New Publications are available for Medical magnetic resonance imaging and spectroscopy
http://dl-live.theiet.org
New Publications are available now online for this publication.
Please follow the links to view the publication.A new mutual information based similarity measure for medical image registration
http://dl-live.theiet.org/content/conferences/10.1049/cp.2012.0424
Medical image registration (IR) is the systematic process of aligning spate images, often involving different modalities with common reference framework, so complementary information can be combined and compared. This paper presents a new similarity measure which uses Expectation Maximization for Principal Component Analysis allied with mutual information (EMPCA-MI) for medical IR. The new measure has been analysed on multimodal, three band magnetic resonance images (MRI) T1, T2 and PD weighted, in the presence of both intensity non-uniformities (INU) and noise. Both quantitative and qualitative experimental results clearly demonstrate both improved robustness and lower computational complexity of the new EMPCA-MI paradigm compared with existing MI-based similarity measures, for various MRI test datasets. (6 pages)Validating the neutrosophic approach of MRI denoising based on structural similarity
http://dl-live.theiet.org/content/conferences/10.1049/cp.2012.0419
This paper focuses on validating the proposed Neutrosophic Set (NS) approach of Magnetic Resonance Image (MRI) denoising based on structural similarity such as Structural Similarity Index (SSIM) and Quality Index based on Local Variance (QILV). The Neutrosophic Set approach of median filter is used to reduce the Rician noise in MR image. The experiments have conducted on real MR image with Rician noise added. The visual and the diagnostic quality of the denoised image is well preserved. The performance of this filter is compared with median filter and non local mean filter (NLM). (6 pages)MRI mammogram image classification using ID3 algorithm
http://dl-live.theiet.org/content/conferences/10.1049/cp.2012.0464
Breast cancer is one of the most common forms of cancer in women. In order to reduce the death rate , early detection of cancerous regions in mammogram images is needed. The existing system is not so accurate and it is time consuming. The Proposed system is mainly used for automatic segmentation of the mammogram images and classify them as benign,malignant or normal based on the decision tree ID3 algorithm. A hybrid method of data mining technique is used to predict the texture features which play a vital role in classification. The sensitivity, the specificity, positive prediction value and negative prediction value of the proposed algorithm accounts to 93.45% , 99.95%,94% and 98.5% which rates very high when compared to the existing algorithms. The size and the stages of the tumor is detected using the ellipsoid volume formula which is calculated over the segmented region. (5 pages)Epicardial contour segmentation by using level set with prior shape and region
http://dl-live.theiet.org/content/conferences/10.1049/cp.2011.0993
A segmentation model is proposed for the epicardial contour of the cardiac MRI image based on level set with prior shape and region. Considering the shape correlation of the left ventricle (LV) in the cardiac image sequences, the segmentation result of the current frame is referred as the prior shape of the next frame and the contraction and diastolic change area of LV is exploited as the prior region. The initial evolving curve is automatically located by calculating the posterior probability of the pixel belonging to the chamber of LV. The prior shape and region prevent the evolving curve from leaking off the outer contour and pull it to the accurate edges. Promising experimental results are obtained on the real cardiac sequences.Segmentation of meningioma MR images using SVM and RBF algorithms
http://dl-live.theiet.org/content/conferences/10.1049/ic.2009.0152
The paper explores the field of medical image analysis through the use of ANN algorithms. The processing is carried out on MRI images which would be helpful in detecting brain tumor tissues through the revelation of infected tissues in the image. After the preprocessing of image, features are extracted from spatial domain and are separately fed to the algorithms. This process requires calculation of co-occurrence matrices of the template (size 3×3) from the image. The outputs obtained after these processes are subjected to segmentation technique. The comparison is done by determining the sensitivity and the specificity of the segmented image with reference to the manually segmented image. The algorithms implemented are Radial Basis Function (RBF) and Support Vector Machine (SVM). The important step in RBF is to determine centroids and width of the clusters. For this purpose, k-means clustering is used. The basis function used for RBF is Gaussian and for learning backpropagation is used. In SVM, kernel function used is linear function. (5 pages)Self organizing feature maps for the fiber tracking in diffusion tensor MR images
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080466
Self-Organizing Maps (SOM) can be utilized for the purpose of discovering underlying structures in a distributed system. In the proposed study, SOM is employed to track neuronal pathways in brain Diffusion Tensor Magnetic Resonance (DTMR) images. DTMR Imaging (DTMRI) provides local directional information along the nerve fiber bundles in a vector form. The directional information is obtained by solving for the eigensystem of the tensors produced by the diffusion of the water molecules during the imaging process in MRI. The idea is to map a network of connected nodes of a SOM structure onto the investigated neuronal tracks. First, the imaging matrix is simulated with random eigenvectors and SOM is implemented and verified on these synthetic diffusivity tracks. Then the analysis follows on real DTMR images. The aim of the study is to propose an alternate method for the DTMRI tractography with SOM. Preliminary results on two dimensional synthetic diffusion images are encouraging to pursue the idea in 3D real images. In this study we discuss the proposed method in detail. (4 pages)Nasopharyngeal carcinoma lesion extraction using clustering via semi-supervised metric learning with side-information
