Carbon nanotube field-effect transistors (CNFETs) are considered to be promising candidate beyond the conventional CMOSFET due to their higher current drive capability, ballistic transport, lesser power delay product and higher thermal stability. CNFETs show great potential to build digital systems on advanced technology nodes with big benefits in terms of power, performance and area (PPA). Hence, there is a great need to develop proven models and CAD tools for performance evaluation of CNFET-based circuits. CNFETs specific parameters, such as number of tubes, pitch (spacing between the tubes) and diameter of CNTs determine current driving capability, speed, power consumption and area of circuits and play a significant role in accurate PPA evaluation. Furthermore, count and density variations in carbon nanotubes (CNTs) due to manufacturing limitations, like the presence of metallic tubes in the CNFET channel, pose major obstacles to robust and energy-efficient CNFET digital circuit designs and degrade the anticipated PPA benefits. CNFET-based circuits can suffer from large performance variations and reduction in functional yield due to these variations in CNFETs. Moreover, modeling the CNFET parameters, CNT variations and etching techniques for CNTs create additional complexity during performance optimization. Hence, for realistic optimization of CNFET circuit's performance, it is imperative to incorporate the impact of these parameters and variations. We present a capacitance-based Logical Effort (LE) framework to investigate design issues of high-speed and low-power circuit designs implemented by considering specific requirements and challenges of the CNFET technology. The LE technique is widely recognized as a pedagogical method to quickly estimate and optimize the propagation delay and transition time in CMOS circuits equivalently without performing transient simulations and detailed delay calculations. In this thesis, we propose novel delay models [Pitch-Aware Logical Effort (PALE) and Position-Aware Pitch Factor (PAPF)] for fast and accurate performance evaluation by including the impact due to CNFET-specific parameters and CNT variations. 1. Ideal case (CNTs variations are not considered): During our research on CNFET-based circuits, we analyzed the impact of CNFET specific parameters, such as CNTs count, diameter and spacing between tubes, on the performance of CNFET-based circuits. The screening effect is critical to take into account for accurate performance evaluation. Hence, PALE model is developed by extending LE formulation to include influence of CNFET specific parameters. 2. Realistic case (CNTs variations are considered): We have studied CNFET-based logic gates and circuits in the presence of major CNTs variations using Monte Carlo simulations. The removal of the initially present unwanted metallic tubes, by the known processing techniques, causes non-uniformity of CNT density in the channel. Such variations in the number of CNTs impact circuit performance and functional yield. We develop variation-aware model (PAPF) based on LE technique to include impact of CNTs variations on the delay of large CNFET-based circuits. Our developed models are correlated with SPICE simulations using different types of gates and circuits with an average error of 3% and 5% for ideal and realistic cases respectively. Our framework is capable of estimating performance more than 100x faster as compared to SPICE simulations methods. Furthermore, using our models (PALE and PAPF), we present an optimization tool to minimize the area and delay product (ADP) of CNFET circuits. We deploy circuit-level techniques (CLT) prior to the optimizing the tubes (CNTs) in the logic gates to achieve highly optimized solution with global approach. For better optimization of the circuits, the impact of wire parasitic in estimating the delay of the individual gates is included as well. Our optimization tool results in maximum and average delay improvement by 27% and 17% respectively, and 2.5X reduction in area for standard ISCAS and OpenSPARC benchmark circuits. Fast and fairly accurate delay computation in our optimization framework offers great runtime benefits as compared to state-of-the-art SPICE simulation and statistical-based methods. Finally, we propose more accurate probabilistic model for yield estimation which incorporates the impact of screening effect on the functional yield after the removal of metallic tubes. Overall, the objective of this thesis is to develop comprehensive LE-based framework and optimization tool and methodology which comprehend CNFET specific parameters for accurate performance evaluation as well as estimation of delay, power, functional yield and do ADP optimization in presence of CNTs variations. Our models are easily scalable to future technology nodes.

The goal of this study was to develop two computer-controlled visual evoked potential (VEP) stimulators. The first device employs a 12 x 12 matrix of 5 mm square Light Emitting Diodes (LEDs) and is housed in an 8 x 8 utility box structure. The second device employs two 8x8 matrices of 3 mm square LEDs, each housed in one eyepiece of a goggle-like structure. A quantitative comparison of the performance of these stimulators was carried out in terms of absolute and interpeak latencies, signal-to-noise ratio (SNR) and cross-correlation between sequential responses obtained from them. Six normal adult subjects were involved in the comparison. Data were acquired from monocular full-field stimulation. The comparison emphasizes potential advantages of the newer, goggle- mounted stimulator.

No description available.

No description available.

