Automatic classification of MR scans in Alzheimer's disease.

To be diagnostically useful, structural MRI must reliably distinguish Alzheimer's disease (AD) from normal aging in individual scans. Recent advances in statistical learning theory have led to the application of support vector machines to MRI for detection of a variety of disease states. The aims of this study were to assess how successfully support vector machines assigned individual diagnoses and to determine whether data-sets combined from multiple scanners and different centres could be used to obtain effective classification of scans. We used linear support vector machines to classify the grey matter segment of T1-weighted MR scans from pathologically proven AD patients and cognitively normal elderly individuals obtained from two centres with different scanning equipment. Because the clinical diagnosis of mild AD is difficult we also tested the ability of support vector machines to differentiate control scans from patients without post-mortem confirmation. Finally we sought to use these methods to differentiate scans between patients suffering from AD from those with frontotemporal lobar degeneration. Up to 96% of pathologically verified AD patients were correctly classified using whole brain images. Data from different centres were successfully combined achieving comparable results from the separate analyses. Importantly, data from one centre could be used to train a support vector machine to accurately differentiate AD and normal ageing scans obtained from another centre with different subjects and different scanner equipment. Patients with mild, clinically probable AD and age/sex matched controls were correctly separated in 89% of cases which is compatible with published diagnosis rates in the best clinical centres. This method correctly assigned 89% of patients with post-mortem confirmed diagnosis of either AD or frontotemporal lobar degeneration to their respective group. Our study leads to three conclusions: Firstly, support vector machines successfully separate patients with AD from healthy aging subjects. Secondly, they perform well in the differential diagnosis of two different forms of dementia. Thirdly, the method is robust and can be generalized across different centres. This suggests an important role for computer based diagnostic image analysis for clinical practice.

A data-dependent skeleton estimate and a scale-sensitive dimension for classification

The classical binary classification problem is investigatedwhen it is known in advance that the posterior probability function(or regression function) belongs to some class of functions. We introduceand analyze a method which effectively exploits this knowledge. The methodis based on minimizing the empirical risk over a carefully selected``skeleton'' of the class of regression functions. The skeleton is acovering of the class based on a data--dependent metric, especiallyfitted for classification. A new scale--sensitive dimension isintroduced which is more useful for the studied classification problemthan other, previously defined, dimension measures. This fact isdemonstrated by performance bounds for the skeleton estimate in termsof the new dimension.

Parametric imaging for characterizing focal liver lesions in contrast-enhanced ultrasound.

The differentiation between benign and malignant focal liver lesions plays an important role in diagnosis of liver disease and therapeutic planning of local or general disease. This differentiation, based on characterization, relies on the observation of the dynamic vascular patterns (DVP) of lesions with respect to adjacent parenchyma, and may be assessed during contrast-enhanced ultrasound imaging after a bolus injection. For instance, hemangiomas (i.e., benign lesions) exhibit hyper-enhanced signatures over time, whereas metastases (i.e., malignant lesions) frequently present hyperenhanced foci during the arterial phase and always become hypo-enhanced afterwards. The objective of this work was to develop a new parametric imaging technique, aimed at mapping the DVP signatures into a single image called a DVP parametric image, conceived as a diagnostic aid tool for characterizing lesion types. The methodology consisted in processing a time sequence of images (DICOM video data) using four consecutive steps: (1) pre-processing combining image motion correction and linearization to derive an echo-power signal, in each pixel, proportional to local contrast agent concentration over time; (2) signal modeling, by means of a curve-fitting optimization, to compute a difference signal in each pixel, as the subtraction of adjacent parenchyma kinetic from the echopower signal; (3) classification of difference signals; and (4) parametric image rendering to represent classified pixels as a support for diagnosis. DVP parametric imaging was the object of a clinical assessment on a total of 146 lesions, imaged using different medical ultrasound systems. The resulting sensitivity and specificity were 97% and 91%, respectively, which compare favorably with scores of 81 to 95% and 80 to 95% reported in medical literature for sensitivity and specificity, respectively.

