Uncertainty Quantification with Support Vector Regression Prediction Models

Uncertainty quantification of petroleum reservoir models is one of the present challenges, which is usually approached with a wide range of geostatistical tools linked with statistical optimisation or/and inference algorithms. The paper considers a data driven approach in modelling uncertainty in spatial predictions. Proposed semi-supervised Support Vector Regression (SVR) model has demonstrated its capability to represent realistic features and describe stochastic variability and non-uniqueness of spatial properties. It is able to capture and preserve key spatial dependencies such as connectivity, which is often difficult to achieve with two-point geostatistical models. Semi-supervised SVR is designed to integrate various kinds of conditioning data and learn dependences from them. A stochastic semi-supervised SVR model is integrated into a Bayesian framework to quantify uncertainty with multiple models fitted to dynamic observations. The developed approach is illustrated with a reservoir case study. The resulting probabilistic production forecasts are described by uncertainty envelopes.

Brain Tissue Hypoxia is a Strong Predictor of Outcome after Severe Traumatic Brain Injury Independent from Elevated Intracranial Pressure

Introduction: Low brain tissue oxygen pressure (PbtO2) is associated with worse outcome in patients with severe traumatic brain injury (TBI). However, it is unclear whether brain tissue hypoxia is merely a marker of injury severity or a predictor of prognosis, independent from intracranial pressure (ICP) and injury severity. Hypothesis: We hypothesized that brain tissue hypoxia was an independent predictor of outcome in patients wih severe TBI, irrespective of elevated ICP and of the severity of cerebral and systemic injury. Methods: This observational study was conducted at the Neurological ICU, Hospital of the University of Pennsylvania, an academic level I trauma center. Patients admitted with severe TBI who had PbtO2 and ICP monitoring were included in the study. PbtO2, ICP, mean arterial pressure (MAP) and cerebral perfusion pressure (CPP = MAP-ICP) were monitored continuously and recorded prospectively every 30 min. Using linear interpolation, duration and cumulative dose (area under the curve, AUC) of brain tissue hypoxia (PbtO2 < 15 mm Hg), elevated ICP >20 mm Hg and low CPP <60 mm Hg were calculated, and the association with outcome at hospital discharge, dichotomized as good (Glasgow Outcome Score [GOS] 4-5) vs. poor (GOS 1-3), was analyzed. Results: A total of 103 consecutive patients, monitored for an average of 5 days, was studied. Brain tissue hypoxia was observed in 66 (64%) patients despite ICP was < 20 mm Hg and CPP > 60 mm Hg (72 +/- 39% and 49 +/- 41% of brain hypoxic time, respectively). Compared with patients with good outcome, those with poor outcome had a longer duration of brain hypoxia (1.7 +/- 3.7 vs. 8.3 +/- 15.9 hrs, P<0.01), as well as a longer duration (11.5 +/- 16.5 vs. 21.6 +/- 29.6 hrs, P=0.03) and a greater cumulative dose (56 +/- 93 vs. 143 +/- 218 mm Hg*hrs, P<0.01) of elevated ICP. By multivariable logistic regression, admission Glasgow Coma Scale (OR, 0.83, 95% CI: 0.70-0.99, P=0.04), Marshall CT score (OR 2.42, 95% CI: 1.42-4.11, P<0.01), APACHE II (OR 1.20, 95% CI: 1.03-1.43, P=0.03), and the duration of brain tissue hypoxia (OR 1.13; 95% CI: 1.01-1.27; P=0.04) were all significantly associated with poor outcome. No independent association was found between the AUC for elevated ICP and outcome (OR 1.01, 95% CI 0.97-1.02, P=0.11) in our prospective cohort. Conclusions: In patients with severe TBI, brain tissue hypoxia is frequent, despite normal ICP and CPP, and is associated with poor outcome, independent of intracranial hypertension and the severity of cerebral and systemic injury. Our findings indicate that PbtO2 is a strong physiologic prognostic marker after TBI. Further study is warranted to examine whether PbtO2-directed therapy improves outcome in severely head-injured patients .

INTERMED predicts non-return to work in an occupational rehabilitation setting for individuals with orthopaedic trauma-Part I

