Development of neural network reconstruction algorithm for optical brain imaging
Page Content
Project type: master student
Abstract: We recently built first-in-class time-domain
near-infrared optical tomography system (TD NIROT). It is has a hand-held
design, and can be used in neonatal ICU. Now, we are looking for a new fast
image reconstruction technique, which combines a finite element method (FEM)
and a neural network. Goal: pave new ways for semi-real-time TD NIROT tomogram reconstruction.
Key words: newral network, TD NIROT, brain
oxygenation.
Date: September 2020
|  |
You will analyse the
state-of-the-art reconstruction process in NIROT and investigate the ways to
facilitate it with the focus on neural network. With the current deterministic
approach, it takes 10-15 minutes to produce a tomogram. Our goal is to reduce
this time to 1 min.
This project is the first step in a cascade of
research. The ultimate goal is to develop new reconstruction algorithm that
enables real-time optical tomography, which is hardly needed in neonatal
intensive care, and will improve (and save) lives of thousands yearly born
extreme preterms.
The project is hosted by Biomedical Optics Research Laboratory (BORL), Dept. of Neonatology, University Hospital Zurich. The lab has an expertise in NIRS and NIROT, including clinical applications. Since we are part of the hospital, we offer insights into clinical everyday life and the true needs of clinicians. In addition, we are an interdisciplinary team with a great spirit that offers you an exciting environment.
Qualification:
- skills in neural network and data processing,
- knowledge in NIROT is very welcome as well.
Contact:
Alexander Kalyanov, E-Mail: alexander.kalyanov@usz.ch, Tel.: 043 253 30 31.
Professor:
Prof. Martin Wolf
E-Mail: martin.wolf@usz.ch Tel: 044 255 53 46
Biomedical Optics Research Laboratory, Neonatologie, UniversitätsSpital Zürich, 8091 Zürich