Natural classification of knots

The principal objective of the knot theory is to provide a simple way of classifying and ordering all the knot types. Here, we propose a natural classification of knots based on their intrinsic position in the knot space that is defined by the set of knots to which a given knot can be converted by individual intersegmental passages. In addition, we characterize various knots using a set of simple quantum numbers that can be determined upon inspection of minimal crossing diagram of a knot. These numbers include: crossing number; average three-dimensional writhe; number of topological domains; and the average relaxation value

Active learning methods for remote sensing image classification

In this paper, we propose two active learning algorithms for semiautomatic definition of training samples in remote sensing image classification. Based on predefined heuristics, the classifier ranks the unlabeled pixels and automatically chooses those that are considered the most valuable for its improvement. Once the pixels have been selected, the analyst labels them manually and the process is iterated. Starting with a small and nonoptimal training set, the model itself builds the optimal set of samples which minimizes the classification error. We have applied the proposed algorithms to a variety of remote sensing data, including very high resolution and hyperspectral images, using support vector machines. Experimental results confirm the consistency of the methods. The required number of training samples can be reduced to 10% using the methods proposed, reaching the same level of accuracy as larger data sets. A comparison with a state-of-the-art active learning method, margin sampling, is provided, highlighting advantages of the methods proposed. The effect of spatial resolution and separability of the classes on the quality of the selection of pixels is also discussed.

Feature Extraction from Whole-Sky Ground-Based Images for Cloud-Type Recognition

Several features that can be extracted from digital images of the sky and that can be useful for cloud-type classification of such images are presented. Some features are statistical measurements of image texture, some are based on the Fourier transform of the image and, finally, others are computed from the image where cloudy pixels are distinguished from clear-sky pixels. The use of the most suitable features in an automatic classification algorithm is also shown and discussed. Both the features and the classifier are developed over images taken by two different camera devices, namely, a total sky imager (TSI) and a whole sky imager (WSC), which are placed in two different areas of the world (Toowoomba, Australia; and Girona, Spain, respectively). The performance of the classifier is assessed by comparing its image classification with an a priori classification carried out by visual inspection of more than 200 images from each camera. The index of agreement is 76% when five different sky conditions are considered: clear, low cumuliform clouds, stratiform clouds (overcast), cirriform clouds, and mottled clouds (altocumulus, cirrocumulus). Discussion on the future directions of this research is also presented, regarding both the use of other features and the use of other classification techniques

A protein interaction atlas for the nuclear receptors: properties and quality of a hub-based dimerisation network.

BACKGROUND: The nuclear receptors are a large family of eukaryotic transcription factors that constitute major pharmacological targets. They exert their combinatorial control through homotypic heterodimerisation. Elucidation of this dimerisation network is vital in order to understand the complex dynamics and potential cross-talk involved. RESULTS: Phylogeny, protein-protein interactions, protein-DNA interactions and gene expression data have been integrated to provide a comprehensive and up-to-date description of the topology and properties of the nuclear receptor interaction network in humans. We discriminate between DNA-binding and non-DNA-binding dimers, and provide a comprehensive interaction map, that identifies potential cross-talk between the various pathways of nuclear receptors. CONCLUSION: We infer that the topology of this network is hub-based, and much more connected than previously thought. The hub-based topology of the network and the wide tissue expression pattern of NRs create a highly competitive environment for the common heterodimerising partners. Furthermore, a significant number of negative feedback loops is present, with the hub protein SHP [NR0B2] playing a major role. We also compare the evolution, topology and properties of the nuclear receptor network with the hub-based dimerisation network of the bHLH transcription factors in order to identify both unique themes and ubiquitous properties in gene regulation. In terms of methodology, we conclude that such a comprehensive picture can only be assembled by semi-automated text-mining, manual curation and integration of data from various sources.

A wavelet-based method for the determination of the relative resolution between remotely sensed images

Spatial resolution is a key parameter of all remote sensing satellites and platforms. The nominal spatial resolution of satellites is a well-known characteristic because it is directly related to the area in ground that represents a pixel in the detector. Nevertheless, in practice, the actual resolution of a specific image obtained from a satellite is difficult to know precisely because it depends on many other factors such as atmospheric conditions. However, if one has two or more images of the same region, it is possible to compare their relative resolutions. In this paper, a wavelet-decomposition-based method for the determination of the relative resolution between two remotely sensed images of the same area is proposed. The method can be applied to panchromatic, multispectral, and mixed (one panchromatic and one multispectral) images. As an example, the method was applied to compute the relative resolution between SPOT-3, Landsat-5, and Landsat-7 panchromatic and multispectral images taken under similar as well as under very different conditions. On the other hand, if the true absolute resolution of one of the images of the pair is known, the resolution of the other can be computed. Thus, in the last part of this paper, a spatial calibrator that is designed and constructed to help compute the absolute resolution of a single remotely sensed image is described, and an example of its use is presented.

