MSc Thesis projects
If you are interested in doing your MSc thesis, there are a number of MSc thesis projects available. These projects involve either image processing, video analysis, neural actvity analysis or computational modelling of the brain.
Automated identification of Extracellular vescicles:
Background: Think about how hormones act in your body. The glands that produce hormones and the organ where those have to be delivered may be widely separated in the body. How do the hormones produced in the gland reach the organ and only that organ? This magic is done through extracellular vesicles (EVs) , which are nano-vesicles generated by all cells. Glands pack their hormones in EVs and send these EVs to blood. EVs carry signatures which are recognized only by their intended targets. Understanding EVs’ biophysical properties is important to understand chemical communication in the body and it has important implications for targeted drug delivery. Study of EVs is now an exciting branch of medicine and biology.
Research problem: We are interested in biophysical properties of EVs, in particular, their deformability which can be measured by atomic force microscopes (AFMs). AFM measures the surface irregularities using force-distance curves. Some of these irregularities are EVs. Experimentalists are interested in identifying EVs so they can do their measurements.
The thesis project: The machine learning challenge is to automatically identify putative EVs. Experimentalists can provide labelled images so this is a supervised learning task. If we can automate AFM measurements of EVs with machine learning algorithms we can speed up the process of measurement of deformability by at least an order of magnitude. This would be a game-changer in the field. This is a thesis project to develop such algorithms based on the images we already have.
Collaborations: Dhrubaditya Mitra and Apurba Dev.
Neural data analysis: We are interested in characterizing brain-wide changes in the neural activity during Parkinson's diseases (PD). To this end we have collected MEG data from healthy controls and patients with PD. This dataset is quite large in terms of number of MEG channels and duration of recording. Our preliminary analysis already shows some crucial differences in healthy controls and PD patients. There are still numerous opportunities to analyze this data to understand how interaction between different regions is altered during PD.
Required expertise (0ne or more of the following) : Machine learning tools, signal processing, stochastic processes
Collaboration : National MEG Facility , Daniel Lundqvist, Per Svenningsson(KS)
Analysis infant movements from videos : Infant movements are often useful in determining whether an infant might develop other disorders such as epilepsy. Clinicians routinely look at the infant's movements for this kind of diagnosis. We wish to make this more objective. And that requires that we have a data analysis pipeline to process infant videos (1) to track hands and legs using markerless tracking methods (eg deeplabcut) and (2) to analyze the tracking data and identify signatures of future disorders from hand/leg movement motives.
Required expertise (0ne or more of the following) : Machine learning tools for marker-less tracking, Unsupervised clustering methods
Collaboration : Heléne Sundelin (KS)
Multi-scale models of biological neuronal networks : Understanding dynamical and information processing properties of biological neuronal networks is fundamental to understand the brain function. In my group we are interested in (1) isolating the role of various neuron and synapse weaknesses and (2) understanding how network structure affects network activity dynamics. Now more and more new data about the neurons, synapses and connectivity is becoming available. So we would like to integrate that in our models. Besides theoretical interest, we can adapt such models to understand mechanisms underlying brain diseases.
Required expertise (0ne or more of the following): Strong background in Physics, Maths or Electrical Engineering and motivation to ask and understand scientific questions (not just training a network).
Start date: As early as possible
Project duration: 3-4 Months
Contact Arvind Kumar ( )
More information about Arvind's research group: NeuroLogic
