Ida Pu

Correction of High-Order Eddy Current Induced Geometric Distortion in Diffusion-Weighted Echo-Planar Images

Diffusion‐weighted images acquired with the echo‐planar imaging technique are highly sensitive to eddy current induced geometric distortions that vary with the magnitude and direction of the diffusion sensitizing gradients. Such distortions cause misalignment of images acquired with different diffusion strengths and orientations. This in turn can result in errors when calculating maps of the apparent diffusion coefficient and diffusion tensor. Previous correction methods either require separate calibration data or only deal with low‐order errors. In this study, we demonstrate a method that can correct for higher‐order errors. The method relies on collecting pairs of images with diffusion sensitizing gradients reversed. This paired data are first corrected for shifts and linear distortion and then combined to cancel higher‐order errors. All acquired data contribute to the final results. The method has been tested by simulation, on phantoms, on adult volunteers, and on neonatal brain examinations. Magn Reson Med 52:1184–1189, 2004.

Enhanced Blocking Expanding Ring Search in Mobile Ad Hoc Networks

We introduce BERS*, an enhanced Blocking Expanding Ring Search (BERS) protocol for route discovery in mobile ad hoc networks (MANETs). BERS is an energy efficient alternative that was developed recently based on the Expanding Ring Search (ERS). ERS is widely applied in reactive routing protocols such as DSR and AODV. BERS* is a faster and more energy-time efficient version of BERS. It reduces the route discovery latency of BERS by nearly half while maintainning a similar level of energy saving. Our results show that, among the three protocol schemes (BERS*, BERS and ERS), BERS* incurs the least search latency when the hop number of the route nodes is greater than 3, and has achieved the best performance in terms of energy-time efficiency when the hop number of the route nodes is greater than 7. We have also discovered the conditions that allow collective optimisation of BERS* and ERS.

Quantification of venous vessel size in human brain in response to hypercapnia and hyperoxia using magnetic resonance imaging

Hypercapnia and hyperoxia give rise to vasodilation and vasoconstriction, respectively. This study investigates the influence of hypercapnia and hyperoxia on venous vessel size in the human brain. Venous vessel radii were measured in response to hypercapnia and hyperoxia. The venous vessel radii were determined by calculation of the changes in R2* and R2 that are induced by breathing 6% CO2 or pure oxygen. The experimental paradigm consisted of two 3‐min intervals of inhaling 6% CO2 or 100% O2 interleaved with three 2‐min intervals of breathing air. Hypercapnic and hyperoxic experiments were performed on eight subjects on a 3T scanner. Parametric maps of mean venous vessel radius were calculated from the changes in R2* and R2, which were measured by simultaneous acquisition of gradient‐echo and spin‐echo signals. The mean venous vessel radii in hypercapnia were 7.3 ± 0.3 μm in gray matter and 6.6 ± 0.5 μm in white matter. The corresponding vessel radii in hyperoxia were 5.6 ± 0.2 μm in gray matter and 5.4 ± 0.2 μm in white matter. These results show that the venous vessel radius was larger in hypercapnia than that in hyperoxia in both gray matter and white matter (P < 0.005), which agrees with the hypothesis that hypercapnia causes vasodilation and hyperoxia induces vasoconstriction. Magn Reson Med, 2013.

Improving time-efficiency in blocking expanding ring search for mobile ad hoc networks

We propose a new strategy for reducing the amount of latency and energy consumption in Blocking Expanding Ring Search (BERS) and enhanced Blocking Expanding Ring Search (BERS*) for mobile ad hoc networks (MANETs). BERS and BERS* are respectively energy and energy-time efficient route discovery protocols for MANETs as compared to conventional Expanding Ring Search (ERS). In this study, we identify unnecessary waiting time caused by a STOP/END instruction in BERS/BERS* and explore the potential of further improvement of their time efficiency. This leads to tBERS and tBERS*, the improved BERS and BERS* respectively. In tBERS/tBERS*, a route node may also issue the STOP/END instruction to terminate flooding. We implement this idea in algorithms, conduct analysis, and achieve further latency reduction in both tBERS and tBERS* as well as the energy saving in tBERS*.

Analytical Studies of Energy–Time Efficiency of Blocking Expanding Ring Search

This paper investigates the energy–time efficiency of the Blocking Expanding Ring Search algorithms (BERS) using an extended graph model. BERS is an energy efficient alternative that was developed recently based on the Expanding Ring Search (ERS). ERS is widely applied in reactive routing protocols for mobile ad hoc networks. Most studies in this area focus on minimising either energy consumption or search latency, but few look into the strong correlation between the energy saving and the increased latency incurred. We analyse, in this study, the concurrency mechanism of BERS and have developed BERS*, an enhanced scheme based on BERS. Our results show that, among the three schemes (BERS*, BERS and ERS), BERS* incurs the least latency when the hop number of the route nodes is greater than 3, and has achieved the best performance in terms of energy–time efficiency when the hop number of the route nodes is greater than 7. We have also discovered the conditions that allow collective optimisation of BERS* and ERS.

