Derek Magee
University of Leeds, Computing, Faculty Member
Research Interests:
This repository contains the data referred to in the paper "Raman spectroscopy of endoscopic colonic biopsies from patients with ulcerative colitis to identify mucosal inflammation and healing". It consists of experimental data... more
This repository contains the data referred to in the paper "Raman spectroscopy of endoscopic colonic biopsies from patients with ulcerative colitis to identify mucosal inflammation and healing". It consists of experimental data for the Raman spectra obtained for mucosally healed (or quiescent) and inflamed colon tissue and statistical data to identify significant differences between the peak intensities in the Raman spectra for the two different sample groups.
Research Interests:
– Cardiac phase differences between perfusion and angiography We are developing a framework for establishing direct correspondence between DCE-MRI findings and a patient-specific model of coronary blood supply obtainable from MRA. Here we... more
– Cardiac phase differences between perfusion and angiography We are developing a framework for establishing direct correspondence between DCE-MRI findings and a patient-specific model of coronary blood supply obtainable from MRA. Here we present an approach for mediated spatiotemporal registration designed to overcome the difficulties associated with 2D perfusion and 3D angiography registration. Our work improves on 2D Affine registration reported in [1]. 3. Mediated – Use Spatiotemporal of wall motion Registration cine for recovering phase difference transform Our method is based on the use of a non-rigid transform obtained from the analysis of temporal series. A – Phase difference transform is used to compensate for non-rigid transform spanning the phases from angiography to perfusion is derived from the 4D cine (vertically sparse); it is used as phase the spatiotemporal differences registration between mediator perfusion for registering and perfusion angiography and angiography data. Angiography phase 2. Cardiac Phase Mismatch The standard methods of rigid and non-rigid registration are not applicable in this context because of the potential cardiac phase mismatch between DCE-MRI and MRA datasets. perf.ap; perf.med; angio; perf.bas.
Research Interests: Biomedical Engineering, Magnetic Resonance Imaging, Biomarkers, Atherosclerosis, Medicine, and 15 moreMolecular Imaging, Adipose tissue, Humans, Clinical Medicine, Male, Lipid metabolism, Aged, Middle Aged, Necrosis, Reproducibility of Results, Statistical models, Sensitivity and Specificity, Carotid Artery, Hemorrhage, and Quantitative Susceptibility Mapping
Research Interests:
Abstract. Systems able to learn from visual observations have a great deal of potential for autonomous robotics, scientific discovery, and many other fields as the necessity to generalise from visual observation (from a quotidian scene or... more
Abstract. Systems able to learn from visual observations have a great deal of potential for autonomous robotics, scientific discovery, and many other fields as the necessity to generalise from visual observation (from a quotidian scene or from the results of a scientific enquiry) is inherent in various domains. We de-scribe an application to learning rules of a dice game using data from a vision system observing the game being played. In this paper, we experimented with two broad approaches: (i) a predictive learning approach with the Progol system, where explicit concept learning problems are posed and solved, and (ii) a descrip-tive learning approach with the HR system, where a general theory is formed with no specific problem solving task in mind and rules are extracted from the theory. 1
Research Interests:
This paper builds on existing work on learning protocol behaviour from observation to propose a new framework for visual attention. The main contribution of this work resides in the fact that attention is not given a priori to the vision... more
This paper builds on existing work on learning protocol behaviour from observation to propose a new framework for visual attention. The main contribution of this work resides in the fact that attention is not given a priori to the vision system but learned by induction from the active observation of patterns in space. These patterns are sequences of coloured objects that are placed by an agent in a camera field of view. Therefore, in this work we propose a method for learning the focus of attention from the visual observation of tasks executed by an agent. The description of objects in space in terms of observer-object relative frames of references, named local cardinal systems is a second contribution of this work. 1
Research Interests:
We present the first results of a new technique to bin Cone Beam projections without imposing any motion model. Such a technique is required for studying motion in regions of the body, such as the pelvis, where motion exists and is... more
We present the first results of a new technique to bin Cone Beam projections without imposing any motion model. Such a technique is required for studying motion in regions of the body, such as the pelvis, where motion exists and is unpredictable. All motion information is obtained directly from the projections and the binning is performed through a type of best first search through the graph of possible complete assignments. Simplifying assumptions coupled with loss-less dimensional reduction using Principal Component Analysis, make the method tractable. 1
Motivated by applications such as automated visual surveillance and video monitoring and annotation, there has been a lot of interest in constructing cognitive vision systems capable of interpreting the high level semantics of dynamic... more
Motivated by applications such as automated visual surveillance and video monitoring and annotation, there has been a lot of interest in constructing cognitive vision systems capable of interpreting the high level semantics of dynamic scenes. In this paper we present a novel approach for automatically inferring models of object interactions that can be used to interpret observed behaviour within a scene. A real-time low-level computer vision system, together with an attentional control mechanism, are used to identify incidents or events that occur in the scene. A data driven approach has been taken in order to automatically infer discrete and abstract representations (symbols) of primitive object interactions; effectively the system learns a set of qualitative spatial relations relevant to the dynamic behaviour of the domain. These symbols then form the alphabet of a VLMM which automatically infers the high level structure of typical interactive behaviour. The learnt behaviour model...
