PubMed Health⌕ Search

Biomedical subjects

P J Edwards

Publications and source records attributed to P J Edwards.

18 recordsLinked to original sources
PubMed2000

Locking acrylic resin dental stent for image-guided surgery.

This article presents a procedure for fabricating a locking acrylic resin dental stent for use in image-guided base-of-skull surgery and neurosurgery. The stent offers advantages over conventional bone screw-anchored systems to surgeons and patients. In view of the increasing use of image guidance in base-of-skull surgery and neurosurgery, prosthodontists will meet a growing demand for this type of device in the future.

Acrylic Resins↗
PubMed2000

A study of early stopping and model selection applied to the papermaking industry.

This paper addresses the issues of neural network model development and maintenance in the context of a complex task taken from the papermaking industry. In particular, it describes a comparison study of early stopping techniques and model selection, both to optimise neural network models for generalisation performance. The results presented here show that early stopping via use of a Bayesian model evidence measure is a viable way of optimising performance while also making maximum use of all the data. In addition, they show that ten-fold cross-validation performs well as a model selector and as an estimator of prediction accuracy. These results are important in that they show how neural network models may be optimally trained and selected for highly complex industrial tasks where the data are noisy and limited in number.

Algorithms↗
PubMed1999

A system for microscope-assisted guided interventions.

We present a system for surgical navigation using stereo overlays in the operating microscope aligned to the operative scene. This augmented reality system provides 3D information about nearby structures and offers a significant advancement over pointer-based guidance, which provides only the location of one point and requires the surgeon to look away from the operative scene. With a previous version of this system, we demonstrated feasibility, but it became clear that to achieve convincing guidance through the magnified microscope view, a very high alignment accuracy was required. We have made progress with several aspects of the system, including automated calibration, error simulation, bone-implanted fiducials and a dental attachment for tracking. We have performed experiments to establish the visual display parameters required to perceive overlaid structures beneath the operative surface. Easy perception of real and virtual structures with the correct transparency has been demonstrated in a laboratory and through the microscope. The result is a system with a predicted accuracy of 0.9 mm and phantom errors of 0.5 mm. In clinical practice errors are 0.5-1.5 mm, rising to 2-4 mm when brain deformation occurs.

Bone Cysts↗
PubMed1999

Stereo augmented reality in the surgical microscope.

We present an augmented reality system that allows surgeons to view features from preoperative radiological images accurately overlaid in stereo in the optical path of a surgical microscope. The purpose of the system is to show the surgeon structures beneath the viewed surface in the correct 3-D position. The technical challenges are registration, tracking, calibration and visualisation. For patient registration, or alignment to preoperative images, we use bone-implanted markers and a dental splint is used for patient tracking. Both microscope and patient are tracked by an optical localiser. Calibration uses an accurately manufactured object with high contrast circular markers which are identified automatically. All ten camera parameters are modelled as a bivariate polynomial function of zoom and focus. The overall system has a theoretical overlay accuracy of better than 1 mm. Implementations of the system have been tested on seven patients. Recent measurements in the operating room conformed to our accuracy predictions. For visualisation the system has been implemented on a graphics workstation to enable high frame rates with a variety of rendering schemes. Several issues of 3-D depth perception remain unsolved, but early results suggest that perception of structures in the correct 3-D position beneath the viewed surface is possible.

Depth Perception↗
PubMed1998

Toward optimally distributed computation.

This article introduces the concept of optimally distributed computation in feedforward neural networks via regularization of weight saliency. By constraining the relative importance of the parameters, computation can be distributed thinly and evenly throughout the network. We propose that this will have beneficial effects on fault-tolerance performance and generalization ability in large network architectures. These theoretical predictions are verified by simulation experiments on two problems: one artificial and the other a real-world task. In summary, this article presents regularization terms for distributing neural computation optimally.

Computer Simulation↗
PubMed1998

A three-component deformation model for image-guided surgery.

In image-guided surgery it is necessary to align preoperative image data with the patient. The rigid-body approximation is usually applied, but is often not valid due to tissue deformation. Non-rigid deformation algorithms have been applied to related, but not identical problems, such as atlas matching and surgery simulation. In image-guided surgery we have the additional information that the deformation is constrained by the physical properties of the different tissues. The most important properties that must be incorporated are the rigidity of bone, the unconstrained nature of fluid regions and the relatively smooth deformation of soft tissue. Hence, we have developed a simplified model of tissue deformation based on a three-component system. Rigid regions are constrained by the rigid-body transformation and fluid regions are unconstrained. A number of energy models for deformable tissues have been compared. The model can be deformed using intraoperative data, in this case landmarks, using a technique similar to active contours. A novel strategy to avoid folding in the transformation is described. Our method was applied to MRI and CT data from a neurosurgery patient with epilepsy. Although the current implementation is only two dimensional, the initial results are promising. As the algorithm must ultimately run in or near 'real-time' an improved implementation of the energy minimization is underway. This paper presents the problem of tissue deformation, which has received little attention in the literature and outlines the framework we have developed for tackling this difficult subject.