Measuring Energy-Time Efficiency of Protocol Performance in Mobile Ad Hoc Networks

This paper introduces two new metrics for assessment of mobile ad hoc network performance in terms of energy-time efficiency. The combined effect of both energy and time consumption is considered and represented in mathematical terms. The measures have demonstrated a number of advantages over the conventional ones in which the energy and time were often considered separately. The proposed new metrics are simple, generic and flexible. As an application, we have compared the energy-time efficiency of Blocking Expanding Ring Search (BERS) and Expanding Ring Search (ERS), two similar Time to Live (TTL)-based expanding ring search algorithms using our new metrics. The results show that the new metrics can be applied efficiently in assessment of different protocols.

Energy-time efficiency of two routing strategies with chase packets in expanding ring search

We investigate the energy-time efficiency of two reactive route discovery protocols (BERS and DTC) with chase packets in the expanding ring search (ERS). BERS introduces deliberately waiting time in broadcasts of RREQs. DTC uses a threshold to switch between two different channel speeds to allow the chase packets to catch up with fulfilled RREQs. Taking an analytical approach, we gain the insight into these protocols and explore the trade-off nature of the energy and time efficiency. An optimal threshold for DTC is identified, and the performance of DTC under this optimal condition is used to compare with the performance of BERS in the light of ERS. Our results have practical values and implications as well as theoretical interests. We found that BERS has achieved the best overall performance in terms of energy-time efficiency, while DTC achieved the least overall performance despite the reduction of latency.

Gray matter nulled and vascular space occupancy dependent fMRI response to visual stimulation during hypoxic hypoxia

Two cerebral blood volume (CBV)-weighted fMRI techniques, gray matter nulled (GMN) and vascular space occupancy (VASO)-dependent techniques at spatial resolution of 2 × 2 × 5 mm(3), were compared in the study investigating functional responses in the human visual cortex to stimulation in normoxia (inspired O(2) = 21%) and mild hypoxic hypoxia (inspired O(2) = 12%). GMN and VASO signals and T(2)* were quantified in activated voxels. While the CBV-weighted signal changes in voxels activated by visual stimulation were similar in amplitude in both fMRI techniques in both oxygenation conditions, the number of activated voxels during hypoxic hypoxia was significantly reduced by 72 ± 22% in GMN fMRI and 66 ± 23% in VASO fMRI. T(2)* prolonged in GMN and VASO activated voxels in normoxia by 1.6 ± 0.5 ms and 1.7 ± 0.5 ms, respectively. In hypoxia, however, T(2)* shortened in GMN-activated voxels by 0.7 ± 0.6 ms (p < 0.001 relative to normoxia), but prolonged in VASO-activated ones by 1.1 ± 0.6 ms (p < 0.05 relative to normoxia). The data show that the hemodynamic responses to visual stimulation were not affected by hypoxic hypoxia, but T(2)* increases by both CBV-weighted fMRI techniques were smaller in activated voxels in hypoxia. The mechanisms influencing GMN fMRI signal in both oxygenation conditions were explored by simulating effects of the oxygen extraction fraction (OEF) and partial voluming with cerebral spinal fluid (CSF) and white matter in imaging voxels. It is concluded that while GMN fMRI data point to increased, rather than decreased OEF during visual stimulation in hypoxia, partial voluming by CSF is likely to affect the CBV quantification by GMN fMRI under the experimental conditions used.

A framework for chase strategies in recent energy or time efficient route discovery protocols for MANETs

We investigate the energy-time efficiency of the use of chase packets for a number of recently developed energy or time efficient protocols based on the expanding ring search (ERS) for MANETs, namely LB, oDTC, cBERS, BERS and BERS*. A framework is developed based on our analytical studies and simulation findings. We focus on the level of energy consumption, the search latency and the energy-time efficiency, and analyse the performance of these protocols for route discovery. Different strategies are compared in their mathematical models and in reference to ERS as a standard. Our results show that BERS* has achieved the optimal performance of this category of protocols in terms of energy-time efficiency. The framework and our results have practical implications on protocol design and analysis, as well as theoretical interests.

Predicting First-Episode Psychosis Associated with Cannabis Use with Artificial Neural Networks and Deep Learning

In recent years, a number of researches started to investigate the existence of links between cannabis use and psychotic disorder. More recently, artificial neural networks and in particular deep learning have set a revolutionary wave in pattern recognition and machine learning. This study proposes a novel machine learning approach based on neural network and deep learning algorithms, to developing highly accurate predictive models for the onset of first-episode psychosis. Our approach is based also on a novel methodology of optimising and post-processing the predictive models in a computationally intensive framework. A study of the trade-off between the volume of the data and the extent of uncertainty due to missing values, both of which influencing the predictive performance, enhanced this approach. Furthermore, we extended our approach by proposing and encapsulating a novel post-processing k-fold cross-testing method in order to further optimise, and test these models. The results show that the average accuracy in predicting first-episode psychosis achieved by our models in intensive Monte Carlo simulation, is about 89%.