Super-resolution of remotely sensed images with variable-pixel linear reconstruction

This paper describes the development and applications of a super-resolution method, known as Super-Resolution Variable-Pixel Linear Reconstruction. The algorithm works combining different lower resolution images in order to obtain, as a result, a higher resolution image. We show that it can make significant spatial resolution improvements to satellite images of the Earth¿s surface allowing recognition of objects with size approaching the limiting spatial resolution of the lower resolution images. The algorithm is based on the Variable-Pixel Linear Reconstruction algorithm developed by Fruchter and Hook, a well-known method in astronomy but never used for Earth remote sensing purposes. The algorithm preserves photometry, can weight input images according to the statistical significance of each pixel, and removes the effect of geometric distortion on both image shape and photometry. In this paper, we describe its development for remote sensing purposes, show the usefulness of the algorithm working with images as different to the astronomical images as the remote sensing ones, and show applications to: 1) a set of simulated multispectral images obtained from a real Quickbird image; and 2) a set of multispectral real Landsat Enhanced Thematic Mapper Plus (ETM+) images. These examples show that the algorithm provides a substantial improvement in limiting spatial resolution for both simulated and real data sets without significantly altering the multispectral content of the input low-resolution images, without amplifying the noise, and with very few artifacts.

Description of a tertiary swiss university hospital palliative population based on the International Classifiaction of Disease (ICD) : a retrospective pilot study

Introduction : L'équipe mobile de soins palliatifs intra hospitalière (EMSP) du Centre Hospitalier Universitaire Vaudois (CHUV) a été mise en place en 1996. Il s'agit d'une des premières équipes interdisciplinaire de consultants mise à disposition d'un hôpital tertiaire. Le CHUV est l'hôpital de proximité de la ville de Lausanne (850 lits) mais aussi un hôpital de référence pour le reste du canton. En 2007, il y a eu 38'359 patients hospitalisés au CHUV. Les facteurs d'évaluation du taux d'utilisation d'une équipe mobile de soins palliatifs consultantes sont variés et complexes. Plusieurs méthodes sont décrites dans la littérature pour tenter de répondre à cette problématique. Avant de pouvoir évaluer l'utilisation de notre équipe mobile consultante de soins palliatifs intra hospitalière, il nous est apparu nécessaire de mieux décrire et définir la population qui meurt dans notre institution. McNamara et collègues ont proposé des critères qui classifient une population palliative comme « minimale », « intermédiaire » ou « maximale ». L'objectif de cette étude est de déterminer le taux de patients décédés au CHUV sur une période de 4 mois (Γ1 février au 31 mai 2007) suivie par notre EMSP en utilisant la méthode de classification «minimal » et « maximal ». Méthode : les archives médicales du CHUV ont été analysées pour chaque patient adulte décédé pendant la période sélectionnée. Les populations « maximal » et « minimal » de ces patients ont été ensuite déterminées selon des critères basés sur les codes diagnostiques figurants sur les certificats de décès. De ces deux populations, nous avons identifié à partir de notre base de données, les patients qui ont été suivie par notre EMSP. Le CHUV utilise les mêmes codes diagnostiques (International Classification of Disease, ICD) que ceux utilisés dans la classification de McNamara. Une recherche pilote effectuée dans les archives médicales du CHUV manuellement en analysant en profondeur l'ensemble du dossier médical a révélé que la classification de la population « minimal » pouvait être biaisée notamment en raison d'une confusion entre la cause directe du décès (complication d'une maladie) et la maladie de base. Nous avons estimé le pourcentage d'erreur de codification en analysé un échantillon randomisé de patients qui remplissait les critères « minimal ». Résultats : sur un total de 294 décès, 263 (89%) remplissaient initialement les critères « maximal » et 83 (28%) les critères «minimal», l'analyse de l'échantillon randomisé de 56 dossiers de patients sur les 180 qui ne remplissaient pas les critères « minimal » ont révélé que 21 (38%) auraient dus être inclus dans la population « minimal ». L'EMSP a vu 67/263 (25.5%) de la population palliative « maximal » et 56/151 (37.1%) de la population palliative « minimal ». Conclusion : cette étude souligne l'utilité de la méthode proposée par McNamara pour déterminer la population de patients palliatifs. Cependant, notre travail illustre aussi une limite importante de l'estimation de la population « minima » en lien avec l'imprécision des causes de décès figurant sur les certificats de décès de notre institution. Nos résultats mettent aussi en lumière que l'EMSP de notre institution est clairement sous- utilisée. Nous prévoyons une étude prospective de plus large envergure utilisant la même méthodologie afin d'approfondir les résultats de cette étude pilote.

High specificity of line-immunoassay based algorithms for recent HIV-1 infection independent of viral subtype and stage of disease.