Research Interests:
A method for automatic 2D image registration based on combining multiple row-wise and column-wise dynamic programming (DP) processes is presented. Combination is pe rformed by means of a parametric transform (either rigid or non-rigid)... more
A method for automatic 2D image registration based on combining multiple row-wise and column-wise dynamic programming (DP) processes is presented. Combination is pe rformed by means of a parametric transform (either rigid or non-rigid) which is estimated from the output of the dynamic programming processes. The method performs particularly well on histopathology images as dynamic programming is a global optimisation method with the ability to overcome local optima often encountered when registering such images. Our method has been applied to reconstruction of volumetric data from histopathology sections by serial application to successi ve pairs of images. An evaluation is presented that confirms our metho d compares favourably with state-of-the-art methods both interms of accuracy and computational cost for this task. Microscopic 3D Histopathology has potential application in the analysi s of anatomy and atlas building (e.g. in transgenic mice) and the analysis of tissue structure ...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests: Mathematics, Algorithms, Biomedical Engineering, Magnetic Resonance Imaging, Medicine, and 15 moreQuantitative analysis, Humans, Female, Male, Middle Aged, Observer Variation, Adult, Myocardium, Reproducibility of Results, Coronary Circulation, Quantitative Analysis, Sensitivity and Specificity, Blood Flow Velocity, Contrast Media, and Magnetic resonance image
Our objective is to efficiently and accurately estimate human upper body pose in ges-ture videos. To this end, we build on the recent successful applications of random forests (RF) classifiers and regressors, and develop a pose estimation... more
Our objective is to efficiently and accurately estimate human upper body pose in ges-ture videos. To this end, we build on the recent successful applications of random forests (RF) classifiers and regressors, and develop a pose estimation model with the following novelties: (i) the joints are estimated sequentially, taking account of the human kinematic chain. This means that we don’t have to make the simplifying assumption of most pre-vious RF methods – that the joints are estimated independently; (ii) by combining both classifiers (as a mixture of experts) and regressors, we show that the learning problem is tractable and that more context can be taken into account; and (iii) dense optical flow is used to align multiple expert joint position proposals from nearby frames, and thereby improve the robustness of the estimates. The resulting method is computationally efficient and can overcome a number of the errors (e.g. confusing left/right hands) made by RF pose estimators that infe...
A qualitative image description grammar with automatic image fitting and object modelling algorithms is presented. The grammar is based on assigning a square sub-region of an image one of a finite number of qualitative labels, based on... more
A qualitative image description grammar with automatic image fitting and object modelling algorithms is presented. The grammar is based on assigning a square sub-region of an image one of a finite number of qualitative labels, based on the occurrence of object boundaries within this region and how these intersect the region boundary. In the general case there is an infinite number of such labels, however the use of a multi-scale approach allows a finite (small) number of labels at each scale. This makes the problem tractable within a constraint satisfaction type framework. Constraints are put on neighboring labels based on the premise that all object boundaries are continuous, having no ending within an image. A minimum description length (MDL) approach is suggested for description hypothesis selection (based on colour histograms) and methods for (constraint based) hypothesis generation/adaption and (Hidden Markov Model based) a-priori shape modelling are presented. 1
Research Interests:
Learning general truths from the observation of simple domains and, further, learning how to use this knowledge are essential capabilities for any intelligent agent to understand and execute informed actions in the real world. The aim of... more
Learning general truths from the observation of simple domains and, further, learning how to use this knowledge are essential capabilities for any intelligent agent to understand and execute informed actions in the real world. The aim of this work is the investigation of the automatic learning of mathematical structures from visual observation. This research was conducted upon a system that combines computer vision with inductive logic programming that was first designed to learn protocol behaviour from observation. In this paper we show how transitivity, reflexivity and symmetry axioms could be induced from the noisy data provided by the vision system. Noise in the data accounts for the generation of a large number of possible generalisations by the ILP system, most of which do not represent interesting concepts about the observed domain. In order to automatically choose the best answers among those generated by induction, we propose a method for combining the results of multiple I...