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080373
In this paper, we consider the extraction of nasopharyngeal carcinoma lesion from magnetic resonance images as a clustering problem. The metric used by the clustering algorithm in our proposed method is a new spatially weighted metric, which is learned by semi-supervised metric learning with side-information. Several experiments have been conducted to compare the performance of the proposed metric with similar metrics for the tumor extraction.Obtaining DSC MRI cerebral blood flow estimates without tissue specific errors
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080456
The singular value decomposition (SVD) deconvolution implementation is in common use in magnetic resonance (MR) dynamic susceptibility contrast (DSC) studies. The zdSVD SVD variant involves computationally manipulating DSC concentration signals to have zero arterial-tissue delay (ATD) prior to using SVD. Our proposed zdSVD improvements show how considering other signal time shifts leads to cerebral blood flow estimates whose accuracy shows a minimal dependency on the tissue mean transit times (MTT) values. This characteristic leads to greater absolute CBF accuracy across all MTT values after scaling MR studies to match PET flow values. A comparison is made between the modified zdSVD algorithm and other DSC-specific deconvolution algorithms. (4 pages)Signal-to-noise ratio improvement of cardiac magnetic resonance spectroscopy signals acquired by phased array coils: a simulation based approach
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080445
Magnetic resonance spectroscopy (MRS) is a non-invasive technique for obtaining in vivo biochemical information. Since the amplitude of the peaks in magnetic resonance (MR) spectra is proportional to the metabolite concentrations, obtaining the best signal-to-noise ratio (SNR) is fundamental for the accurate quantification of the metabolites. New acquisition strategies for the improvement of the intrinsic low SNR of MRS signals have been designed, without increasing the examination time. These approaches are based on the use of multiple receiving coils, called phased array coils. In this context one of the main challenges is to determine the best combination of the acquired signals that optimize the resulting SNR. This paper describes a novel method for the combination of MR signals acquired by phased array coils, even in presence of correlated noise between the acquisition channels. Performance evaluation is carried out on simulated <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H-MRS signals and experimental results are obtained on in vivo cardiac <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H-MR spectra. (4 pages)Automatic segmentation of left ventricular myocardium tissue in series of full short axis MRI
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080467
We proposed a new method for computer-based segmentation of myocardium tissue of left ventricular in series of full short axis MRI. First we locate the cardiac region in full short axis MRI data and then left ventricular cavity is searched and segmented in that region. Using this new approach, great result has been obtained in determination of left ventricular cavity. After finding the left ventricular cavity, the myocardium tissue is separated from other parts using a dilating circle starting from the left ventricular cavity. Separating myocardium tissue in short axis MRI data may lead to a very good analytic tool for interpretation of heart problems and analyzing left ventricular. In addition to preciseness, the proposed method is fully adaptive and has no dependence on a particular image. Beside it is deploys computational effective algorithm which results in an easy and direct implementation using a simple DSP hardware. (4 pages)A geographic network for MRI assessment of iron overload in thalassemia: the MIOT experience
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080458
Today there are about 6000 thalassemia patients in Italy, with a marked prevalence in the south, Sardinia and Sicily. Cardiac magnetic resonance (CMR) measurement of myocardial iron overload allows early diagnosis and treatment of these patients. To optimize patient care, CMR centers should be easily accessible for the patients and should be able to perform iron overload measurement in a standardized way. To accomplish this objective, an analysis system was developed and distributed to six CMR centers. Fifty haematological and paediatric centers specialized in thalassemia care were also involved. All centers are linked by a web-based network, configured to collect and share patient data, assuring personal data protection. The inter-center variability of the proposed methodology was evaluated, demonstrating the effectiveness of the network. (4 pages)Realistic head model preparation for EEG forward problem
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080470
Electroencephalography (EEG) forward solution computes the electrical potentials over the electrodes that are lying on the scalp surface. Forward solution requires a numerical head model to work on. The head model can be spherical or a realistic one. Accuracy of the forward problem can be improved by using realistic head models. Triangulated realistic head model is an input for the Boundary Element method (BEM) which solves the EEG forward problem. In this paper, a software is introduced which constructs 3-D triangulated realistic head models from the Magnetic Resonance images. (4 pages)Tractography of corpus callosum connections to other brain structures
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080468
