Population specific and up to date cardiovascular risk charts can be efficiently obtained with record linkage of routine and observational data.

BACKGROUND: Only few countries have cohorts enabling specific and up-to-date cardiovascular disease (CVD) risk estimation. Individual risk assessment based on study samples that differ too much from the target population could jeopardize the benefit of risk charts in general practice. Our aim was to provide up-to-date and valid CVD risk estimation for a Swiss population using a novel record linkage approach. METHODS: Anonymous record linkage was used to follow-up (for mortality, until 2008) 9,853 men and women aged 25-74 years who participated in the Swiss MONICA (MONItoring of trends and determinants in CVD) study of 1983-92. The linkage success was 97.8%, loss to follow-up 1990-2000 was 4.7%. Based on the ESC SCORE methodology (Weibull regression), we used age, sex, blood pressure, smoking, and cholesterol to generate three models. We compared the 1) original SCORE model with a 2) recalibrated and a 3) new model using the Brier score (BS) and cross-validation. RESULTS: Based on the cross-validated BS, the new model (BS = 14107×10(-6)) was somewhat more appropriate for risk estimation than the original (BS = 14190×10(-6)) and the recalibrated (BS = 14172×10(-6)) model. Particularly at younger age, derived absolute risks were consistently lower than those from the original and the recalibrated model which was mainly due to a smaller impact of total cholesterol. CONCLUSION: Using record linkage of observational and routine data is an efficient procedure to obtain valid and up-to-date CVD risk estimates for a specific population.

Classifying minimally disabled multiple sclerosis patients from resting state functional connectivity.

Multiple sclerosis (MS), a variable and diffuse disease affecting white and gray matter, is known to cause functional connectivity anomalies in patients. However, related studies published to-date are post hoc; our hypothesis was that such alterations could discriminate between patients and healthy controls in a predictive setting, laying the groundwork for imaging-based prognosis. Using functional magnetic resonance imaging resting state data of 22 minimally disabled MS patients and 14 controls, we developed a predictive model of connectivity alterations in MS: a whole-brain connectivity matrix was built for each subject from the slow oscillations (<0.11Hz) of region-averaged time series, and a pattern recognition technique was used to learn a discriminant function indicating which particular functional connections are most affected by disease. Classification performance using strict cross-validation yielded a sensitivity of 82% (above chance at p<0.005) and specificity of 86% (p<0.01) to distinguish between MS patients and controls. The most discriminative connectivity changes were found in subcortical and temporal regions, and contralateral connections were more discriminative than ipsilateral connections. The pattern of decreased discriminative connections can be summarized post hoc in an index that correlates positively (ρ=0.61) with white matter lesion load, possibly indicating functional reorganisation to cope with increasing lesion load. These results are consistent with a subtle but widespread impact of lesions in white matter and in gray matter structures serving as high-level integrative hubs. These findings suggest that predictive models of resting state fMRI can reveal specific anomalies due to MS with high sensitivity and specificity, potentially leading to new non-invasive markers.

PAX3/FOXO1 fusion gene status is the key prognostic molecular marker in rhabdomyosarcoma and significantly improves current risk stratification.

PURPOSE: To improve the risk stratification of patients with rhabdomyosarcoma (RMS) through the use of clinical and molecular biologic data. PATIENTS AND METHODS: Two independent data sets of gene-expression profiling for 124 and 101 patients with RMS were used to derive prognostic gene signatures by using a meta-analysis. These and a previously published metagene signature were evaluated by using cross validation analyses. A combined clinical and molecular risk-stratification scheme that incorporated the PAX3/FOXO1 fusion gene status was derived from 287 patients with RMS and evaluated. RESULTS: We showed that our prognostic gene-expression signature and the one previously published performed well with reproducible and significant effects. However, their effect was reduced when cross validated or tested in independent data and did not add new prognostic information over the fusion gene status, which is simpler to assay. Among nonmetastatic patients, patients who were PAX3/FOXO1 positive had a significantly poorer outcome compared with both alveolar-negative and PAX7/FOXO1-positive patients. Furthermore, a new clinicomolecular risk score that incorporated fusion gene status (negative and PAX3/FOXO1 and PAX7/FOXO1 positive), Intergroup Rhabdomyosarcoma Study TNM stage, and age showed a significant increase in performance over the current risk-stratification scheme. CONCLUSION: Gene signatures can improve current stratification of patients with RMS but will require complex assays to be developed and extensive validation before clinical application. A significant majority of their prognostic value was encapsulated by the fusion gene status. A continuous risk score derived from the combination of clinical parameters with the presence or absence of PAX3/FOXO1 represents a robust approach to improving current risk-adapted therapy for RMS.

