Algorithmic depth compensation improves quantification and noise suppression in functional diffuse optical tomography

Brain space image reconstruction of functional near-infrared spectroscopy using a Bayesian adaptive fused sparse overlapping group lasso model

Significance

Functional near-infrared spectroscopy (fNIRS) is a noninvasive technology that uses low levels of nonionizing light in the range of red and near-infrared to record changes in the optical absorption and scattering of the underlying tissue that can be used to infer blood flow and oxygen changes during brain activity. The challenges and difficulties of reconstructing spatial images of hemoglobin changes from fNIRS data are mainly caused by the illposed nature of the optical inverse model.

Aim

We describe a Bayesian approach combining several lasso-based regularizations to apply anatomy-prior information to solving the inverse model.

Approach

We built a Bayesian hierarchical model to solve the Bayesian adaptive fused sparse overlapping group lasso (Ba-FSOGL) model. The method is evaluated and validated using simulation and experimental datasets.

Results

We apply this approach to the simulation and experimental datasets to reconstruct a known brain activity. The reconstructed images and statistical plots are shown.

Conclusion

We discuss the adaptation of this method to fNIRS data and demonstrate that this approach provides accurate image reconstruction with a low false-positive rate, through numerical simulations and application to experimental data collected during motor and sensory tasks.

Prior Compensation Algorithm for Cerenkov Luminescence Tomography From Single-View Measurements

Cerenkov luminescence tomography (CLT) has attracted much attention because of the wide clinically-used probes and three-dimensional (3D) quantification ability. However, due to the serious morbidity of 3D optical imaging, the reconstructed images of CLT are not appreciable, especially when single-view measurements are used. Single-view CLT improves the efficiency of data acquisition. It is much consistent with the actual imaging environment of using commercial imaging system, but bringing the problem that the reconstructed results will be closer to the animal surface on the side where the single-view image is collected. To avoid this problem to the greatest extent possible, we proposed a prior compensation algorithm for CLT reconstruction based on depth calibration strategy. This method takes full account of the fact that the attenuation of light in the tissue will depend heavily on the depth of the light source as well as the distance between the light source and the detection plane. Based on this consideration, a depth calibration matrix was designed to calibrate the attenuation between the surface light flux and the density of the internal light source. The feature of the algorithm was that the depth calibration matrix directly acts on the system matrix of CLT reconstruction, rather than modifying the regularization penalty items. The validity and effectiveness of the proposed algorithm were evaluated with a numerical simulation and a mouse-based experiment, whose results illustrated that it located the radiation sources accurately by using single-view measurements.

Optical breast atlas as a testbed for image reconstruction in optical mammography

We present two optical breast atlases for optical mammography, aiming to advance the image reconstruction research by providing a common platform to test advanced image reconstruction algorithms. Each atlas consists of five individual breast models. The first atlas provides breast vasculature surface models, which are derived from human breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data using image segmentation. A finite element-based method is used to deform the breast vasculature models from their natural shapes to generate the second atlas, compressed breast models. Breast compression is typically done in X-ray mammography but also necessary for some optical mammography systems. Technical validation is presented to demonstrate how the atlases can be used to study the image reconstruction algorithms. Optical measurements are generated numerically with compressed breast models and a predefined configuration of light sources and photodetectors. The simulated data is fed into three standard image reconstruction algorithms to reconstruct optical images of the vasculature, which can then be compared with the ground truth to evaluate their performance.

Deep learning-based mesoscopic fluorescence molecular tomography: an in silico study

Fluorescence molecular tomography (FMT), as well as mesoscopic FMT (MFMT) is widely employed to investigate molecular level processes ex vivo or in vivo. However, acquiring depth-localized and less blurry reconstruction still remains challenging, especially when fluorophore (dye) is located within large scattering coefficient media. Herein, a two-stage deep learning-based three-dimensional (3-D) reconstruction algorithm is proposed. The key point for the proposed algorithm is to employ a 3-D convolutional neural network to correctly predict the boundary of reconstructions, leading refined results. Compared with conventional algorithm, in silico experiments show that relative volume and absolute centroid error reduce over ∼ 50 % whereas intersection over union increases over 15% for most situations. These results preliminarily indicate the promising future of appropriately applying machine learning (deep learning)-based methods in MFMT.

Dimensionality Reduction Based Optimization Algorithm for Sparse 3-D Image Reconstruction in Diffuse Optical Tomography

Diffuse optical tomography (DOT) is a relatively low cost and portable imaging modality for reconstruction of optical properties in a highly scattering medium, such as human tissue. The inverse problem in DOT is highly ill-posed, making reconstruction of high-quality image a critical challenge. Because of the nature of sparsity in DOT, sparsity regularization has been utilized to achieve high-quality DOT reconstruction. However, conventional approaches using sparse optimization are computationally expensive and have no selection criteria to optimize the regularization parameter. In this paper, a novel algorithm, Dimensionality Reduction based Optimization for DOT (DRO-DOT), is proposed. It reduces the dimensionality of the inverse DOT problem by reducing the number of unknowns in two steps and thereby makes the overall process fast. First, it constructs a low resolution voxel basis based on the sensing-matrix properties to find an image support. Second, it reconstructs the sparse image inside this support. To compensate for the reduced sensitivity with increasing depth, depth compensation is incorporated in DRO-DOT. An efficient method to optimally select the regularization parameter is proposed for obtaining a high-quality DOT image. DRO-DOT is also able to reconstruct high-resolution images even with a limited number of optodes in a spatially limited imaging set-up.