A microstructural damage of corpus callosum (cc) is demonstrated by histological post-mortem study in many neurodegenerative diseases, while at present the only in-vivo investigation is provided by DTI and tractography. However, the lesional load which often accompanies these pathologies can severely limit the image registration steps necessary for reliable DTI computation. A technical study for specific cc investigation and for the enhancement of main tracts connecting cc and other brain structures is presented, analyzing 5 subjects with different lesional loads. Atlas of 116 brain structures and segmented cc, derived from high-resolution-Tl images, are coregistered onto DTI directly or by 2 passages. For 2 patients with high lesional load, the 2- steps coregistration gives best results, indeed mutual information between Tl and DTI is higher for Tl coregistered with 2 steps than directly. Volumes of entire cc, portions of cc, brain structures connected to cc are computed. Also DTI-derived indices (MD, FA) are computed for the same structures. For all the subjects, the volume of cc anterior portion was less than posterior one; for 4 subjects FA of anterior bundles were less then FA of posteriors. The parameters of fiber tracking allowed to extract bundles connecting cc to other brain structures and to compute indices of their microstructural integrity, within the limits, in terms of acquisition time, of a clinically feasible protocol. (4 pages)Accurate and fast 3D interactive segmentation system applied to MR brain quantification
http://dl-live.theiet.org/content/conferences/10.1049/cp_20080443
This work presents an efficient interactive segmentation system for volumetric data-sets based on advanced 3D morphological analyses and an interaction paradigm that allows a good match with user intentions. This system has been designed to produce accurate results under the complete control of the user, to minimize the interaction time and to address a generality of 3D segmentation tasks. The system has been tested and compared with other softwares on normal MR brain structure quantification and on a challenging clinical setting pointed to the detection of the presence of subtle brain atrophy associated to primitive immunodeficiency (PID). (4 pages)Realistic breast models for second generation tissue sensing adaptive radar system
http://dl-live.theiet.org/content/conferences/10.1049/ic.2007.1310
Radar-based breast cancer detection methods require rigorous testing before clinical application. Therefore, realistic breast models in terms of shape, size and electrical properties are needed. Magnetic resonance images are typically segmented to create these realistic breast models. This paper introduces five breast models of increasing complexity. Tumour reflections from each model are focussed with the TSAR algorithm, and detection results are calculated for each model. (4 pages)Practical issues in hospitals
http://dl-live.theiet.org/content/conferences/10.1049/ic_20060013
The physical implementation of the Physical Agents (EMF) Directive has significant implications for medical MRI. There are a number of clinical situations where occupational exposure to low frequency time varying magnetic fields exceeds the relevant action values, and almost certainly also the exposure limits. Restrictions imposed by the Directive adversely affects current clinical activities and prevent the development of new techniques that could have a significant impact on medical practice and also bring health and safety benefits for both patients and staff. There is no scientific justification for these limits, which are based on cautious interpretation of limited data and overly inflexible incorporation of the resulting guidelines into law.Approaches to validating the "quantity" in quantitative MR cerebral perfusion studies
http://dl-live.theiet.org/content/conferences/10.1049/cp_20060344
Stroke is a major killer across the world; with staggering financial implications. Magnetic resonance dynamic susceptibility contrast studies are a plausible approach to provide quantitative cerebral blood flow (CBF) measurements. However, what is the correct approach to take when a proven method from another scientific area is suggested as a means for a minor increase in CBF accuracy, but leads to results that totally disagree with the established literature? Agile methodologies are being introduced into business as a new "light-footed" approach to developing software with considerable success. We describe our experiences of adapting Agile software engineering practices to establish a repeatable hardware-software co-design framework for biomedical algorithm validation. (4 pages)Automated segmentation of the articular space in MR images of the hip joint
http://dl-live.theiet.org/content/conferences/10.1049/cp_20060581
Segmentation of the cartilages in the hip joint is important in diagnosis of hip diseases. Automatic localization of the hip joint articular space is a major step for finding the central line of this space in this regard. In this study, we propose an automatic technique for segmentation the articular space using clinically obtained multi-slice TI-weighted MR data. To perform an isotropic dataset, we resampled data sets with a modified sine interpolation technique which reduces the effect of unwanted Gibbs ringing. By assuming a spherical shape for the femur, we estimated the center of the femoral head by a Hough transform. Radial derivative filters were then performed to enhance the edges of the cartilages, articular spaces and pelvic bones in the data sets. We localized the articular space by customizing a Canny edge detector to the problem. Finally, we refined the articular space edges by estimating and discarding the bony edges by an Otsu adaptive thresholding technique. We implemented the techniques in C++ and Matlab programming languages. The feasibility of the proposed techniques was successfully evaluated in the presence of 40 hips including 1200 MR images.Segmentation of brain MR images using fuzzy sets and modified co-occurrence matrix
http://dl-live.theiet.org/content/conferences/10.1049/cp_20060551