Testing a new multigroup inference approach to reconstructing past environmental conditions

A new, quantitative, inference model for environmental reconstruction (transfer function), based for the first time on the simultaneous analysis of multigroup species, has been developed. Quantitative reconstructions based on palaeoecological transfer functions provide a powerful tool for addressing questions of environmental change in a wide range of environments, from oceans to mountain lakes, and over a range of timescales, from decades to millions of years. Much progress has been made in the development of inferences based on multiple proxies but usually these have been considered separately, and the different numeric reconstructions compared and reconciled post-hoc. This paper presents a new method to combine information from multiple biological groups at the reconstruction stage. The aim of the multigroup work was to test the potential of the new approach to making improved inferences of past environmental change by improving upon current reconstruction methodologies. The taxonomic groups analysed include diatoms, chironomids and chrysophyte cysts. We test the new methodology using two cold-environment training-sets, namely mountain lakes from the Pyrenees and the Alps. The use of multiple groups, as opposed to single groupings, was only found to increase the reconstruction skill slightly, as measured by the root mean square error of prediction (leave-one-out cross-validation), in the case of alkalinity, dissolved inorganic carbon and altitude (a surrogate for air-temperature), but not for pH or dissolved CO2. Reasons why the improvement was less than might have been anticipated are discussed. These can include the different life-forms, environmental responses and reaction times of the groups under study.

Automatic Segmentation of the Eye in 3D Magnetic Resonance Imaging: A Novel Statistical Shape Model for Treatment Planning of Retinoblastoma.

PURPOSE: Proper delineation of ocular anatomy in 3-dimensional (3D) imaging is a big challenge, particularly when developing treatment plans for ocular diseases. Magnetic resonance imaging (MRI) is presently used in clinical practice for diagnosis confirmation and treatment planning for treatment of retinoblastoma in infants, where it serves as a source of information, complementary to the fundus or ultrasonographic imaging. Here we present a framework to fully automatically segment the eye anatomy for MRI based on 3D active shape models (ASM), and we validate the results and present a proof of concept to automatically segment pathological eyes. METHODS AND MATERIALS: Manual and automatic segmentation were performed in 24 images of healthy children's eyes (3.29 ± 2.15 years of age). Imaging was performed using a 3-T MRI scanner. The ASM consists of the lens, the vitreous humor, the sclera, and the cornea. The model was fitted by first automatically detecting the position of the eye center, the lens, and the optic nerve, and then aligning the model and fitting it to the patient. We validated our segmentation method by using a leave-one-out cross-validation. The segmentation results were evaluated by measuring the overlap, using the Dice similarity coefficient (DSC) and the mean distance error. RESULTS: We obtained a DSC of 94.90 ± 2.12% for the sclera and the cornea, 94.72 ± 1.89% for the vitreous humor, and 85.16 ± 4.91% for the lens. The mean distance error was 0.26 ± 0.09 mm. The entire process took 14 seconds on average per eye. CONCLUSION: We provide a reliable and accurate tool that enables clinicians to automatically segment the sclera, the cornea, the vitreous humor, and the lens, using MRI. We additionally present a proof of concept for fully automatically segmenting eye pathology. This tool reduces the time needed for eye shape delineation and thus can help clinicians when planning eye treatment and confirming the extent of the tumor.

ASTRAL-R score predicts non-recanalisation after intravenous thrombolysis in acute ischaemic stroke.