A novel method called LOGISMOS - Layered Optimal Graph Image Segmentation of Multiple Objects and Surfaces - for simultaneous segmentation of multiple interacting surfaces belonging to multiple interacting objects is reported. The approach is based on representation of the multiple inter-relationships in a single n-dimensional graph, followed by graph optimization that yields a globally optimal solution. Three major contributions for LOGISMOS are made and illustrated in this thesis: 1) multi-object multi-surface optimal surface detection graph design, 2) implementation of a novel and reliable cross-object surface mapping technique and 3) pattern recognition-based graph cost design. The LOGISMOS method's utility and performance are demonstrated on a knee joint bone and cartilage segmentation task. Although trained on only a small number of nine example images, this system achieved good performance as judged by Dice Similarity Coefficients (DSC) using a leave-one-out test, with DSC values of 0.84+-0.04, 0.80+-0.04 and 0.80+-0.04 for the femoral, tibial, and patellar cartilage regions, respectively. These are excellent values of DSC considering the narrow-sheet character of the cartilage regions. Similarly, very low signed mean cartilage thickness errors were observed when compared to manually-traced independent standard in 60 randomly selected 3D MR image datasets from the Osteoarthritis Initiative database - 0.11+-0.24, 0.05+-0.23, and 0.03+-0.17 mm for the femoral, tibial, and patellar cartilage thickness, respectively. The average signed surface positioning error for the 6 detected surfaces ranged from 0.04+-0.12 mm to 0.16+-0.22 mm, while the unsigned surface positioning error ranged from 0.22+-0.07 mm to 0.53+-0.14 mm. The reported LOGISMOS framework provides robust and accurate segmentation of the knee joint bone and cartilage surfaces of the femur, tibia, and patella. As a general segmentation tool, the developed framework can be applied to a broad range of multi-object multi-surface segmentation problems. Following the LOGISMOS-based cartilage segmentation, a fully automated meniscus segmentation system was build using pattern recognition technique. The leave-one-out test for the nine training images showed very good mean DSC 0.80+-0.04. The signed and unsigned surface positioning error when compared to manually-traced independent standard in the 60 randomly selected 3D MR image datasets is 0.65+-0.20 and 0.68+-0.20 mm respectively.

In medical image registration the goal is to find point by point correspondences between a source image and a target image such that the two images are aligned. There are rigid and non-rigid registration algorithms. Rigid registration uses rigid transformation methods which preserve distances between every pair of points. Non-rigid registration uses transformation methods that do not have to preserve the distances. Image registration has many medical applications -tracking tumors, anatomical changes over time, differences between characteristics like age and gender, etc. A gold standard transformation to compare and evaluate the registration algorithms would be ideal to use to verify if the two images are perfectly aligned. However, there is hardly if ever a gold standard transformation for non-rigid registration algorithms. The reason why there is no gold standard transformation for non-rigid registration algorithms is that pointwise correspondence between two registered points is not unique. In the absence of a gold standard various evaluation methods are used to gauge registration performance. However, each evaluation method only evalutes the error in the transformation from a limited perspective and therefore has its advantages and drawbacks. The Non-Rigid Image Registration Evaluation Project (NIREP) was was created to provide one central tool that has a collection of evaluation methods to perform the evaluations on non-rigid image registration algorithms and rank the registration algorithms based on the outputs of the evaluation methods in the absence of without having to use a gold standard.

The main goal of this research is to design a snake-like robot arm to provide control of a cardiac catheter for use in endovascular aortic repair that is small, cheap, and easy to use. This will help increase the number of aortic aneurysms eligible for endovascular repair and make the procedure simpler and safer for both the patient and the operator. The arm surrounds the catheter and is comprised of two joints that can independently move in any direction giving the operator the ability to easily navigate complicated paths and to control the arm remotely. The arm is controlled by Flexinol actuator wire which is comprised of a nickel titanium alloy that contracts when heated. This allows the arm to be controlled electrically by sending current through the actuator wire thereby heating it. The level of current can be controlled using a microcontroller to generate a pulse width modulated signal to vary the average current. The arm can then be controlled remotely by an operator.

This thesis work describes a technique to estimate local lung tissue expansion using non-rigid inverse consistent image registration on multiple respiratory-gated CT images of lung acquired at different inflation levels. The information of local lung tissue expansion is revealed by the Jacobian of transformation, intensity-based specific volume change, and strain resulted from the registration displacement field. Comparisons between the registration-derived regional lung expansion estimates and xenon CT-derived regional ventilation measure on six sheep at supine position were made to validate the estimates based on registration and find the suitable inflation level change to characterize the lung expansion using this technique.

Matrix factorization algorithms are emerging as popular tools in many applications, especially dictionary learning method for recovering biomedical image data from noisy and ill-conditioned measurements. We introduce a novel dictionary learning algorithm based on augmented Lagrangian (AL) approach to learn dictionaries from exemplar data and it can be extended to general matrix factorization problems due to different constraints. Specically, we use the alternating minimization strategy to decouple the dictionary learning scheme into three main subproblems, which can be solved efficiently. The proposed algorithm can accommodate arbitrary priors on the dictionary, which enables us to inject prior information into the learning process. We validate the algorithm using simulated data and demonstrate its utility in the context of denoising. Comparisons with existing methods show a considerable speedup over other methods. More importantly, we observe that the proposed algorithm is able to recover the dictionaries correctly, even at high sparsity levels and is relatively insensitive to initialization.

Cortical bone is an osseous tissue forming the cortex in our skeleton that supports and protects skeletal functions. Cortical bone segmentation is usually the first step for quantitative cortical bone imaging research. Quality of cortical bone segmentation is one of the most critical factor in determining effectiveness and usefulness of cortical bone measures in a bone imaging study aimed at understanding disease effects, fracture risk and or interventional outcomes. Previous methods primarily focus on local image features and ignore ad therefore fail to utilize larger geometric and topologic contextual knowledge into the segmentation algorithm. Such methods often results in compromised performance under in vivo imaging conditions suffering from low signal to noise ratio and low spatial resolution leaving significant partial volume effects. This thesis presents a new cortical bone segmentation method that utilizes larger contextual and topologic knowledge of distal tibia bone through fuzzy distance transform and connectivity analyses. The input of the method is one threshold and other steps are automatic. An accuracy of 95.1% in terms of percent of volume agreement with gold standard segmentation results and a repeat MD-CT scan intra-class correlation of 98.0% were observed on a cadaveric study. An in vivo study involving sixteen age and body mass index order matched pairs of male and female volunteers has shown that male subjects on average have 16.3% thicker cortex and 4.7% increased porosity as compared to females, and athletes have 3.9% less porosity as compared to control group.