-Nett Vest. Lundervold speaks proudly of his newest machine: ”It is actually a computer for computer gamers, a higher-end but still pretty standard machine, equipped with two NVIDIA GeForce 1080Ti graphics cards. We use it to develop our deep learning technology. I connect to it remotely from home”. This will be duplicated, and scaled up at the new Medical Imaging and Visualization Center. The center at the Department of Radiology will include clinical imaging equipment like magnetic resonance imagers (high field MRI) and computer tomography (PET/CT), and will be supplemented with the computer technology that can run the advanced analytical methods. Development is already well on its way: the formal opening of the Centre is planned for the middle of December 2017, where the three beneficiaries of the most recent call will give scientific talks.

Training and recruitment

“People with hands-on experience in deep learning remain scarce”, is an argument in the research proposal. The training of students at the University is therefore a central aspect of the project, and the team is already busy with designing e-learning courses and modules for various platforms. Deep learning will be modules in Lundervold’s class BMED360: In vivo imaging and physiological modelling. The newly developed course ELMED219 will start up next spring semester: Introduction to Computational Medicine and Biomedical Engineering. Both of these classes are open to students at the University of Bergen, as well as the Western Norway University of Applied Sciences, and other students can participate as guests.

To develop a teaching infrastructure on the topics of Translational Digital Pathology, Computational Biomedicine and Machine Learning and Biomedical Ethics is a focus area that has recently attracted substantial funding from the Erasmus + programme. The NordBioMedNet consortium, coordinated by Prof. Marit Bakke at our department and with Lundervold as a collaborator, was recently awarded for their multinational and multidisciplinary proposal “Open Educational Resources in Biomedicine”.

For those who want to develop professional skills in deep learning, starting up a master’s thesis with Lundervold or his collaborators at the Department of Biomedicine may be an exciting option. The BFS-funding will also allow the team to offer a PhD-position and a Post Doc position in the near future.