Introduction.- Knowledge of predictors of an unfavourable outcome, e.g. non-return to work after an injury enables to identify patients at risk and to target interventions for modifiable predictors. It has been recently shown that INTERMED; a tool to measure biopsychosocial complexity in four domains (biologic, psychologic, social and care, with a score between 0-60 points) can be useful in this context. The aim of this study was to set up a predictive model for non-return to work using INTERMED in patients in vocational rehabilitation after orthopaedic injury.Patients and methods.- In this longitudinal prospective study, the cohort consisted of 2156 consecutively included inpatients with orthopaedic trauma attending a rehabilitation hospital after a work, traffic or sport related injury. Two years after discharge, a questionnaire regarding return to work was sent (1502 returned their questionnaires). In addition to INTERMED, 18 predictors known at baseline of the rehabilitation were selected based on previous research. A multivariable logistic regression was performed.Results.- In the multivariate model, not-returning to work at 2 years was significantly predicted by the INTERMED: odds-ratio (OR) 1.08 (95% confidence interval, CI [1.06; 1.11]) for a one point increase in scale; by qualified work-status before the injury OR = 0.74, CI (0.54; 0.99), by using French as preferred language OR = 0.60, CI (0.45; 0.80), by upper-extremity injury OR = 1.37, CI (1.03; 1.81), by higher education (> 9 years) OR = 0.74, CI (0.55; 1.00), and by a 10 year increase in age OR = 1.15, CI (1.02; 1.29). The area under the receiver-operator-characteristics curve (ROC)-curve was 0.733 for the full model (INTERMED plus 18 variables).Discussion.- These results confirm that the total score of the INTERMED is a significant predictor for return to work. The full model with 18 predictors combined with the total score of INTERMED has good predictive value. However, the number of variables (19) to measure is high for the use as screening tool in a clinic.

The essential role of radiotherapy in the treatment of Merkel cell carcinoma: a study from the Rare Cancer Network.

PURPOSE: To evaluate the role of postoperative radiotherapy (RT) in Merkel cell carcinoma (MCC). METHODS AND MATERIALS: A retrospective multicenter study was performed in 180 patients with MCC treated between February 1988 and September 2009. Patients who had had surgery alone were compared with patients who received surgery and postoperative RT or radical RT. Local relapse-free survival (LRFS), regional relapse-free survival (RRFS), and distant metastasis-free survival (DMFS) rates were assessed together with disease-free survival (DFS), cancer-specific survival (CSS), and overall survival (OS) rates. RESULTS: Seventy-nine patients were male and 101 patients were female, and the median age was 73 years old (range, 38-93 years). The majority of patients had localized disease (n = 146), and the remaining patients had regional lymph node metastasis (n = 34). Forty-nine patients underwent surgery for the primary tumor without postoperative RT to the primary site; the other 131 patients received surgery for the primary tumor, followed by postoperative RT (n = 118) or a biopsy of the primary tumor followed by radical RT (n = 13). Median follow-up was 5 years (range, 0.2-16.5 years). Patients in the RT group had improved LRFS (93% vs. 64%; p < 0.001), RRFS (76% vs. 27%; p < 0.001), DMFS (70% vs. 42%; p = 0.01), DFS (59% vs. 4%; p < 0.001), and CSS (65% vs. 49%; p = 0.03) rates compared to patients who underwent surgery for the primary tumor alone; LRFS, RRFS, DMFS, and DFS rates remained significant with multivariable Cox regression analysis. However OS was not significantly improved by postoperative RT (56% vs. 46%; p = 0.2). CONCLUSIONS: After multivariable analysis, postoperative RT was associated with improved outcome and seems to be an important component in the multimodality treatment of MCC.

Complete and long-lasting regression of disseminated multiple skin melanoma metastases under treatment with cyclooxygenase-2 inhibitor.

Experimental and clinical evidence indicates that non-steroidal anti-inflammatory drugs and cyclooxygenase-2 inhibitors may have anti-cancer activities. Here we report on a patient with a metastatic melanoma of the leg who experienced a complete and sustained regression of skin metastases upon continuous single treatment with the cyclooxygenase-2 inhibitor rofecoxib. Our observations indicate that the inhibition of cyclooxygenase-2 can lead to the regression of disseminated skin melanoma metastases, even after failure of chemotherapy.

What is the influence of vaccination's routes on the regression of tumors located at mucosal sites?

Tumor-regressions following tumor-associated-antigen vaccination in animal models contrast with the limited clinical outcomes in cancer patients. Most animal studies however used subcutaneous-tumor-models and questions arise as whether these are relevant for tumors growing in mucosae; whether specific mucosal-homing instructions are required; and how this may be influenced by the tumor.

Bootstrapping pairs in Distance-Based Regression

La regressió basada en distàncies és un mètode de predicció que consisteix en dos passos: a partir de les distàncies entre observacions obtenim les variables latents, les quals passen a ser els regressors en un model lineal de mínims quadrats ordinaris. Les distàncies les calculem a partir dels predictors originals fent us d'una funció de dissimilaritats adequada. Donat que, en general, els regressors estan relacionats de manera no lineal amb la resposta, la seva selecció amb el test F usual no és possible. En aquest treball proposem una solució a aquest problema de selecció de predictors definint tests estadístics generalitzats i adaptant un mètode de bootstrap no paramètric per a l'estimació dels p-valors. Incluim un exemple numèric amb dades de l'assegurança d'automòbils.

