Introduction

Electrical Engineering Senior Capstone Project

Kevin Bernal-Lazo (kbernal-lazo@wustl.edu) and Jordan Thomson (d.thomson@wustl.edu)

Advised by Jason Trobaugh, Professor of Practice

Project Overview

Diffuse Optical Tomography (DOT) is a medical imaging technology that uses near-infrared spectroscopy to image tissue. It is lighter and more portable than conventional imaging technologies like fMRI, but has the drawback of limited ability to image deep into tissue because the system depends on light propagation.

Our project examined the use of DOT to image pigeon brains with various cap designs. Using MATLAB to simulate the system response for each cap, we collected data that characterizes the imaging resolution throughout the region of interest.

Technical Background

A DOT cap is comprised of many variable-wavelength light emitters interspersed in a grid-like pattern with light detectors. The measurements obtained by this cap are used to reconstruct a volumetric image of the tissue, which in the case of brain imaging corresponds to blood oxygenation and thus neural activity.

An example of a DOT cap fitted to a model of a pigeon head. The blue nodes are light detectors and the red nodes are light emitters.

Our work consisted of cap design, the alignment of a pigeon atlas (a record of brain sub-regions of interest) with the CT scan used to design DOT caps, and the characterization of the imaging system based on mathematical simulations of a specific cap.

We use prior work done that characterizes the system performance based on the arrangement of the DOT cap and the volume that we are interested in imaging. Imaging systems are often viewed in terms of point-spread functions (PSFs), which are the response the system sees given activation at a single voxel (a volumetric, or 3-dimensional, pixel).

A three-plane view of a point-spread function, with the point of activation indicated.

There are three measures of PSFs that interest us:

  1. The size of the PSF. This is measured in terms of full width at half maximum (FWHM), meaning we identify the area of the PSF at greater in value than half of the maximum value and then measure its width. In the figure above, you can see from the scale that the FWHM of the PSF is the only area displayed – values below 0.5 are not shown. A larger FWHM, or a larger PSF, corresponds to poorer imaging quality – it means that the system can less precisely reconstruct the true tissue image.
  2. The localization error of the PSF. This is the distance between the point of activation and the center of the resulting PSF. In the figure above, this would be the distance between the blue point and the center of the most intense red region. A larger localization error corresponds to poorer imaging quality – is PSFs emerge far from the activation that caused them, reconstructing the true image accurately becomes difficult.
  3. Signal-to-Noise ratio, or SNR. Part of our simulation is noise inherent to the measurement process. We can measure the relative strength of the simulated signal to the noise. A higher SNR is desirable as it means the image reconstruction contains more useful information.