Image segmentation is an indispensable process in the visualization of human tissues, particularly during clinical analysis of magnetic resonance (MR) images. A robust segmentation technique based on fuzzy set theory for brain MR images is proposed in this paper. The histogram of the given image is thresholded according to the similarity between gray levels. The similarity is assessed through second order fuzzy correlation. To calculate the second order fuzzy correlation, a modified co-occurrence matrix is used to extract the local information more accurately. Two parameters - ambiguity and the strength of ambiguity, are introduced to determine the thresholds of the given histogram. The effectiveness of the proposed algorithm, along with a comparison with other methods, has been demonstrated on a set of brain MR images.Quantitative analysis and comparison of diffusion tensor imaging tractography algorithms
http://dl-live.theiet.org/content/conferences/10.1049/cp_20060421
This paper describes a platform developed for analysis of diffusion tensor image data using different tractography algorithms. The platform allows these algorithms to be compared both quantitatively and qualitatively. Two specific tractography algorithms were compared, STT (streamlines tracking technique) and TEND (tensor-deflection algorithm). The platform was assessed on a publicly available DTI dataset to analyse and quantify the performance of these tractography algorithms. Based on specific tests using this platform, results indicate that the STT algorithm is better at dealing with fibres containing curves and TEND is more appropriate for straighter fibres. A methodology is also proposed to help differentiate between nerve fibres that meet or cross.Automatic tumor segmentation using optimal texture features
http://dl-live.theiet.org/content/conferences/10.1049/cp_20060381
This paper presents an automatic segmentation of malignant tumor in magnetic resonance images (MRI's) of brain using optimal texture features. Texture features are extracted from normal and tumor regions (ROI) in the brain images under study using spatial gray level dependence method and wavelet transform. The normal and tumor regions are classified using an artificial neural network. A very difficult problem in classification techniques is the choice of features to distinguish between classes. In the proposed method, the optimal texture features that distinguish between the brain tissue and malignant tumor tissue is found using genetic algorithm (GA). The optimal features are used to segment the tumor. The proposed feature based segmentation technique is compared with few existing techniques. The performance of the algorithm is evaluated on a series of brain tumor images. The results illustrate that the proposed method outperforms the existing methods. (4 pages)Computer-assisted identification and modelling of internal carotid arteries from 3D CE-MR angiograms for stenosis quantification
http://dl-live.theiet.org/content/conferences/10.1049/ic_20050328
In this study a methodology is presented to identify and quantify the stenosed internal carotid artery from Contrast-Enhanced MR Angiograms. The internal carotid artery automatically identified by tracking the carotid bifurcation and selecting the artery branch with no further arterial branches. An approach is proposed to segment the arteries by applying adaptive thresholds along the central axis of the artery based on 3D cross-sectional intensity profiles. Isosurface models from the original MR data are used to measure the arterial area, in planes perpendicular to the central axis. Experimental results are included to demonstrate the performance of this proposed extraction and modelling tool, giving effective identification of the ICA and enabling consistent and repeatable stenosis quantification.Convergent technologies in personalized healthcare
http://dl-live.theiet.org/content/conferences/10.1049/ic_20050603
A collection of slides from the author's conference presentation is given. (14 pages)The use of LS-SVM in the classification of brain tumors based on <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H-MR spectroscopy signals
http://dl-live.theiet.org/content/conferences/10.1049/ic_20020293
Least Squares Support Vector Machines (LS-SVM) have been developed and successfully applied to classification problems in many areas. In comparison with several other classical methods this technique consistently performs very well on a large variety of problems. Here, results on the application of LS-SVM for classification of brain tumors based on <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H-Magnetic Resonance Spectroscopy (<sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H-MRS) signals are presented. Radial Basis Function (RBF) and linear kernels are used and compared to find the optimal classifier. The improvement of this classification based on MRS signals will lead to an advanced tool for the discrimination of brain tumors, which is presently under development for the INTERPRET project. (5 pages)The feasibility of using medical imaging modalities for in-vivo detecting of physiological effects of electromagnetic radiation in the human brain during the use of cellular phones
http://dl-live.theiet.org/content/conferences/10.1049/ic_19990218
The widespread use of mobile phones has caused a substantial public concern about a possible connection between the RF energy emitted by the phone and adverse effects to the brain. The thermal effect is agreed to be the main demonstrable effect of RF energy on the human body. Despite this fact, it is now accepted the need to at least investigate observations, which do not seem to be linked to the thermal deposition of energy in the human body. One of the important issues in this subject is what really happens in the brain during the use of cellular phones. This can only be answered with an in-vivo experiment. We review some of the medical imaging modalities (MRI, PET, and SPECT) as powerful tools for accurate investigation either anatomically or physiologically. The feasibility of utilising any of these techniques for an in-vivo assessment of the electromagnetic radiation from the cellular phones to the human head (brain) is mentioned. The main objective of this review is to discuss techniques that can detect specific macroscopic physiological effects of cellular phones EM fields, rather than the effects of these fields in the ionic levels and enzymatic responses or regulatory mechanisms of cell growth. (6 pages)Automatic tracking of vortical flow patterns with MR velocity mapping
http://dl-live.theiet.org/content/conferences/10.1049/cp_19990353
