Recent years have seen an explosive increase in the use of lasers for medical applications, particularly in the field of dermatology where they are commonly used to achieve aesthetic, surgical, and therapeutic clinical objectives. Effective cutaneous laser procedures are achieved by tailoring the operating parameters of the laser to the physical and optical properties of the skin. Ideal laser parameters are selected to optimize therapeutic efficacy while minimizing unwanted side effects and tissue damage. Laser-induced tissue injury is known to occur via oxidative, photothermal, photomechanical, and photochemical mechanisms. However, the specific cellular and molecular pathways that initiate and govern these mechanisms are poorly understood. The primary objective of this research was to develop skin models and in vivo molecular imaging techniques to investigate laser-skin interactions. In this work, human skin cells, skin equivalent models, and animal models were developed to assess cellular damage associated with aesthetic, ablative, and therapeutic laser procedures. Thermal damage was assessed using a reporter gene system in which the activation of a thermo-responsive gene (hsp70A1) acts like an on-off switch for the expression and production of light emitting reporter genes. The models were used in this study to: (1) evaluate sublethal cellular damage in aesthetic laser procedures, (2) assess collateral damage in laser surgical ablation procedures, and (3) develop a therapeutic laser preconditioning protocol to enhance cutaneous wound repair. The use of skin models in conjunction with a thermally responsive reporter is a useful strategy for assessing sublethal thermal damage and is a valuable tool for improving medical laser procedures.

Image to physical space registration is a very challenging problem in image guided surgical procedures for the liver due to deformation and paucity of prominent surface anatomical landmarks. Iterative closest point (ICP) algorithm, the surface registration method used for registering the intraoperative laser range scanner (LRS) data with the preoperative CT data in image guided liver surgery, requires a good starting pose to reduce the number of iterations. Currently anatomical landmarks such as vessel bifurcations are used for an initial registration. This paper presents a computational approach to obtain the initial alignment that would reduce contact with probes for registration during surgical procedures. A priori user information about the anatomical orientation of the liver is incorporated and used to orient the point clouds for segmented CT data and LRS liver data. Four points are computationally selected on the anatomical anterior surface of CT point cloud data and corresponding points are localized on the LRS data using the orientation information. These four points are then used to find the rigid transformation using the singular value decomposition method. Nine datasets were tested using the computational approach and the results were compared using the anatomical landmarks method as the "gold standard". Seven of the nine datasets converged to the same solution using both the methods. The computational method, being an approximated approach may increase the number of iterations to converge to the solution. However since the method does not require precise localization of anatomical landmarks, it could potentially reduce OR time.

In a complex visual scene, different objects with specific features are presented in combinations, which raises the question of how the visual system processes and represents structure in the huge amount of visual information that it receives every moment. Besides independent response rate modulation, which has been traditionally considered to be the primary coding mechanism in the visual system, correlated neural responses in the form of synchronized neural firing have been proposed as providing a versatile means of encoding. In this dissertation, the role of neural correlation in visual perception was explored by analyzing synchronized responses in cat primary visual cortex. We first tested the neural response to collinear and cocircular contours and found that the strength of synchrony between cells is not only affected by the receptive field properties but also determined by the effectiveness of the visual stimulus in driving cells. Synchrony was found to be more reliable for detecting cocircular contours than the independent firing rate, suggesting that contour integration through neural synchrony could start as early as in the lowest level of visual cortex. We then explored the relationship between spike timing synchronization and coherent frequency oscillation. Strong correlation between cross-correlation analysis and coherence analysis suggests synchrony and coherence are internally related, though these two estimates reflect neural connectivity from different perspectives. By systematically perturbing the timing accuracy in neural responses, we also discovered that the temporal structures of spike trains are important in maintaining neural correlation. We last studied how the synchronized neural response modulates with the change of spatial integrity in visual stimulation. We found that the general association between neural spike trains depends strongly on spatial integrity, with coherence in the gamma band showing greater sensitivity to the change of spatial structure than other frequency bands. Temporal integrity, and not spatial integrity, generates synchrony; spatial integrity however is critical in triggering subsequent gamma band synchronization.