ABSTRACT: BACKGROUND: Serologic testing algorithms for recent HIV seroconversion (STARHS) provide important information for HIV surveillance. We have shown that a patient's antibody reaction in a confirmatory line immunoassay (INNO-LIATM HIV I/II Score, Innogenetics) provides information on the duration of infection. Here, we sought to further investigate the diagnostic specificity of various Inno-Lia algorithms and to identify factors affecting it. METHODS: Plasma samples of 714 selected patients of the Swiss HIV Cohort Study infected for longer than 12 months and representing all viral clades and stages of chronic HIV-1 infection were tested blindly by Inno-Lia and classified as either incident (up to 12 m) or older infection by 24 different algorithms. Of the total, 524 patients received HAART, 308 had HIV-1 RNA below 50 copies/mL, and 620 were infected by a HIV-1 non-B clade. Using logistic regression analysis we evaluated factors that might affect the specificity of these algorithms. RESULTS: HIV-1 RNA <50 copies/mL was associated with significantly lower reactivity to all five HIV-1 antigens of the Inno-Lia and impaired specificity of most algorithms. Among 412 patients either untreated or with HIV-1 RNA ≥50 copies/mL despite HAART, the median specificity of the algorithms was 96.5% (range 92.0-100%). The only factor that significantly promoted false-incident results in this group was age, with false-incident results increasing by a few percent per additional year. HIV-1 clade, HIV-1 RNA, CD4 percentage, sex, disease stage, and testing modalities exhibited no significance. Results were similar among 190 untreated patients. CONCLUSIONS: The specificity of most Inno-Lia algorithms was high and not affected by HIV-1 variability, advanced disease and other factors promoting false-recent results in other STARHS. Specificity should be good in any group of untreated HIV-1 patients.

Population-based registration of phenotypic anatomic and familial data for melanoma: results from a Swiss multicentric study

CONTEXT AND OBJECTIVES: A multicentric study was set up to assess the feasibility for Swiss cancer registries of actively retrieving 3 additional variables of epidemiological and a etiological relevance for melanoma, and of potential use for the evaluation of prevention campaigns. MATERIAL AND METHODS: The skin type, family history of melanoma and precise anatomical site were retrieved for melanoma cases registered in 5 Swiss cantons (Neuchâtel, St-Gall and Appenzell, Vaud and Wallis) over 3 to 6 consecutive years (1995-2002). Data were obtained via a short questionnaire administered by the physicians - mostly dermatologists - who originally excised the lesions. As the detailed body site was routinely collected in Ticino, data from this Cancer Registry were included in the body site analysis. Relative melanoma density (RMD) was computed by the ratio of observed to expected numbers of melanomas allowing for body site surface areas, and further adjusted for site-specific melanocyte density. RESULTS: Of the 1,645 questionnaires sent, 1,420 (86.3%) were returned. The detailed cutaneous site and skin type were reliably obtained for 84.7% and 78.7% of questionnaires, and family history was known in 76% of instances. Prevalence of sun-sensitive subjects and patients with melanoma affected first-degree relatives, two target groups for early detection and surveillance campaigns were 54.1% and 3.4%, respectively. After translation into the 4th digit of the International Classification of Diseases for Oncology, the anatomical site codes from printed (original information) and pictorial support (body chart from the questionnaire) concurred for 94.6% of lesions. Discrepancies occurred mostly for lesions on the upper, outer part of the shoulder for which the clinician's textual description was "shoulder blade". This differential misclassification suggests under-estimation by about 10% of melanomas of the upper limbs and an over-estimation of 5% for truncal melanomas. Sites of highest melanoma risk were the face, the shoulder and the upper arm for sexes, the back for men and the leg for women. Three major features of this series were: (1) an unexpectedly high RMD for the face in women (6.2 vs 4.2 in men), (2) the absence of a male predominance for melanomas on the ears, and (3) for the upper limbs, a steady gradient of increasing melanoma density with increasing proximity to the trunk, regardless of sex. DISCUSSION AND CONCLUSION: The feasibility of retrieving the skin type, the precise anatomical location and family history of melanoma in a reliable manner was demonstrated thanks to the collaboration of Swiss dermatologists. Use of a schematic body drawing improves the quality of the anatomical site data and facilitate the reporting task of doctors. Age and sex patterns of RMD paralleled general indicators of sun exposure and behaviour, except for the hand (RMD=0.2). These Swiss results support some site or sun exposure specificity in the aetiology of melanoma.

A Novel Volume-Age-KPS (VAK) Glioblastoma Classification Identifies a Prognostic Cognate microRNA-Gene Signature.