This paper describes an automatic method based on the phase portrait theory for studying vortical flow features. The potential of this technique for studying blood flow features using multi-directional magnetic resonance velocity imaging is demonstrated both in normal subjects and in patients. The method relies on detecting critical flow features prior to analysing dynamical indices of the fluid, and therefore is well suited to the topological study of complex flow patterns.Components of brain activity - data analysis for fMRI
http://dl-live.theiet.org/content/conferences/10.1049/cp_19991247
Functional magnetic resonance imaging (fMRI) is a promising method to determine noninvasively the spatial distribution of brain activity in a given situation, e.g. in response to a stimulus or during task solving. The fMRI signal is very small and often cannot be identified from the anatomical images. Thus data analysis methods are required to localize the activity. We discuss different data analysis methods, a simple correlation analysis, principal component analysis (PCA) and independent component analysis (ICA), in the context of a motor task experiment with predefined stimulus time course. We show how it is possible to detect even weak activity without prior knowledge about the stimulus time course with PCA and ICA. The stimulus time course is extracted and major components of the signal, e.g. head movements are also identified.Reduced FOV imaging with motion adaptation
http://dl-live.theiet.org/content/conferences/10.1049/cp_19990373
This paper addresses the use of reduced FOV (field-of-view) imaging with spatially variable resolution for 3-D MR coronary angiography. Details involved in using the technique for reducing imaging time and facilitating the management of respiratory motion are discussed. The technique was evaluated with phantom experiments and right coronary artery images of 11 asymptomatic volunteers using a 0.5 T MR system.Mosaic image segmentation with bias field correction
http://dl-live.theiet.org/content/conferences/10.1049/cp_19990447
This paper presents an automatic method of correcting nonuniform RF coil response for the classification of body composition using MR imaging. By linear mosaic modelling, the smoothly but nonlinearly varying bias field which modulates tissue intensities within the image was corrected. The overlapping between adjacent mosaics ensured consistent segmentation of body fat content and the effectiveness of the technique was validated by both phantom and in vivo experiments. Ten whole body composition data sets, each with 39 trans-axial slices, were acquired. Automatic segmentation results using the proposed technique were compared with those from manual delineations. The automatic segmentation method was found to be highly accurate and the mean percentage error between the two methods was less than 1.5%.KBANNs and the classification of <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">31</sup>P MRS of malignant mammary tissues
http://dl-live.theiet.org/content/conferences/10.1049/cp_19991240
Knowledge-based artificial neural networks (KBANNs) is a hybrid methodology that combines knowledge of a domain in the form of simple rules with connectionist learning. This combination allows the use of small sets of data (typical of medical diagnosis tasks) to train the network. The initial structure is set from the dependencies of a set of rules and it is only necessary to refine these rules by training. Here, the authors present such KBANNs with a topology derived from knowledge elicited from the domain of metabolic features of malignant mammary tissues. KBANN performance is assessed over the classification of 26 in vivo <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">31</sup>P spectra of normal and cancerous breast tissues. Results presented in this paper confirm the suitability of KBANNs a computational aid capable of classifying complex and limited data in a medical domain. The present study is part of an ongoing investigation into normal and abnormal breast physiology which may allow noninvasive early detection of breast cancer.An adaptive gradient-based boundary detector for MRI images of the brain
http://dl-live.theiet.org/content/conferences/10.1049/cp_19990360
Some methods for segmenting MRI data, especially those which employ multi-spectral data, can be very time-consuming. Hence, it is desirable to reduce the dimensionality of the multi-spectral data by eliminating regions which are of no interest. When, for example, the objective is the segmentation of the brain, it is clear that a considerable amount of the image could be discarded without compromising (in some cases even enhancing) the quality of the segmented image. This paper presents an algorithm to isolate the brain from the extra-cranial tissues irrespective of the nature of the MRI brain data (sagittal, coronal or axial slices). Given a starting point, or seed, at the boundary (or near it) between the intra-cranial and extra-cranial region, the algorithm “walks” along the boundary by calculating at each point the direction of the local edge. The gradient operator chosen to calculate the edge direction was the integrated directional derivative gradient (IDDG) operator as defined by Zuniga and Haralick (1987) and is taken from the row and column directional derivatives, D<sub xmlns="http://pub2web.metastore.ingenta.com/ns/">r</sub> and D<sub xmlns="http://pub2web.metastore.ingenta.com/ns/">c</sub>.Automatic motion analysis of bones from MR sequences
http://dl-live.theiet.org/content/conferences/10.1049/cp_19990351