USING DIFFUSION TENSOR IMAGING TO ASSESS WHITE MATTER INTEGRITY IN CHILDREN WITH MATH DIFFICULTIES CRAIG THOMAS LORANG Thesis under the direction of Professor Adam Anderson Dyscalculia is a learning disability that interferes with a persons ability to understand and manipulate numbers. This condition affects up to 6% of all children. Previous studies have shown cortical functional activation in the parietal lobe related to number processing, however no studies have investigated the relationship between white matter integrity and number processing. Thirty-three subjects (mean age: 9.6 years) were imaged using a 3 Tesla Philips Achieva MRI scanner. Anatomical and diffusion weighted datasets were pre-processed and registered to a common space. Fractional anisotropy maps were mapped into the common space. Group t-tests were performed on a voxel-by-voxel basis on the FA maps between control and math difficulty (MD) groups. Linear correlations were performed on a voxel-by-voxel basis between FA and Wide Range Achievement Test Third Edition (WRAT) performance in the math and reading subtests. Regions in the left parietal and occipital lobes were found to have FA values correlating with math performance. A frontal lobe region was found to correlate with both reading and math performance, suggesting these two complex operations share portions of white matter bundles. These findings suggest white matter disorganization in regions critical for number processing. Further investigation will be needed to determine if intervention can change the developmental trajectory of these white matter pathways.

Early detection of breast lesions with malignant potential plays an important role in patient prognosis and survival. While X-ray mammography is the current clinical standard for screening and detection, traditional techniques such as palpation in the physical exam still play an important diagnostic role, and additional alternative means are actively being sought for the identification of suspicious lesions. This work has been focused on the development of a novel method termed modality-independent elastography (MIE) for the purpose of quantitatively extracting the material properties of tissue. This is quite relevant in the context of breast cancer, as solid tumors are typically stiffer than the surrounding unremarkable normal tissue. MIE performs an iterative non-rigid, model-constrained image registration analysis of a tissue under differing states of mechanical loading to produce a spatial mapping of elastic modulus values, which can be then used to characterize and/or localize a lesion. Simulation and phantom experiments were performed for two- and three-dimensional systems with a variety of image data acquired from CT, MR, and digital photography. Additional work produced a clinically compatible proof-of-concept system that is suitable for undergoing further refinement in preparation for initial human trials. Preliminary results have been encouraging and hold promise for the use of MIE in detection and characterization of abnormal stiffness within the breast.

The goal of this work was to develop methods to overcome the practical difficulties of 3T human myocardial imaging and to determine reliable reference values for transverse relaxation in normal human myocardium. Nine healthy volunteers were investigated with three multi-echo, turbo spin-echo (TSE) methods. Each method involved tradeoffs between acquired phase encoding lines per image and the number of echo-image sample points obtained along the T2 decay curve. Three multi-echo turbo field-echo (TFE) methods were also tested. The TFE methods highlighted differences between achievable bandwidth per pixel and echo time constraints versus the number of sample points obtained along the T2* decay curve. Measured transverse relaxation values in pixel maps and regions of interest, quality of monoexponential curve fits, and signal-to-noise ratio (SNR) were assessed among methods to determine accuracy and repeatability. Measured T2 and T2* values were consistent in reported means and in SNR across all scan methods. T2 for the ventricular septum was 59.5 ± 7.9 ms (N=9) across all TSE methods. The 4-echo method gave the best curve fits. T2* for the ventricular septum was 31.6 ± 6.1 ms (N = 9) with the 4-echo method yielding the highest quality curve fits. Significant differences between measured endo- and epicardial transverse relaxation due to myocardial perfusion were not observed. These results indicate that the 4- echo methods are best for optimal T2 and T2* sampling in the mid-ventricular septum.