Algorithms↗
PubMed1995

Augmentation of reality using an operating microscope for otolaryngology and neurosurgical guidance.

The operating microscope is an integral part of many neurosurgery and otolaryngology procedures; the surgeon often uses the microscopic view for a large portion of the operation. Information from preoperative radiological images is often viewed only on X-ray films. The surgeon then has the difficult task of relating this information to the appearance of the surgical view. Image guidance techniques attempt to relate these two sets of information by registering the patient in the operating room to preoperative images using locating devices. Conventionally, image data are presented on a computer monitor, which requires the surgeon to look away from the operative scene. We describe a guidance system, for procedures in which the operating microscope is used, which super-imposes image-derived data upon the operative scene. We create a model of relevant structures (e.g., tumor volume, blood vessels, and nerves) from multimodality preoperative images. By calibrating microscope optics, registering the patient to image coordinates, and tracking the microscope and patient intraoperatively, we can generate stereo projections of the three-dimensional model and project them into the microscope eyepieces, allowing critical structures to be overlaid on the operative scene in the correct position. Measurements with a head phantom gave a root mean square (RMS) error of 1.08 mm, and the estimated error for a human volunteer is between 2 and 3 mm. Initial evaluation in the operating room was very promising.

Computer Simulation↗
PubMed1995

Can deterministic penalty terms model the effects of synaptic weight noise on network fault-tolerance?

This paper investigates fault tolerance in feedforward neural networks, for a realistic fault model based on analog hardware. In our previous work with synaptic weight noise we showed significant fault tolerance enhancement over standard training algorithms. We proposed that when introduced into training, weight noise distributes the network computation more evenly across the weights and thus enhances fault tolerance. Here we compare those results with an approximation to the mechanisms induced by stochastic weight noise, incorporated into training deterministically via penalty terms. The penalty terms are an approximation to weight saliency and therefore, in addition, we assess a number of other weight saliency measures and perform comparison experiments. The results show that the first term approximation is an incomplete model of weight noise in terms of fault tolerance. Also the error Hessian is shown to be the most accurate measure of weight saliency.

Neural Networks, Computer↗
PubMed1993

Pulse stream VLSI circuits and systems: the EPSILON neural network chipset.

An analogue CMOS VLSI neural processing chip has been designed and fabricated. The device employs "pulse-stream" neural state signalling and is capable of computing some 360 million synaptic connections per second. In addition to basic characterisation results, the performance of the chip in solving "real-world" problems is also demonstrated. The experience gained from the development of this device has resulted in the design of a second "pulse-stream" chip with improved performance and features. It is anticipated that this second device will be integrated into a standard bus-based system and find early application in robotic control.

Computers, Analog↗
PubMed1993

Analogue synaptic noise--implications and learning improvements.

We analyse the effects of analogue noise on the synaptic arithmetic during multilayer perceptron training by expanding the cost function to include noise-mediated penalty terms. Predictions are made in the light of these calculations which suggest that fault tolerance, generalisation ability and learning trajectory should be improved by such noise-injection. Extensive simulation experiments on two distinct classification problems substantiate the claims. The results appear to be perfectly general for all training schemes where weights are adjusted incrementally, and have wide-ranging implications for all applications, particularly those involving "inaccurate" analogue neural VLSI.

Algorithms↗
PubMed1987

Application of the GTGO-AD procedure in the scanning electron microscopy of streptococci adhered to pellicle-coated dental enamel.

The scanning electron microscope (SEM) offers a direct and visually effective means of examining streptococci adhering to pellicle-coated dental enamel. The GTGO-AD (Glutaraldehyde-Tannic acid-Guanidine hydrochloride-Osmium tetroxide-Air-Drying procedure) has the advantage that critical point-drying or freeze-drying need not be employed. It was found that both the structural and spatial integrity of the adherent streptococci and the pellicle were maintained with the GTGO-AD procedure.

Bacterial Adhesion↗