In many cases articular damages cannot be diagnosed through an examination of a single image. A motion analysis of a joint's bones might be necessary to make a reliable diagnosis. Examples are lesions of the ligaments and cartilage of the knee or in the cervical and lumbar regions of the vertebral. This paper presents a novel system to diagnose lesions of the ligaments of the wrist (carpal instabilities). The method is particularly well-suited to aid in the diagnosis of the scapho-lunate instability. This damage is a common injury after accidents involving the wrist. The lesion occurs when the ligaments between the scaphoid and the lunate are torn. Motion graphs show the rotation as well as the translation of the carpal bones. The measurement is performed relative to an anatomic co-ordinate system defined by the distal end of the radius. Compared to other applications a motion analysis of the wrist bones is more difficult because there are many bones with a similar shape which complicates their identification. Furthermore some of the bones may tilt, that is they may rotate around axes not perpendicular to the view plane. This results in a varying appearance of the bones in the sliced magnetic resonance (MR) images.Magnetoencephalography
http://dl-live.theiet.org/content/conferences/10.1049/ic_19980707
When comparing the findings of magnetic source imaging (MSI) with other presurgical evaluations (EEG, MRI and intraoperative ECoG) in temporal lobe epilepsy lobar or even intralobar congruence can be found. The dipolar activity that can be recognized during a spike-wave event is localized in temporal neocortical or mesial regions. Further origin of epileptic activity can be localized by the method of spike averaging. The combination of MEG and MRI helps to build a bridge between morphological and functional localization. For clinical use MSI can serve as a guide for invasive recordings. Additionally it helps to detect functionally important brain regions and can serve as a pointer to discrete lesions in the MRI. (5 pages)Functional imaging of the brain in epilepsy
http://dl-live.theiet.org/content/conferences/10.1049/ic_19980705
In the evaluation of patients with seizure disorders, the integration of structural and functional data is key to the successful formulation of the epilepsy syndrome, and the localization of seizure onset and its aetiology, Current MRI reveals the structural cerebral basis of refractory partial seizures in up to 80% of patients referred for consideration of surgical treatment. The remainder are cryptogenic and it is in this group that functional imaging techniques may be particularly clinically useful. The author discusses the application of positron emission tomography (PET), single photon emission computerised tomography (SPECT), functional MRI and magnetic resonance spectroscopy. (4 pages)Anatomical imaging of the brain
http://dl-live.theiet.org/content/conferences/10.1049/ic_19980704
In this work, the need for anatomical imaging is first discussed. The findings of modern neuroimaging in epilepsy are then considered, with particular reference to the many ways in which MRI data may be helpful in enlarging one's understanding of the structural basis of epilepsy. Some newer analytic techniques and their application are then described, ending with a statement of the current position of imaging of the anatomy of the brain and a contemplation of the needs for the future. (9 pages)Multi-dimension modelling techniques for reciprocal space of NMR imaging
http://dl-live.theiet.org/content/conferences/10.1049/cp_19980472
In this paper the application of multidimensional modelling techniques to a reciprocal space components of a nuclear magnetic resonance (NMR) imaging system is considered. These include a 2-dimensional input/output autoregressive moving average (ARMA) model, Fornasini and Marchesini (1976) state-space model and identifications based on multidimensional artificial neural networks (ANN).High precision current tracking PWM amplifier with optimal type 1 digital controller for generating magnetic field in MRI systems
http://dl-live.theiet.org/content/conferences/10.1049/cp_19980565
This paper presents a two-paralleled PWM amplifier using switch-mode power current tracking technique in order to generate a gradient magnetic field in a magnetic resonance imaging (MRI) system. This circuit has 8-IGBTs at their inputs/outputs so as to realize further high-power density. A digital control scheme can minimize the current ripple and improve its control response in the gradient coils. It is proved that the proposed technique will highly enlarge the diagnostic target and improve the image quality of MRI.MRS: a clinical perspective
http://dl-live.theiet.org/content/conferences/10.1049/ic_19970471
Magnetic Resonance Spectroscopy (MRS) is a method with unique capabilities-it is the only practical, non-invasive way to measure body chemistry and intracellular pH (pHi). Many of the thousands of MR imaging instruments in the world could easily be adapted to perform MRS-it adds about 10% to the capital cost. It is already widely used in clinical research, but the anticipated routine clinical methods have not yet developed. The aims of this lecture are: (i) to introduce MRS to non-specialists; (ii) to explain the biological basis of some of the spectra that are likely to be met in pattern recognition work; and (iii) to consider the classes of clinical problems amenable to this approach. (3 pages)Genetic programming for the analysis of nuclear magnetic resonance spectroscopy data
http://dl-live.theiet.org/content/conferences/10.1049/ic_19970474
Good classification of human brain tumours based on <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR spectra of biopsy extracts were be obtained using a genetic programming (GP) approach. In addition, the most significant aspect of the analysis was that very simple functions gave classification results that were almost as good as the `best-ever' functions. The results from classification using GP are unclear. GP copes very well with the binary classification on brain tumours where the data is noisy, due to, e.g. diverse tumour types and uncertainties in histological classification. GP performs at least as well as NN, and finds solutions that are simple. On the multi-class rat data, where the data is more homogenous, GP performs much less well than NN. Only when some extra preprocessing is applied to the data does GP perform as well as the results from the human brain classification would lead one to expect. (3 pages)Pattern recognition analysis of <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR spectra from human tumour biopsy extracts: a European Union Concerted Action Project
http://dl-live.theiet.org/content/conferences/10.1049/ic_19970472
