Optimization of probe geometry for diffuse optical brain imaging based on measurement density and distribution

Abstract

Optode geometry plays an important role in achieving both good spatial resolution and spatial uniformity of detection in diffuse-optical-imaging-based brain activation studies. The quality of reconstructed images for six optode geometries were studied and compared using a laboratory tissue phantom model that contained an embedded object at two separate locations. The number of overlapping measurements per pixel (i.e., the measurement density) and their spatial distributions were quantified for all six geometries and were correlated with the quality of the resulting reconstructed images. The latter were expressed by the area ratio (AR) and contrast-to-noise ratio (CNR) between reconstructed and actual objects. Our results revealed clearly that AR and CNR depended on the measurement density asymptotically, having an optimal point for measurement density beyond which more overlapping measurements would not significantly improve the quality of reconstructed images. Optimization of probe geometry based on our method demonstrated that a practical compromise can be attained between DOI spatial resolution and measurement density.

Fig. 1.

Six optode geometries investigated in this study (×, source fibers; ∘, detector fibers; ⊗, bifurcated fibers for both source and detector). (a) to (f) correspond to G-I to G-VI. The optodes in each geometry were placed over 1cm rectangular grids (dotted line) within a constant FOV of 4cm × 4 cm; they were utilized in the phantom experiments.

Fig. 2.

Schematic diagram of the experimental setup. A 5 × 5 square optode array was placed on the top surface of an Intralipid phantom. The separation of neighboring optodes was 1.0cm. A thin cylindrical absorbing object (ϕ= 1.1 cm) was placed into the phantom during the measurements with one circular side facing up. The depth of the object in the phantom was 1.5cm below the surface. Two object positions were measured separately; one was at the center (P1) and the other one was one grid (= 1 cm) shifted along the y direction (P2).

Fig. 3.

Spatial distributions of overlapping measurements in six geometries. (a) to (f) correspond to G-I to G-VI. The locations of sources and detectors in each geometry (×, source fibers; ∘, detector fibers; ⊗, bifurcated fibers for sources/detectors) are superimposed in the corresponding map. The black circles (P1, solid circle; P2, dotted circle) indicate the two object locations that were placed and measured separately.

Fig. 4.

Reconstructed images of the absorbing object at P1. (a) to (f) correspond to the reconstructed images from G-I to G-VI. The dotted circle indicates the true object region. The locations of sources and detectors in each geometry (×, sources; ∘, detectors; ⊗, bifurcated for sources/detectors) are superimposed in the corresponding images.

Fig. 5.

Reconstructed images of the absorbing object at P2. (a) to (f) correspond to the reconstructed images from G-I to G-VI. The dotted circle indicates the true object region. The locations of sources and detectors in each geometry (×, sources; ∘, detectors; ⊗, bifurcated for sources/detectors) are superimposed in the corresponding images.

Fig. 6.

Correlation between the reconstructed image quality and the mean measurement density in the object region: (a) AR versus mean measurement density and (b) CNR versus mean measurement density. In each graph, two sets of data, one each from the P1 (circle) and P2 (plus) positions, were combined.

Fig. 7.

Schematic diagram to demonstrate an insensitive region of an optode. The ROI is right under detector D1. In this case, little portion of the light received by D1 passes through the ROI and, thus, D1 is not sensitive to the absorption perturbation in this region. Some neighboring detecting optodes, i.e., D2 and D3, are needed to detect the absorption perturbation in this region.