Geomodelling of a fluvial system with semi-supervised support vector regression

Fluvial deposits are a challenge for modelling flow in sub-surface reservoirs. Connectivity and continuity of permeable bodies have a major impact on fluid flow in porous media. Contemporary object-based and multipoint statistics methods face a problem of robust representation of connected structures. An alternative approach to model petrophysical properties is based on machine learning algorithm ? Support Vector Regression (SVR). Semi-supervised SVR is able to establish spatial connectivity taking into account the prior knowledge on natural similarities. SVR as a learning algorithm is robust to noise and captures dependencies from all available data. Semi-supervised SVR applied to a synthetic fluvial reservoir demonstrated robust results, which are well matched to the flow performance

Hospitalized women experiencing an episode of excessive oral anticoagulation had a higher bleeding risk than men.

OBJECTIVE: Overanticoagulated medical inpatients may be particularly prone to bleeding complications. Among medical inpatients with excessive oral anticoagulation (AC), we sought to identify patient and treatment factors associated with bleeding. METHODS: We prospectively identified consecutive patients receiving oral AC admitted to the medical ward of a university hospital (February-July 2006) who had at least one international normalized ratio (INR) value >3.0 during the hospital stay. We recorded patient characteristics, AC-related factors, and concomitant treatments (e.g., platelet inhibitors) that increase the bleeding risk. The outcome was overall bleeding, defined as the occurrence of major or minor bleeding during the hospital stay. We used logistic regression to explore patient and treatment factors associated with bleeding. RESULTS: Overall, 145 inpatients with excessive oral AC comprised our study sample. Atrial fibrillation (59%) and venous thromboembolism (28%) were the most common indications for AC. Twelve patients (8.3%) experienced a bleeding event. Of these, 8 had major bleeding. Women had a somewhat higher risk of major bleeding than men (12.5% vs 4.1%, p = 0.08). Multivariable analysis demonstrated that female gender was independently associated with bleeding (odds ratio [OR] 4.3, 95% confidence interval [95% C1] 1.1-17.8). Age, history of major bleeding, value of the index INR, and concomitant treatment with platelet inhibitors were not independent predictors of bleeding. CONCLUSIONS: We found that hospitalized women experiencing an episode of excessive oral AC have a 4-fold increased risk of bleeding compared with men. Whether overanticoagulated women require more aggressive measures of AC reversal must be examined in further studies.

Higher-order colored-noise effects in multivariable systems

We extend the partial resummation technique of Fokker-Planck terms for multivariable stochastic differential equations with colored noise. As an example, a model system of a Brownian particle with colored noise is studied. We prove that the asymmetric behavior found in analog simulations is due to higher-order terms which are left out in that technique. On the contrary, the systematic ¿-expansion approach can explain the analog results.

Modelling non-industrial private forest landowners' strategic decision making by using logistic regression and neural networks: case of predicting the choice of forest taxation basis

Summary

Neurological Prognostication after Cardiac Arrest and Therapeutic Hypothermia: a Prospective Validation Study

Introduction: Clinical examination and electroencephalography study (EEG) have been recommended to predict functional recovery in comatose survivors of cardiac arrest (CA), however their prognostic value in patients treated with induced hypothermia (IH) has not been evaluated. Hypothesis: We aimed to validate the prognostic ability of clinical examination and EEG in predicting outcome of patients with coma after CA treated with IH and sought to derive a score with high predictive value for poor functional outcome in this setting. Methods: We prospectively studied 100 consecutive comatose survivors of CA treated with IH. Repeated neurological examination and EEG were performed early after passive rewarming and off sedation. Mortality was assessed at hospital discharge, and functional outcome at 3 to 6 months with Cerebral Performance Categories (CPC), and was dichotomized as good (CPC 1-2) vs. poor (CPC 3-5). Independent predictors of outcome were identified by multivariable logistic regression and used to assess the prognostic value of a Reproducible Electro-clinical Prognosticators of Outcome Score (REPOS). Results: Patients (20/100) with good outcome had all a reactive EEG background. Incomplete recovery of brainstem reflexes, myoclonus, time to return of spontaneous circulation (ROSC) > 25 min, and unreactive EEG background were all independent predictors of death and severe disability, and were added to construct the REPOS. Using a cut-off of 0 or 1 variables for good vs. 2 to 4 for poor outcome, the REPOS had a positive predictive value of 1.00 (95% CI: 0.92-1.00), a negative predictive value of 0.43 (95% CI: 0.29-0.58) and an accuracy of 0.81 for poor functional recovery at 3 to 6 months. Conclusions: In comatose survivors of CA treated with IH, a prognostic score, including clinical and EEG examination, was highly predictive of death and poor functional outcome at 3 to 6 months. Lack of EEG background reactivity strongly predicted poor neurological recovery after CA. Our findings show that clinical and electrophysiological studies are effective in predicting long-term outcome of comatose survivors after CA and IH, and suggest that EEG improves early prognostic assessment in the setting of therapeutic cooling.