An automated data analysis approach has been developed for processing of <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR from tumour extracts. At present, the only manual interventions involve spectrum phasing (although automation methods are available) and choice of the number of principal components (usually chosen to account for about 99% of data variance). Unsupervised learning was important for identifying errors in the automatic processing scheme and for finding outliers (e.g. due to technical failures during extraction or NMR spectroscopy). it can also reveal underlying structure in the dataset (i.e. which classes of samples may be most easily separated). Factor analysis was useful for reducing data dimensionality (important for subsequent analysis) and, after vector rotation, for identifying important biochemical metabolites. Supervised learning (backpropagation NN) provided a robust classification method and was good for distinguishing between meningiomas and other types of brain tumour. Genetic programming analysis of a subset of these data gave comparable classification to NN but with quite simple `programs', facilitating biochemical interpretation. In general, classification of astrocytic tumours according to grade was not reliable based on <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR spectra from chemical extracts. Although in vivo <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR spectra of higher grade brain tumours might be characterised by elevated lipid signals, this information will be lost during extraction of water-soluble metabolites (as here). It would be expected that the best classification based on in vivo <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H NMR spectra would involve short echo-time measurements since these are most sensitive to glutamine signals (important for distinguishing tumour type) as well as lipid signals (possibly dependant on tumour grade). (3 pages)Towards an automated system to classify in vivo <sup xmlns="http://pub2web.metastore.ingenta.com/ns/">1</sup>H spectra of brain tumours
http://dl-live.theiet.org/content/conferences/10.1049/ic_19970476
In this study the authors have only considered pairs of classes and have used a very simple classification approach. However, the results are useful because they show which data points vary most between tumour types and suggest that a binary approach may be useful for discrimination. The authors' preliminary results show that it is possible to discriminate between the different classes of tumours using data point values which are extracted from the spectra after a minimum of pre-processing. These data points provided better discrimination than points selected from the tops of the peaks. This study differs from previous work in that the feature selection and normalisation procedures are completely objective. Linear discriminant analysis using selected data points seems a promising approach for the automated classification of in vivo human brain tumours. (3 pages)Pattern recognition methods for MRS analysis and classification
http://dl-live.theiet.org/content/conferences/10.1049/ic_19970473
It is clear that statistical classification of MRS has considerable potential as an accurate diagnostic advice tool. However, sample sizes are not yet sufficient to guarantee performance in large-scale clinical trials. In addition, the most commonly used classification strategies, namely Principal Components Analysis of in vitro spectra and Linear Discriminants Analysis of in vivo data, while offering considerable accuracy, are not optimal even within the currently available linear statistical classifiers. The issue of non-linearity and the consequent need of neural network analysis remains rests also on the outcome of larger studies, although the performance of these methods has already been shown to be competitive with that of statistical methods. Further work on principled methods for neural network design and rule-extraction offer the potential, in the near future, for high-performing, transparent non-linear classifiers. Alternative algorithms can also be used for automatic labelling, of clustering, of spectra. Overall, the results of the preliminary studies reported here are encouraging both for the accuracy that has been achieved and the consistency of the variable relevance ranking with clinical expectation. (9 pages)3-dimensional medical image reconstruction from serial cross sections
http://dl-live.theiet.org/content/conferences/10.1049/cp_19970949
In the field of medicine, it is useful to generate a three-dimensional (3D) view of human tissue or organ from its serial cross sectional images. This investigation is to carry out the 3D reconstruction and visualization of the human brain from its serial cross sectional magnetic resonance images (MRIs). The techniques involved are the pre-processing, texture analysis, image segmentation, interpolation, surface fitting and visualization. T1 weighted MRIs of the brain of a human male are used to assess the performance. The tissues which are of interest are the bone, gyrus and lobe. Fractal dimension and entropy are employed as the textural parameters. A novel technique of co-matching correspondence finding (CMCF) method is developed for the interpolation. A modified Phong shading is used to provide a realistic 3D view of the human skull.Point-to-point registration of non-rigid medical images using local elastic transformation methods
http://dl-live.theiet.org/content/conferences/10.1049/cp_19971002
Although a considerable amount of research has been devoted to the registration of medical images, a good proportion of this has been performed on rigid regions of the anatomy. This has tended to limit the practical value of such research. The registration of medical images which are non-rigid in nature is a seemingly difficult and not fully investigated problem. This paper presents an elastic transformation method which is used to perform matching of non-rigid medical images, that is, two-dimensional images of a region of the body such as the thorax or abdomen which exhibits a high level of elasticity. Some forms of elastic transformation may be more localized in scope than others, which leads to the categorization of elastic transformation methods into local and semi-local methods. This paper demonstrates the use of local elastic spatial transformations in matching non-rigid medical images. These transformations are illustrated by matching corresponding computed tomography and magnetic resonance images of the thorax.Mathematical morphology and active contours for object extraction and localization in medical images
