Computer Aided Detection and Quantification of Brain Micro-Bleeds
Brain microbleeds (BMB) are hemosiderin deposits in the brain that are caused by leakage of red blood cells from small blood vessels. A BMB appears as a round or ovoid black signal on gradient echo T2* MRI or a susceptibility weighted MRI (SWI) scan. The prevalence of BMBs is about 5% in healthy people (but increases with age), 20% in people with Alzheimer, and about 34% in ischemic and 60% in hemorrhagic stroke patients. Current research focuses on assessing if BMBs could be an indication for an underlying pathology.
BMBs can also be a result of severe head trauma (as shown in the image above). The amount of micro-bleeds, their volume and their location may affect clinical status and outcome. For example, brain-stem injury is a predictor of mortality and poor functional outcome, whereas lesions at the temporal and frontal lobes may provide an explanation for memory and executive functioning disorders.
To find answers about the importance of BMBs in the assessment of the mentioned pathologies, an accurate annotation of these BMBs is required for each patient. The manual annotation of micro-bleeds is a very tedious and time consuming task (varying from 20 minutes to several hours per scan).
The goal of this project is to develop a computer-aided detection and quantification system that is able detect brain micro-bleeds, locate them on a brain atlas, and compute their volume.
This project is of particular interest to students who want to obtain skills in medical image analysis and machine learning.
The work will be executed in a high level environment for medical image analysis, at the Diagnostic Image Analysis Group (DIAG) of the Radboud University, in close collaboration with Fraunhofer MEVIS.
The DIAG group is highly renowned for their work in computer-aided detection and diagnosis (CAD). In this project we will build on the knowledge gained from CAD systems for breast, lung and prostate to create an algorithm for the automatic detection and quantification of brain micro-bleeds.
Fraunhofer MEVIS, one of the leading research and development centers for computer assistance in image-based medicine, will be a partner in this research. They will provide supervision, feedback, and a software environment containing a large amount of algorithms than can be used for this project.
- Do a literature search on existing methods for micro-bleed detection
- Consult neurologists and radiologists and find out about their requirements
- Research and develop features for brain micro-bleed detection
- Implement these features in our existing computer aided diagnosis (CAD) framework
- Train the CAD system
- Validate the results by comparing them to the annotations made by experts
- Discuss the outcomes with the clinical users
- Students with a major in computer science, biomedical engineering, artificial intelligence, or a related area in the final stage of master level studies are invited to apply
- Some affinity with programming is required