Jul 3, 2012
Light patterns improve image reconstruction
Fluorescence diffuse optical tomography is an established molecular imaging technique that's well suited to reconstructing fluorescence distributions in small animals. A recent focus has been on reducing both the acquisition and reconstruction times, without sacrificing image quality. A key part of this work has involved the evolution of pattern-based illumination, and determining the most appropriate structured patterns of light to project onto the sample.
To address this question, researchers in Italy and the UK have performed what they believe is the first experimental comparison of different source patterns. Writing in Physics in Medicine & Biology, the team also introduces its virtual source patterns (VSP) method, which is designed to take full advantage of structured light (Phys. Med. Biol. 57 3811).
"Structured light reduces the number of illumination steps, which leads to the fast acquisition of a limited amount of data with a rich information content while preserving reconstruction quality," said Nicolas Ducros from the Department of Physics at the Politecnico di Milano, Italy. "Our VSP method allows us to take advantage of any illumination pattern, including those with negative and complex intensities."
A realistic phantom
The team's experimental set-up comprised three main components: an illumination system, a suitable phantom and a detector. In terms of illumination, the starting point was a helium-neon laser focused onto a digital micromirror device (DMD). The DMD consisted of 720 x 521 independently controllable tilting mirrors, capable of projecting the various source patterns on to the phantom.
An epoxy resin phantom with a radius of 20 mm and a height of 45 mm was chosen in order to mimic the optical properties and dimensions of a mouse. The phantom itself contained three fluorescent inclusions each with a radius of 2 mm. Fluorescence emission was then recorded by a CCD camera and used to reconstruct the internal fluorescence concentration.
Investigating source patterns
The researchers investigated five different illumination patterns, three of which were actual source patterns while two were derived using the VSP method. Every source pattern was projected onto the phantom for 35 s at 16 different rotational angles.
The first of the real source patterns considered was uniform illumination, where a constant intensity was projected onto every element of the DMD. The other two real source patterns were sinusoids and the scaling functions of the Haar wavelet basis. The VSP method was then used to generate a phase-shifted "phasor" pattern and virtual Haar wavelets. The reconstructed images for all five patterns are shown in the group's research paper.
"Real source patterns contain only positive values, whereas the VSP method offers the possibility of using any negative or complex pattern," commented Ducros. "VSP exploits the linearity of the forward problem. Consider two source patterns and the resulting fluorescence images: we can virtually obtain the response to the difference of the two patterns by forming the difference to the two fluorescence images. By taking the appropriate linear combination of positive patterns, one virtually considers any source pattern."
Improved reconstruction quality
In their results, the researchers consider the acquisition and reconstruction times, as well as metrics such as the contrast-to-noise ratio and reconstruction error for each source pattern. "The gain in reconstruction quality is substantial for VSP," report the authors. "A key feature of the VSP method is its ability to remove the DC components in the fluorescence images. Of the two VSP implementations, the wavelet approach offers the best performance, whether in terms of contrast-to-noise ratio, contrast or reconstruction error."
The team is now applying this technique to mouse measurements. "Our goal is to make a fully tomographic reconstruction of the concentration of highly selective fluorophores in small animals in a time of one to two minutes," said Ducros.
• Related articles in PMB
A virtual source pattern method for fluorescence tomography with structured light
Nicolas Ducros et al Phys. Med. Biol. 57 3811
Joint L1 and total variation regularization for fluorescence molecular tomography
Joyita Dutta et al Phys. Med. Biol. 57 1459
Excitation-resolved fluorescence tomography with simplified spherical harmonics equations
Alexander D Klose and Thomas Pöschinger Phys. Med. Biol. 56 1443
Illumination pattern optimization for fluorescence tomography: theory and simulation studies
Joyita Dutta et al Phys. Med. Biol. 55 2961
About the author
Jacqueline Hewett is a freelance science and technology journalist based in Bristol, UK.