http://dl-live.theiet.org/content/conferences/10.1049/cp_19970907
Segmentation refers to the process of extracting meaningful regions from images. Such regions typically correspond to objects of interest or to their parts. The segmentation of medical images of soft tissues into regions (corresponding to meaningful biological structures such as cells and organs) is a difficult problem because of the large variety of their characteristics. Numerous segmentation methods have been proposed; their choice depend on the type of images, and of a priori knowledge about the objects to be detected. We present a method of segmentation that is combination of two attractive tools for segmentation, morphological segmentation and active contours. We describe the principle of morphological segmentation and active contours segmentation. We also present a method integrating the two approaches. Finally, we present three examples of the use of this method: segmentation of isolated nuclei for DNA quantification; segmentation of tumoral lobules in histological sections and extraction of the cerebellum in MR image of a human brain.ROI approach to wavelet-based, hybrid compression of MR images
http://dl-live.theiet.org/content/conferences/10.1049/cp_19971013
This paper presents a novel medical image compression technique for inhomogeneous spatial reconstruction of MR images of the brain. The images are decomposed to various scales using the wavelet transform and a new multiscale segmentation algorithm is used to select areas of high diagnostic importance (regions of interest-ROI). Those areas, corresponding to brain tissue, tumours, and other structures in the head are compressed for maximum reconstruction quality, while neighbouring areas are coarsely approximated. The background is rejected since it contains only noise and no useful data. The quality of the reconstructed image is very good, even at low bit-rates, since the bit allocation is performed after diagnosis and reflects the diagnostic importance of each region. No useful data is lost at the selected ROI.Optimal feature extraction for the segmentation of medical images
http://dl-live.theiet.org/content/conferences/10.1049/cp_19971009
Many image segmentation algorithms use a small local area around each pixel for the extraction of features, in order to minimise the effect of image anomalies. The main drawback of this approach is its generation of classification errors at region boundaries, where the local area can contain pixels from more than one region. In this paper, a novel method of determining the optimal position of the local area for feature extraction is presented. The proposed technique avoids overlap into adjacent regions by examining the intensity gradients of neighbouring pixels and shifting the area for feature extraction accordingly. The improvement obtained using this technique is demonstrated on a variety of MRI medical images.On a hybrid watershed-neural network approach for the segmentation of high field MR images of the spinal cord
http://dl-live.theiet.org/content/conferences/10.1049/cp_19970922
We describe a general method for extraction of an image object based on an approximate knowledge of its shape and its position in the image. The method is a combination of the watershed approach and neural networks. It is applied to high field MR images of a section of a spinal cord for the extraction of the shape of the grey matter.Motion analysis for magnetic resonance myocardial perfusion imaging
http://dl-live.theiet.org/content/conferences/10.1049/cp_19971014
The assessment of regional myocardial perfusion using magnetic resonance (MR) imaging during the first-pass of a contrast agent bolus requires tracking of the signal time course for each myocardial segment so that a detailed perfusion map can be derived. To obtain such a map in practice, however, is not trivial because deformation of the shape of the myocardium and respiratory induced motion render a major difficulty in this process. This study describes a practical implementation of a real-time interactive MR echo-planar (EPI) myocardial perfusion imaging system and demonstrates an automated approach for motion and deformation tracking of functional myocardial perfusion images.Identification of boundaries in MRI medical images using artificial neural networks
http://dl-live.theiet.org/content/conferences/10.1049/ic_19960641
In the area of medical imaging, fully-automatic and robust segmentation techniques would have an enormous beneficial impact on clinical practice and research, by decreasing dramatically the manual effort which must otherwise be devoted to this task. Deployment of conventional image processing techniques has not so far led to a fully-automatic solution, although semi-automatic systems do exist. Since no known, robust segmentation algorithm exists, the ability of neural networks to discover regularities and features in complex data is appealing. Indeed, many preliminary attempts at neural segmentation have been described, although none yet achieves the necessary level of performance for routine application. Southampton General Hospital have a requirement to obtain lung-boundary data within an asthma research project. In connection with this requirement, we have previously reported on work in which multilayer perceptrons (MLPs) are trained using backpropagation to segment the region of the lungs in magnetic resonance images of the thorax. This is achieved by training the network to classify voxels as either boundary (voxels on the boundary between lung interior and surrounding tissue) or non-boundary. In this paper, we present the latest results using this technique. We also show how the generalisation performance of the MLP can be improved using a variety of techniques, including weight pruning algorithms. (6 pages)