GenIII Results

6.2.1

Testing and Certification

After_∼30 minutes warm up, the CW laser intensity were stable to<1%; remaining power variations were normalized using the low power arm of a 97:3 splitter placed between the wavelength and position optical switch. The position switch exhibited similar stability. The time domain system requires_∼4 hrs. to stabilize, primarily due to jitter in the laser diode driver oscillator.

DOT image reconstruction requires both forward and inverse calculations (see Section2.2.8). We used Finite Element Modeling (FEM) to compute the forward problem, with an optimized mesh created with Tet- Gen (http://tetgen.berlios.de/). The inverse problem was solved using the Time-resolved Optical Absorption and Scattering Tomography (TOAST85, 86_{) package made available by Arridge and collaborators at Univer-}

sity College London. We chose to utilize Conjugate Gradient (CG) reconstruction as this permitted us to reconstruct more voxels (i.g. larger volumes or smaller voxels) at a lower computational cost. However, this technique is known to effectively provide a low pass spatial filter over heterogeneities (e.g., as in Corlu244)

reducing the apparent contrast. Future work will include reconstruction with the more computationally intensive Levenberg-Marquardt (LM) method.

Simultaneous imaging with DOT and MRI also provides the opportunity to utilize the MR data in con- straining the DOT reconstruction by setting spatial bounds on tissue types (e.g. so-called hard priors, as in Carpenteret al.8) or by incorporating MRI information into a reconstruction, but not assuming perfect

alignment of contrasts across modalities (e.g. so-called soft priors, as in Inteset al.188and Guvenet al.235).

The phantom data presented below was collected using a Siemens TIM Trio 3T clinical MR scanner and the joint optical/MR imaging platform, using clinical research imaging protocols. We constructed various phantoms from Liposyn, sodium chloride, various inks, ICG, gelatin, and water. Recipes are described in AppendixC.

We undertook a set of measurements with sealed liquid phantoms and a suspended balloon target. The background Liposyn and saline solution had optical properties ofµa=0.04 cm−1andµa =8 cm−1at 800

nm. ICG was added to the target solution create a 4:1 optical contrast along with_∼0.5 mM MultiHance (clinical Gd-Chelate contrast agent) to create an MR contrast. A latex balloon (_∼8 cc) was filled with the target solution and suspended partially in the optical field of view. Data was normalized to measurements performed on the phantom without the target balloon in place; optical reconstructions were performed using TOAST85, 86_{. Figure6.15shows preliminary reconstructions from both modalities. Note that the optical}

reconstructions in Figure6.15 do not utilize spatially dependent regularization as this has not yet been implemented in the current incarnation of TOAST.

MRI

DOT

Figure 6.15: GenIII phantom imaging: Slices through the target center in a 3D reconstruction of a multi- modality phantom from MR (a,b,c) and DOT (f,g,i) imaging; (d,e,h) show schematics of the phantom geom- etry. Sources (red diamonds) and Detectors (yellow circles) are marked in all images. All dimensions are in mm. MRI is 3D T1-weighted spoiled gradient echo imaging with fat saturation. (a,h,i) are axial, (b,e,f) are coronal, (c,d,g) are sagittal slices. See text for details.

Figure 6.16: Example phantom reconstructions from the GenIII system. (left) Phantom Schematic. (center) 3D T1-weighted spoiled gradient echo imaging with fat saturation; slice through a gelatin-saline phantom with Gd chelate (0.5 mM MultiHance) in targets. (right) Optical reconstruction using TOAST (see text). Optical contrast is 4:1; background does not contain any MultiHance MR contrast agent.

=8 cm−1_{, at 800 nm. Cylindrical inclusions were cut in one of the phantoms and filled with a gelatin mixture} similar to the background, to which ICG to create a 4:1 optical contrast and_∼0.5 mM MultiHance for MR contrast had been added. These phantoms were reconstructed utilizing spatially dependent regularization as described in Section2.2.8. Gelatin phantoms are convenient for measurement in the MRI suite, but are vulnerable to air bubbles and are difficult to integrate into dynamic phantoms.

The results presented above demonstrate that the GenIII system can generate DOT images during simul- taneous MR data acquisition. Reconstruction and data collection are not optimized however. Future data sets will have more optode positions, extending over a greater vertical distance (posterior-anterior, currently

∼4 cm), additional spectral data as more laser wavelengths are added, and improved data collection proto- cols. Future reconstructions well benefit from our expanding spatial regularization techniques, optimized reconstruction parameters, and improved data pre-processing.

Figure 6.17: Initial human subject MR imaging with GenIII Opt/MR imaging platform. No optical data was collected during this MR certification test, but source and detector fiber modules were in place. Source modules are mounted in the medial 5 cm space; a fiducial marker on the source plate can be seen in this image. This bilateral MR scan is intended for Gd-Uptake imaging (3D T1-weighted spoiled gradient echo imaging with fat saturation- fat produces less signal and is therefore darker).

6.2.2

Initial Human Subject Data

We are currently awaiting recruitment of our first human subject; research has been approved by the Univer- sity of Pennsylvania Institutional Review Board under protocol 809792. MR only imaging from a healthy volunteer is shown in Figure6.17. In addition to the phantom measurements discussed in Section6.2.1, a single source-detector pair of the GenIII opto-electronic system has been utilized in the compression studies discussed in Section8.4.