Janine D. Bijsterbosch, PhD, and Aristeidis Sotiras, PhD, assistant professors of radiology at MIR, use advanced computer technology to quantify large volumes of data to, respectively, study functional brain connectivity and develop novel computational algorithms for brain image analysis. Although they take different approaches, their goals are the same: to better understand and interpret the brain’s complexities, and therefore advance personalized treatments to best meet individual patients’ needs.
Janine D. Bijsterbosch, PhD
Bijsterbosch’s research focuses on how functional connectivity networks in the brain differ from person to person. By identifying the variables in these “personalized connectomes,” there exists the potential for understanding and predicting differences in behavior, performance, mental state, disease risk, treatment response and physiology.
To characterize connectivity, Bijsterbosch uses functional magnetic resonance imaging (fMRI) of individuals in a resting state. A major resource for these images is the UK Biobank Imaging Study, which aims to scan 100,000 people over five to six years. With no explicit inclusion or exclusion criteria, the genetic diversity of the UK Biobank provides Bijsterbosch with a rich set of phenotypic variables. From these, she aims to differentiate between shared connectivity abnormalities and unique markers of disease through cross-diagnostic research.
The ultimate goal of Bijsterbosch’s research is finding markers that indicate where individuals fall within a mental health continuum. An MRI could then inform part of patients’ treatment indication. For instance, those at risk for developing anxiety or depression could be given the tools they need to deal with stressors that may trigger mental health issues. Another long-term goal may be treatment response predictions that indicate which antidepressant is best suited for someone with particular markers along the continuum.
Aristeidis Sotiras, PhD
The goal of Sotiras’ work is to teach computers to see images not as a collection of pixels of varying intensity, but instead to interpret them as humans do — with the ability to easily identify anatomical structures such as the heart and brain.
Combining the sensitivity of human perception with a computer’s ability to quantify large volumes of information is vital to advancing diagnostic and prognostic abilities and treatment planning, as well as understanding brain health in general. To that end, Sotiras is interested in developing unique computational algorithms and applying them to various problems. His methodological work concentrates on developing novel algorithms for image analysis that involves image segmentation, for example, labeling regions of interest. Another use is image registration — the alignment of images in the same coordinate system — to compare individuals and build models.
Finally, Sotiras is developing machine-learning techniques for indentifying patterns in data and creating predictive models. His work has applications in brain development, brain aging and brain tumor segmentation. Sotiras hopes to advance automation and quality assurance workflows in clinical practice in order to help clinicians manage increasing data volume. But ultimately, he hopes to use machine learning to fully utilize large volumes of complex imaging and nonimaging data to advance the model of precision medicine.