Depth-compensated diffuse optical tomography enhanced by general linear model analysis and an anatomical atlas of human head

One of the main challenges in functional diffuse optical tomography (DOT) is to accurately recover the depth of brain activation, which is even more essential when differentiating true brain signals from task-evoked artifacts in the scalp. Recently, we developed a depth-compensated algorithm (DCA) to minimize the depth localization error in DOT. However, the semi-infinite model that was used in DCA deviated significantly from the realistic human head anatomy. In the present work, we incorporated depth-compensated DOT (DC-DOT) with a standard anatomical atlas of human head. Computer simulations and human measurements of sensorimotor activation were conducted to examine and prove the depth specificity and quantification accuracy of brain atlas-based DC-DOT. In addition, node-wise statistical analysis based on the general linear model (GLM) was also implemented and performed in this study, showing the robustness of DC-DOT that can accurately identify brain activation at the correct depth for functional brain imaging, even when co-existing with superficial artifacts.

Sparsity enhanced spatial resolution and depth localization in diffuse optical tomography

In diffuse optical tomography (DOT), researchers often face challenges to accurately recover the depth and size of the reconstructed objects. Recent development of the Depth Compensation Algorithm (DCA) solves the depth localization problem, but the reconstructed images commonly exhibit over-smoothed boundaries, leading to fuzzy images with low spatial resolution. While conventional DOT solves a linear inverse model by minimizing least squares errors using L2 norm regularization, L1 regularization promotes sparse solutions. The latter may be used to reduce the over-smoothing effect on reconstructed images. In this study, we combined DCA with L1 regularization, and also with L2 regularization, to examine which combined approach provided us with an improved spatial resolution and depth localization for DOT. Laboratory tissue phantoms were utilized for the measurement with a fiber-based and a camera-based DOT imaging system. The results from both systems showed that L1 regularization clearly outperformed L2 regularization in both spatial resolution and depth localization of DOT. An example of functional brain imaging taken from human in vivo measurements was further obtained to support the conclusion of the study.

Clustered targets imaged by optical tomography guided by ultrasound

Clustered small breast lesions may be present in the neighboring areas and are difficult to accurately resolve and quantify in diffuse optical tomography. In addition, larger cancers are often accompanied by clustered satellite lesions in the neighboring areas, which are also difficult to resolve and quantify. To improve the light quantification of clustered lesions, a new multi-zone reconstruction algorithm guided by co-registered ultrasound (US) was investigated using simulations, phantoms, and clinical examples. This method separated one larger region-of-interest (ROI) into several ROIs based on the location information provided by co-registered US. In general, the single-ROI method cannot resolve two smaller targets when their separations were less than 2.5 cm and the depth was greater than 2.0 cm. The multi-zone reconstruction method improved the resolving ability and reconstruction accuracy. As a result, two targets located at 2.5 cm depth with separation greater than 2.0 cm could be distinguished, and reconstruction improved by more than 20% as compared with that of the single-ROI method. When two targets, one larger and one smaller, were located closer to each other, the location of the reconstructed absorption mass was shifted toward the larger target and the quantification of the smaller target was limited.

Enhanced functional brain imaging by using adaptive filtering and a depth compensation algorithm in diffuse optical tomography

Reflectance diffuse optical tomography (rDOT) of brain function is limited by its high sensitivity to the superficial tissues (i.e., the scalp and skull) and by its severe decrease in measurement sensitivity with increased depth. Significant interference in rDOT results from spontaneous fluctuations that are embedded in both the superficial tissues and brain, such as arterial pulsation and vasomotion. In this study, first we investigate coherence and phase shift of the spontaneous fluctuations in the resting state, within the superficial tissues and at various depths of the brain, respectively. We demonstrate that the spontaneous fluctuations originating from arterial pulsations ( ∼ 1 Hz) are spatially global and temporally coherent, while the fluctuations originating from vasomotion ( ∼ 0.1 Hz) tend to have less coherence with increased depth. Second, adaptive cancellation of spontaneous fluctuations with a frequency-specific strategy is utilized and validated in both resting and activation (evoked by a finger-tapping task) states. Third, improved depth localization of motor activation in reconstructed rDOT images is achieved by combining adaptive cancellation with a depth compensation algorithm that we recently reported.

Depth-correction algorithm that improves optical quantification of large breast lesions imaged by diffuse optical tomography

Optical quantification of large lesions imaged with diffuse optical tomography in reflection geometry is depth dependence due to the exponential decay of photon density waves. We introduce a depth-correction method that incorporates the target depth information provided by coregistered ultrasound. It is based on balancing the weight matrix, using the maximum singular values of the target layers in depth without changing the forward model. The performance of the method is evaluated using phantom targets and 10 clinical cases of larger malignant and benign lesions. The results for the homogenous targets demonstrate that the location error of the reconstructed maximum absorption coefficient is reduced to the range of the reconstruction mesh size for phantom targets. Furthermore, the uniformity of absorption distribution inside the lesions improve about two times and the median of the absorption increases from 60 to 85% of its maximum compared to no depth correction. In addition, nonhomogenous phantoms are characterized more accurately. Clinical examples show a similar trend as the phantom results and demonstrate the utility of the correction method for improving lesion quantification.