Research Interests:
An efficient and general framework for the incorporation of statistical prior information, based on a wide variety of detectable point features, into level set based object tracking is presented. Level set evolution is based on the... more
An efficient and general framework for the incorporation of statistical prior information, based on a wide variety of detectable point features, into level set based object tracking is presented. Level set evolution is based on the maximisation of a set of likelihoods on mesh values at features, which are located using a stochastic sampling process. This evolution is based on the interpolation of likelihood gradients using kernels centred at the features. Feature detectors implemented are based on moments of colour histogram segmented images and learned image patches located using normalised correlation, although a wide variety of feature detectors could be used. A computationally efficient level set implementation is presented along with a method for the incorporation of a motion model into the scheme. 1
Research Interests:
With the advent of digital histopathology imaging and automatic image analysis, colour constancy across multiple microscope slides has become an impor tant issue. Colour variation due to chemical, user or protocol inconsistency is... more
With the advent of digital histopathology imaging and automatic image analysis, colour constancy across multiple microscope slides has become an impor tant issue. Colour variation due to chemical, user or protocol inconsistency is widespread. This paper presents an approach for computationally efficient context aware colo ur c assification. A ‘context vector’ derived from the colour distribution of the complete ima ge is combined with the per-pixel information to improve pixel classification perform ance. The context vector implicitly encodes global image information such as whether the slide is under/over stained, or cut thinly, or thickly. The method is evaluated for s egmentation accuracy on two data sets with different stains, and as a pre-processin g method for a cell nuclei detection algorithm.
Research Interests:
Stack misalignment in cardiac MR cine series distorts the correct appearance of anatomical features and reduces the reproducibility of volumetric measurements computed for diagnostic purposes. This paper describes a method for correction... more
Stack misalignment in cardiac MR cine series distorts the correct appearance of anatomical features and reduces the reproducibility of volumetric measurements computed for diagnostic purposes. This paper describes a method for correction of stack misalignment in cardiac MR series. Our method involves registration against a reference volume and features a prominent enhancement designed to circumvent the weaknesses associated with slice-to-volume registration. The core of the presented method is a custom stack alignment transform which parametrises the in-plane movement for all slices independently of each other; at the same time the image similarity metric for every optimisation iteration is calculated on the whole stack with all slice correction parameters contributing to the result. The method was evaluated on 50 clinical datasets with a 100 simulated optimisation runs for every dataset. The results show that on average the method is able to recover alignment parameters with sub-vo...
Research Interests:
This paper presents a novel technique for classifying both cell nuclei and tissue regions in liver specimens by incorporating context information, linking cell nuclei and tissue regions using Bayesian networks. The method works in two... more
This paper presents a novel technique for classifying both cell nuclei and tissue regions in liver specimens by incorporating context information, linking cell nuclei and tissue regions using Bayesian networks. The method works in two stages: (i) initial classification of cell nuclei and tissue regions; and (ii) integrating the initial classifications using a Bayesian network to enforce consistancy (thus including context). Results demonstrate that our method of incorporating context information is superior to the classification that uses only object based features for both nucleus and region classification.
Research Interests:
Abstract. Systems able to learn from visual observations have a great deal of potential for autonomous robotics, scientific discovery, and many other fields as the necessity to generalise from visual observation (from a quotidian scene or... more
Abstract. Systems able to learn from visual observations have a great deal of potential for autonomous robotics, scientific discovery, and many other fields as the necessity to generalise from visual observation (from a quotidian scene or from the results of a scientific enquiry) is inherent ...
Research Interests:
Research Interests:
Research Interests:
Human ear reconstruction is recognized as the emblematic enterprise in tissue engineering. Up to now, it has failed to reach human applications requiring appropriate tissue complexity along with an accessible vascular tree. We hereby... more
Human ear reconstruction is recognized as the emblematic enterprise in tissue engineering. Up to now, it has failed to reach human applications requiring appropriate tissue complexity along with an accessible vascular tree. We hereby propose a new method to process human auricles in order to provide a poorly immunogenic, complex and vascularized ear graft scaffold. 12 human ears with their vascular pedicles were procured. Perfusion-decellularization was applied using a SDS/polar solvent protocol. Cell and antigen removal was examined by histology and DNA was quantified. Preservation of the extracellular matrix (ECM) was assessed by conventional and 3D-histology, proteins and cytokines quantifications. Biocompatibility was assessed by implantation in rats for up to 60 days. Adipose-derived stem cells seeding was conducted on scaffold samples and with human aortic endothelial cells whole graft seeding in a perfusion-bioreactor. Histology confirmed cell and antigen clearance. DNA reduc...