Intravenous thrombolysis (IVT) as treatment in acute ischaemic strokes may be insufficient to achieve recanalisation in certain patients. Predicting probability of non-recanalisation after IVT may have the potential to influence patient selection to more aggressive management strategies. We aimed at deriving and internally validating a predictive score for post-thrombolytic non-recanalisation, using clinical and radiological variables. In thrombolysis registries from four Swiss academic stroke centres (Lausanne, Bern, Basel and Geneva), patients were selected with large arterial occlusion on acute imaging and with repeated arterial assessment at 24 hours. Based on a logistic regression analysis, an integer-based score for each covariate of the fitted multivariate model was generated. Performance of integer-based predictive model was assessed by bootstrapping available data and cross validation (delete-d method). In 599 thrombolysed strokes, five variables were identified as independent predictors of absence of recanalisation: Acute glucose > 7 mmol/l (A), significant extracranial vessel STenosis (ST), decreased Range of visual fields (R), large Arterial occlusion (A) and decreased Level of consciousness (L). All variables were weighted 1, except for (L) which obtained 2 points based on β-coefficients on the logistic scale. ASTRAL-R scores 0, 3 and 6 corresponded to non-recanalisation probabilities of 18, 44 and 74 % respectively. Predictive ability showed AUC of 0.66 (95 %CI, 0.61-0.70) when using bootstrap and 0.66 (0.63-0.68) when using delete-d cross validation. In conclusion, the 5-item ASTRAL-R score moderately predicts non-recanalisation at 24 hours in thrombolysed ischaemic strokes. If its performance can be confirmed by external validation and its clinical usefulness can be proven, the score may influence patient selection for more aggressive revascularisation strategies in routine clinical practice.

Accuracy and Precision of Head Motion Information in Multi-Channel Free Induction Decay Navigators for Magnetic Resonance Imaging.

Free induction decay (FID) navigators were found to qualitatively detect rigid-body head movements, yet it is unknown to what extent they can provide quantitative motion estimates. Here, we acquired FID navigators at different sampling rates and simultaneously measured head movements using a highly accurate optical motion tracking system. This strategy allowed us to estimate the accuracy and precision of FID navigators for quantification of rigid-body head movements. Five subjects were scanned with a 32-channel head coil array on a clinical 3T MR scanner during several resting and guided head movement periods. For each subject we trained a linear regression model based on FID navigator and optical motion tracking signals. FID-based motion model accuracy and precision was evaluated using cross-validation. FID-based prediction of rigid-body head motion was found to be with a mean translational and rotational error of 0.14±0.21 mm and 0.08±0.13(°) , respectively. Robust model training with sub-millimeter and sub-degree accuracy could be achieved using 100 data points with motion magnitudes of ±2 mm and ±1(°) for translation and rotation. The obtained linear models appeared to be subject-specific as inter-subject application of a "universal" FID-based motion model resulted in poor prediction accuracy. The results show that substantial rigid-body motion information is encoded in FID navigator signal time courses. Although, the applied method currently requires the simultaneous acquisition of FID signals and optical tracking data, the findings suggest that multi-channel FID navigators have a potential to complement existing tracking technologies for accurate rigid-body motion detection and correction in MRI.

Prediction of multiple infections after severe burn trauma: a prospective cohort study.

OBJECTIVE: To develop predictive models for early triage of burn patients based on hypersusceptibility to repeated infections. BACKGROUND: Infection remains a major cause of mortality and morbidity after severe trauma, demanding new strategies to combat infections. Models for infection prediction are lacking. METHODS: Secondary analysis of 459 burn patients (≥16 years old) with 20% or more total body surface area burns recruited from 6 US burn centers. We compared blood transcriptomes with a 180-hour cutoff on the injury-to-transcriptome interval of 47 patients (≤1 infection episode) to those of 66 hypersusceptible patients [multiple (≥2) infection episodes (MIE)]. We used LASSO regression to select biomarkers and multivariate logistic regression to built models, accuracy of which were assessed by area under receiver operating characteristic curve (AUROC) and cross-validation. RESULTS: Three predictive models were developed using covariates of (1) clinical characteristics; (2) expression profiles of 14 genomic probes; (3) combining (1) and (2). The genomic and clinical models were highly predictive of MIE status [AUROCGenomic = 0.946 (95% CI: 0.906-0.986); AUROCClinical = 0.864 (CI: 0.794-0.933); AUROCGenomic/AUROCClinical P = 0.044]. Combined model has an increased AUROCCombined of 0.967 (CI: 0.940-0.993) compared with the individual models (AUROCCombined/AUROCClinical P = 0.0069). Hypersusceptible patients show early alterations in immune-related signaling pathways, epigenetic modulation, and chromatin remodeling. CONCLUSIONS: Early triage of burn patients more susceptible to infections can be made using clinical characteristics and/or genomic signatures. Genomic signature suggests new insights into the pathophysiology of hypersusceptibility to infection may lead to novel potential therapeutic or prophylactic targets.

