Image Processing of Brain Activity via functional Ultrasound Thesis Topics

Job Description

Techniques for live brain imaging grew out of radiology, initially involving radioactive tracers (for example, positron-emission tomography (PET)), followed by the momentous discovery that MRI (Magentic-Resonance Imaging) could measure intrinsic hemodynamic signals linked to neural activity (functional MRI or fMRI), a significant technological advancement of the past two decades. The next big technological leap of late is functional Ultrasound (fUS). Functional ultrasound imaging is able to assess local changes in cerebral blood volume during cognitive tasks, with sufficient temporal resolution to measure the directional propagation of signals. With the advent of novel mathematical, statistical and computational techniques in science at large, and with the dramatic growth of technology and computing power, only now a fUS imaging system is possible.

We want to explore and study the performance of typical imaging and analysis algorithms, applied on real raw fUS brain scans (animal & human subjects). There is a need to determine the most promising among the available algorithms that can functionally correlate the images and, even more importantly, profile the computational workload they introduce.
This theme entails multiple diploma theses each having as a main objective to study, optimize and accelerate – if necessary – different imaging and statistical-analysis techniques, including FFT transformations, beamforming and convolution filters, ICA/PCA/FCMA analysis.
Per thesis topic, modeling, implementation, benchmarking, acceleration, exploration or – even – surveying of diverse imaging and analytics algorithms may be involved.
These theses are given in conjunction with the Neuroscience department of the Erasmus MC (neuro.nl).

 

Reading Material

(fUS uses similar analysis techniques with fMRI, but the computational load is orders of magnitude higher.)
1) “Computational approaches to fMRI analysis”,
Cohen et al., in Nature of Neuroscience, 2017.
2) “Full correlation matrix analysis of MRI data“,
Wang et al., in Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, 2015


Contact:
Assist. Prof. Christos Strydis: c.strydis@erasmusmc.nl
Harry Sidiropoulos: harry.sidi@gmail.com
or use form Below

Minimum Qualifications

Required Knowledge (depending on the particular topic selected):
1) C/C++, Python, CUDA, openMP, Vivado C
2) Codebase Version Control with Git
3) Linux Operating System
4) Signal processing

Prefered Qualifications

Optional: Computational statistical/informatics tools (PCA, ICA, etc.)

Academic Advisors:
Assist. Prof. Christos Strydis: c.strydis@erasmusmc.nl


Apply for Job