Research Interests:
To investigate if the early treatment effects of radiofrequency ablation (RFA) on renal cell carcinoma (RCC) can be detected with dynamic contrast enhanced (DCE)-MRI and to correlate RCC perfusion with RFA treatment time. 20 patients... more
To investigate if the early treatment effects of radiofrequency ablation (RFA) on renal cell carcinoma (RCC) can be detected with dynamic contrast enhanced (DCE)-MRI and to correlate RCC perfusion with RFA treatment time. 20 patients undergoing RFA of their 21 RCCs were evaluated with DCE-MRI before and at one month after RFA treatment. Perfusion was estimated using the maximum slope technique at two independent sittings. Total RCC blood flow was correlated with total RFA treatment time, tumour location, size and histology. DCE-MRI examinations were successfully evaluated for 21 RCCs (size from 1.3 to 4 cm). Perfusion of the RCCs decreased significantly (p < 0.0001) from a mean of 203 (±80) mL/min/100 mL before RFA to 8.1 (±3.1) mL/min/100 mL after RFA with low intra-observer variability (r ≥ 0.99, p < 0.0001). There was an excellent correlation (r = 0.95) between time to complete ablation and pre-treatment total RCC blood flow. Tumours with an exophytic location exhibit the l...
Research Interests:
Research Interests:
Margin status and invasion pattern are prognostic factors for oral tongue squamous cell carcinoma (OTSCC). Current methods to identify these factors are limited to 2D observation; it is necessary to explore 3D reconstruction with... more
Margin status and invasion pattern are prognostic factors for oral tongue squamous cell carcinoma (OTSCC). Current methods to identify these factors are limited to 2D observation; it is necessary to explore 3D reconstruction with whole-mount sample to improve the accuracy of analysis. This study aimed to study the tissue preparation, section generation, and 3D reconstruction with whole-mount OTSCC specimen. Two OTSCC samples were retrieved from Nanjing Stomatological Hospital, Medical School of Nanjing University. One sample was sliced into 3 equal-sized pieces and subjected to different processing schedules to determine the best method. The second sample was processed accordingly. Serial whole-mount sections of the second sample were generated, stained with HE/anticytokine antibody in intersection manner, and scanned into digital images. Digital images were aligned and reconstructed into 3D images with Hetero Genius Medical Image Manager 3D Pathology Add-On [HGMIM3D]. Successful se...
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Respiratory motion is a significant obstacle to the use of quantitative perfusion in clinical practice. Increasingly complex motion correction algorithms are being developed to correct for respiratory motion. However, the impact of these... more
Respiratory motion is a significant obstacle to the use of quantitative perfusion in clinical practice. Increasingly complex motion correction algorithms are being developed to correct for respiratory motion. However, the impact of these improvements on the final diagnosis of ischemic heart disease has not been evaluated. The aim of this study was to compare the performance of four automated correction methods in terms of their impact on diagnostic accuracy. Three strategies for motion correction were used: (1) independent translation correction for all slices, (2) translation correction for the basal slice with transform propagation to the remaining two slices assuming identical motion in the remaining slices, and (3) rigid correction (translation and rotation) for the basal slice. There were no significant differences in diagnostic accuracy between the manual and automatic motion-corrected datasets ([Formula: see text]). The area under the curve values for manual motion correction...
Research Interests:
Research Interests:
Widespread screening mammography programmes mean that ductal carcinoma in situ (DCIS), a pre-invasive breast lesion, is now more frequently diagnosed. However, not all diagnosed DCIS lesions progress to invasive breast cancer, which... more
Widespread screening mammography programmes mean that ductal carcinoma in situ (DCIS), a pre-invasive breast lesion, is now more frequently diagnosed. However, not all diagnosed DCIS lesions progress to invasive breast cancer, which presents a dilemma for clinicians. As such, there is much interest in studying DCIS in the laboratory, in order to help understand more about its biology and determine the characteristics of those that progress to invasion. Greater knowledge would lead to targeted and better DCIS treatment. Here, we outline some of the models available to study DCIS, with a particular focus on animal-free systems.