New kernel functions and learning methods for text and data mining

Recent advances in machine learning methods enable increasingly the automatic construction of various types of computer assisted methods that have been difficult or laborious to program by human experts. The tasks for which this kind of tools are needed arise in many areas, here especially in the fields of bioinformatics and natural language processing. The machine learning methods may not work satisfactorily if they are not appropriately tailored to the task in question. However, their learning performance can often be improved by taking advantage of deeper insight of the application domain or the learning problem at hand. This thesis considers developing kernel-based learning algorithms incorporating this kind of prior knowledge of the task in question in an advantageous way. Moreover, computationally efficient algorithms for training the learning machines for specific tasks are presented. In the context of kernel-based learning methods, the incorporation of prior knowledge is often done by designing appropriate kernel functions. Another well-known way is to develop cost functions that fit to the task under consideration. For disambiguation tasks in natural language, we develop kernel functions that take account of the positional information and the mutual similarities of words. It is shown that the use of this information significantly improves the disambiguation performance of the learning machine. Further, we design a new cost function that is better suitable for the task of information retrieval and for more general ranking problems than the cost functions designed for regression and classification. We also consider other applications of the kernel-based learning algorithms such as text categorization, and pattern recognition in differential display. We develop computationally efficient algorithms for training the considered learning machines with the proposed kernel functions. We also design a fast cross-validation algorithm for regularized least-squares type of learning algorithm. Further, an efficient version of the regularized least-squares algorithm that can be used together with the new cost function for preference learning and ranking tasks is proposed. In summary, we demonstrate that the incorporation of prior knowledge is possible and beneficial, and novel advanced kernels and cost functions can be used in algorithms efficiently.

Equating accelerometer estimates among youth: The Rosetta Stone 2.

OBJECTIVES: Different accelerometer cutpoints used by different researchers often yields vastly different estimates of moderate-to-vigorous intensity physical activity (MVPA). This is recognized as cutpoint non-equivalence (CNE), which reduces the ability to accurately compare youth MVPA across studies. The objective of this research is to develop a cutpoint conversion system that standardizes minutes of MVPA for six different sets of published cutpoints. DESIGN: Secondary data analysis. METHODS: Data from the International Children's Accelerometer Database (ICAD; Spring 2014) consisting of 43,112 Actigraph accelerometer data files from 21 worldwide studies (children 3-18 years, 61.5% female) were used to develop prediction equations for six sets of published cutpoints. Linear and non-linear modeling, using a leave one out cross-validation technique, was employed to develop equations to convert MVPA from one set of cutpoints into another. Bland Altman plots illustrate the agreement between actual MVPA and predicted MVPA values. RESULTS: Across the total sample, mean MVPA ranged from 29.7MVPAmind(-1) (Puyau) to 126.1MVPAmind(-1) (Freedson 3 METs). Across conversion equations, median absolute percent error was 12.6% (range: 1.3 to 30.1) and the proportion of variance explained ranged from 66.7% to 99.8%. Mean difference for the best performing prediction equation (VC from EV) was -0.110mind(-1) (limits of agreement (LOA), -2.623 to 2.402). The mean difference for the worst performing prediction equation (FR3 from PY) was 34.76mind(-1) (LOA, -60.392 to 129.910). CONCLUSIONS: For six different sets of published cutpoints, the use of this equating system can assist individuals attempting to synthesize the growing body of literature on Actigraph, accelerometry-derived MVPA.

Geographic assignment of seabirds to their origin: combining morphology, genetics and biogeochemical analyses