BACKGROUND: Several studies have established Glioblastoma Multiforme (GBM) prognostic and predictive models based on age and Karnofsky Performance Status (KPS), while very few studies evaluated the prognostic and predictive significance of preoperative MR-imaging. However, to date, there is no simple preoperative GBM classification that also correlates with a highly prognostic genomic signature. Thus, we present for the first time a biologically relevant, and clinically applicable tumor Volume, patient Age, and KPS (VAK) GBM classification that can easily and non-invasively be determined upon patient admission. METHODS: We quantitatively analyzed the volumes of 78 GBM patient MRIs present in The Cancer Imaging Archive (TCIA) corresponding to patients in The Cancer Genome Atlas (TCGA) with VAK annotation. The variables were then combined using a simple 3-point scoring system to form the VAK classification. A validation set (N = 64) from both the TCGA and Rembrandt databases was used to confirm the classification. Transcription factor and genomic correlations were performed using the gene pattern suite and Ingenuity Pathway Analysis. RESULTS: VAK-A and VAK-B classes showed significant median survival differences in discovery (P = 0.007) and validation sets (P = 0.008). VAK-A is significantly associated with P53 activation, while VAK-B shows significant P53 inhibition. Furthermore, a molecular gene signature comprised of a total of 25 genes and microRNAs was significantly associated with the classes and predicted survival in an independent validation set (P = 0.001). A favorable MGMT promoter methylation status resulted in a 10.5 months additional survival benefit for VAK-A compared to VAK-B patients. CONCLUSIONS: The non-invasively determined VAK classification with its implication of VAK-specific molecular regulatory networks, can serve as a very robust initial prognostic tool, clinical trial selection criteria, and important step toward the refinement of genomics-based personalized therapy for GBM patients.

Domain Adaptation in remote sensing: increasing the portability of land-cover classifiers

Among the types of remote sensing acquisitions, optical images are certainly one of the most widely relied upon data sources for Earth observation. They provide detailed measurements of the electromagnetic radiation reflected or emitted by each pixel in the scene. Through a process termed supervised land-cover classification, this allows to automatically yet accurately distinguish objects at the surface of our planet. In this respect, when producing a land-cover map of the surveyed area, the availability of training examples representative of each thematic class is crucial for the success of the classification procedure. However, in real applications, due to several constraints on the sample collection process, labeled pixels are usually scarce. When analyzing an image for which those key samples are unavailable, a viable solution consists in resorting to the ground truth data of other previously acquired images. This option is attractive but several factors such as atmospheric, ground and acquisition conditions can cause radiometric differences between the images, hindering therefore the transfer of knowledge from one image to another. The goal of this Thesis is to supply remote sensing image analysts with suitable processing techniques to ensure a robust portability of the classification models across different images. The ultimate purpose is to map the land-cover classes over large spatial and temporal extents with minimal ground information. To overcome, or simply quantify, the observed shifts in the statistical distribution of the spectra of the materials, we study four approaches issued from the field of machine learning. First, we propose a strategy to intelligently sample the image of interest to collect the labels only in correspondence of the most useful pixels. This iterative routine is based on a constant evaluation of the pertinence to the new image of the initial training data actually belonging to a different image. Second, an approach to reduce the radiometric differences among the images by projecting the respective pixels in a common new data space is presented. We analyze a kernel-based feature extraction framework suited for such problems, showing that, after this relative normalization, the cross-image generalization abilities of a classifier are highly increased. Third, we test a new data-driven measure of distance between probability distributions to assess the distortions caused by differences in the acquisition geometry affecting series of multi-angle images. Also, we gauge the portability of classification models through the sequences. In both exercises, the efficacy of classic physically- and statistically-based normalization methods is discussed. Finally, we explore a new family of approaches based on sparse representations of the samples to reciprocally convert the data space of two images. The projection function bridging the images allows a synthesis of new pixels with more similar characteristics ultimately facilitating the land-cover mapping across images.

Automatic Mapping and Classification of Spatial Environmental Data

The paper deals with the development and application of the generic methodology for automatic processing (mapping and classification) of environmental data. General Regression Neural Network (GRNN) is considered in detail and is proposed as an efficient tool to solve the problem of spatial data mapping (regression). The Probabilistic Neural Network (PNN) is considered as an automatic tool for spatial classifications. The automatic tuning of isotropic and anisotropic GRNN/PNN models using cross-validation procedure is presented. Results are compared with the k-Nearest-Neighbours (k-NN) interpolation algorithm using independent validation data set. Real case studies are based on decision-oriented mapping and classification of radioactively contaminated territories.