Modern neuroscience seeks to understand the human brain and determine how stimulus features are directly mapped to neuronal representations that govern emergent properties like perception and behavior. One prominent proposal for population-based encoding of information is synchronization in the firing of two or more cells. Although synchrony has tremendous potential as a coding mechanism, understanding its relevance is difficult since the techniques to measure and analyze synchrony are relatively new. The work presented here combines simultaneous recordings from dozens of neurons with a novel method for identification of cellular assemblies defined by synchrony to investigate the dynamic associations among small populations during natural stimulation. We found that synchronous activity was able to discriminate changes in structural integrity and overcome the ambiguity of firing rate to identify contour structure reliably and consistently. The time course of synchrony suggests that it is directly related to spatial stimulus properties. Using a large natural image sequence with a variety of visual features to optimize stimulation of the entire recorded population, we showed that synchronous activity was correlated with the receptive field properties of proximity, orientation, and continuity. We used these properties to create a contour index, which quantitatively described how well an assembly's configuration matched a contour structure. Synchrony was well-correlated with this measure, which indicates that cooperation may be selective for local contrast structure arranged in continuous, well-defined contours. Synchronous activity is particularly sensitive to structural content in natural images, which is preserved in the phase regularities in the image and not the power spectrum. We found that synchrony measured between assemblies representing different contours was severely reduced. Responses were sparse for each assembly across the image set and well as across all assemblies for each condition. Our results demonstrate that synchrony has the potential to encode image properties not apparent from changes in firing rate. As a fundamental mechanism of sensory cognition, synchrony may act as a sparse code to help facilitate the detection of contour information for integration and processing in higher visual areas.

Recent advances in breast cancer imaging have generated new ways to characterize the disease, and many of these novel analysis techniques require the registration of breast surfaces during a non-rigid deformation. In this paper, a semi-automated method to register breast surfaces before and after a non-rigid deformation is presented. This method involves solving the Laplace or diffusion equations over the undeformed and deformed breast surfaces, yielding potential fields and isocontours that are used to establish surface correspondence. The proposed partial differential equation (PDE)-based registration method is compared to a thin-plate spline (TPS) interpolation of surface displacements tracked by fiducial markers. These registration methods are tested on a realistic finite element simulation of a breast compression, a breast phantom, and on a clinical data set. The results indicate that the TPS registration technique is the most accurate; however, if fiducial information is not available, the PDE-based registration methods may be viable alternatives. The PDE-based and TPS registration methods have the potential to be used by analyses requiring the non-rigid registration of breast surfaces before and after compression, and both have been used in Modality-Independent Elastography (MIE) to determine the boundary conditions needed for elasticity imaging.

This dissertation covers research regarding the use of computational models during brain tumor resection therapies. Systematic studies have shown that the brain tissue shifts during tumor resection therapies and that current image-guided systems do not account for this shift. Compensating for intraoperative brain shift using computational models has been used with promising results. For computational models to be clinically useful in tumor resection guidance, these models should meet the real-time constraints of neurosurgery and they should also provide images that mirror their intraoperative counterparts. The primary goal behind this research involves developing one such computational framework. More specifically, this framework involves combining a computational model with a linear inverse model to predict intraoperative brain shift. The framework reported in this dissertation relies on relatively inexpensive small scale computer clusters and can compute image updates on a time scale that is compatible with the surgical removal of tumor. In-vivo validation shows that the framework presented in this dissertation increases the efficiency and accuracy of image-guided systems. Results obtained have also been presented as graphical images for qualitative assessment. In summary this research constitutes a significant step towards using computational models for neuronavigation.

Most magnetic resonance imaging techniques offer tissue contrast but provide limited information regarding the variation of the magnetic resonance signal that exists on a smaller scale. The magnetic resonance signal arising from a heterogeneous tissue, such as spinal cord white matter, is the sum of signals from each tissue compartment within the imaging voxel. Analysis of this signal can better characterize the micro-anatomical heterogeneity tissue, white matter in particular. Many questions remain with regard to the compartmental contributions for the various types of magnetic resonance imaging (MRI) contrast. This project utilizes a variety of in vitro studies as well as simulations to better characterize the contribution of different water compartments to conventional MRI methods. Such an understanding of the complex combination of the various relaxation and exchange properties is important in developing an anatomical basis for interpreting magnetization transfer and T2 weighted images, particularly with respect to myelination.