Longline fisheries, oil spills, and offshore wind farms are some of the major threats increasing seabird mortality at sea, but the impact of these threats on specific populations has been difficult to determine so far. We tested the use of molecular markers, morphometric measures, and stable isotope (δ15N and δ13C) and trace element concentrations in the first primary feather (grown at the end of the breeding period) to assign the geographic origin of Calonectris shearwaters. Overall, we sampled birds from three taxa: 13 Mediterranean Cory's Shearwater (Calonectris diomedea diomedea) breeding sites, 10 Atlantic Cory's Shearwater (Calonectris diomedea borealis) breeding sites, and one Cape Verde Shearwater (C. edwardsii) breeding site. Assignment rates were investigated at three spatial scales: breeding colony, breeding archipelago, and taxa levels. Genetic analyses based on the mitochondrial control region (198 birds from 21 breeding colonies) correctly assigned 100% of birds to the three main taxa but failed in detecting geographic structuring at lower scales. Discriminant analyses based on trace elements composition achieved the best rate of correct assignment to colony (77.5%). Body measurements or stable isotopes mainly succeeded in assigning individuals among taxa (87.9% and 89.9%, respectively) but failed at the colony level (27.1% and 38.0%, respectively). Combining all three approaches (morphometrics, isotopes, and trace elements on 186 birds from 15 breeding colonies) substantially improved correct classifications (86.0%, 90.7%, and 100% among colonies, archipelagos, and taxa, respectively). Validations using two independent data sets and jackknife cross-validation confirmed the robustness of the combined approach in the colony assignment (62.5%, 58.8%, and 69.8% for each validation test, respectively). A preliminary application of the discriminant model based on stable isotope δ15N and δ13C values and trace elements (219 birds from 17 breeding sites) showed that 41 Cory's Shearwaters caught by western Mediterranean long-liners came mainly from breeding colonies in Menorca (48.8%), Ibiza (14.6%), and Crete (31.7%). Our findings show that combining analyses of trace elements and stable isotopes on feathers can achieve high rates of correct geographic assignment of birds in the marine environment, opening new prospects for the study of seabird mortality at sea.

Quimiometria I: calibração multivariada, um tutorial

The aim of this work is to present a tutorial on Multivariate Calibration, a tool which is nowadays necessary in basically most laboratories but very often misused. The basic concepts of preprocessing, principal component analysis (PCA), principal component regression (PCR) and partial least squares (PLS) are given. The two basic steps on any calibration procedure: model building and validation are fully discussed. The concepts of cross validation (to determine the number of factors to be used in the model), leverage and studentized residuals (to detect outliers) for the validation step are given. The whole calibration procedure is illustrated using spectra recorded for ternary mixtures of 2,4,6 trinitrophenolate, 2,4 dinitrophenolate and 2,5 dinitrophenolate followed by the concentration prediction of these three chemical species during a diffusion experiment through a hydrophobic liquid membrane. MATLAB software is used for numerical calculations. Most of the commands for the analysis are provided in order to allow a non-specialist to follow step by step the analysis.

Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

DWI and complex brain network analysis predicts vascular cognitive impairment in spontaneous hypertensive rats undergoing executive function tests

The identification of biomarkers of vascular cognitive impairment is urgent for its early diagnosis. The aim of this study was to detect and monitor changes in brain structure and connectivity, and to correlate them with the decline in executive function. We examined the feasibility of early diagnostic magnetic resonance imaging (MRI) to predict cognitive impairment before onset in an animal model of chronic hypertension: Spontaneously Hypertensive Rats. Cognitive performance was tested in an operant conditioning paradigm that evaluated learning, memory, and behavioral flexibility skills. Behavioral tests were coupled with longitudinal diffusion weighted imaging acquired with 126 diffusion gradient directions and 0.3 mm(3) isometric resolution at 10, 14, 18, 22, 26, and 40 weeks after birth. Diffusion weighted imaging was analyzed in two different ways, by regional characterization of diffusion tensor imaging (DTI) indices, and by assessing changes in structural brain network organization based on Q-Ball tractography. Already at the first evaluated times, DTI scalar maps revealed significant differences in many regions, suggesting loss of integrity in white and gray matter of spontaneously hypertensive rats when compared to normotensive control rats. In addition, graph theory analysis of the structural brain network demonstrated a significant decrease of hierarchical modularity, global and local efficacy, with predictive value as shown by regional three-fold cross validation study. Moreover, these decreases were significantly correlated with the behavioral performance deficits observed at subsequent time points, suggesting that the diffusion weighted imaging and connectivity studies can unravel neuroimaging alterations even overt signs of cognitive impairment become apparent.

Application of GA-PLS and GA-KPLS calculations for the prediction of the retention indices of essential oils

Genetic algorithm and partial least square (GA-PLS) and kernel PLS (GA-KPLS) techniques were used to investigate the correlation between retention indices (RI) and descriptors for 117 diverse compounds in essential oils from 5 Pimpinella species gathered from central Turkey which were obtained by gas chromatography and gas chromatography-mass spectrometry. The square correlation coefficient leave-group-out cross validation (LGO-CV) (Q²) between experimental and predicted RI for training set by GA-PLS and GA-KPLS was 0.940 and 0.963, respectively. This indicates that GA-KPLS can be used as an alternative modeling tool for quantitative structure-retention relationship (QSRR) studies.