Retrospective feasibility study of simultaneous integrated boost in cervical cancer using Tomotherapy: the impact of organ motion and tumor regression.

BACKGROUND: Whole pelvis intensity modulated radiotherapy (IMRT) is increasingly being used to treat cervical cancer aiming to reduce side effects. Encouraged by this, some groups have proposed the use of simultaneous integrated boost (SIB) to target the tumor, either to get a higher tumoricidal effect or to replace brachytherapy. Nevertheless, physiological organ movement and rapid tumor regression throughout treatment might substantially reduce any benefit of this approach. PURPOSE: To evaluate the clinical target volume - simultaneous integrated boost (CTV-SIB) regression and motion during chemo-radiotherapy (CRT) for cervical cancer, and to monitor treatment progress dosimetrically and volumetrically to ensure treatment goals are met. METHODS AND MATERIALS: Ten patients treated with standard doses of CRT and brachytherapy were retrospectively re-planned using a helical Tomotherapy - SIB technique for the hypothetical scenario of this feasibility study. Target and organs at risk (OAR) were contoured on deformable fused planning-computed tomography and megavoltage computed tomography images. The CTV-SIB volume regression was determined. The center of mass (CM) was used to evaluate the degree of motion. The Dice's similarity coefficient (DSC) was used to assess the spatial overlap of CTV-SIBs between scans. A cumulative dose-volume histogram modeled estimated delivered doses. RESULTS: The CTV-SIB relative reduction was between 31 and 70%. The mean maximum CM change was 12.5, 9, and 3 mm in the superior-inferior, antero-posterior, and right-left dimensions, respectively. The CTV-SIB-DSC approached 1 in the first week of treatment, indicating almost perfect overlap. CTV-SIB-DSC regressed linearly during therapy, and by the end of treatment was 0.5, indicating 50% discordance. Two patients received less than 95% of the prescribed dose. Much higher doses to the OAR were observed. A multiple regression analysis showed a significant interaction between CTV-SIB reduction and OAR dose increase. CONCLUSIONS: The CTV-SIB had important regression and motion during CRT, receiving lower therapeutic doses than expected. The OAR had unpredictable shifts and received higher doses. The use of SIB without frequent adaptation of the treatment plan exposes cervical cancer patients to an unpredictable risk of under-dosing the target and/or overdosing adjacent critical structures. In that scenario, brachytherapy continues to be the gold standard approach.

Socio-demographic and behavioural determinants of weight gain in the Swiss population.

BackgroundIn Switzerland, socio-demographic and behavioural factors are associated with obesity, but no study ever assessed their impact on weight gain using prospective data.MethodsData from 4,469 participants (53.0% women), aged 35 to 75 years at baseline and followed for 5.5 years. Weight gain was considered as a rate (kg/year) or as gaining ¿5 kg during the study period.ResultsRate of weight gain was lower among participants who were older (mean¿±¿standard deviation: 0.46¿±¿0.92, 0.33¿±¿0.88, 0.21¿±¿0.86 and 0.06¿±¿0.74 kg/year in participants aged [35-45[, [45-55[, [55¿65[and [65+ years, respectively, P<0.001); physically active (0.27¿±¿0.82 vs. 0.35¿±¿0.95 kg/year for sedentary, P¿<¿0.005) or living in a couple (0.29¿±¿0.84 vs. 0.35¿±¿0.96 kg/year for living single, P¿<¿0.05), and higher among current smokers (0.41¿±¿0.97, 0.26¿±¿0.84 and 0.29±0.85 kg/year for current, former and never smokers, respectively, p<0.001). These findings were further confirmed by multivariable analysis. Multivariable logistic regression showed that receiving social help, being a current smoker or obese increased the likelihood of gaining ¿5Kg: Odds ratio (OR) and 95% confidence interval (CI) 1.43 (1.16-1.77); 1.63 (1.35-1.95) and 1.95 (1.57-2.43), respectively, while living in couple or being physically active decreased the risk: 0.73 (0.62-0.86) and 0.72 (0.62-0.83), respectively. No association was found between weight gain and gender, being born in Switzerland or education.ConclusionsIn Switzerland, financial difficulties (indicated by receiving social help) and current smoking were associated with increases in body weight over a 5 years follow-up. Living in couple, being older or physically active were protective against weight gain.

Regression analysis of spatial data.

Many of the most interesting questions ecologists ask lead to analyses of spatial data. Yet, perhaps confused by the large number of statistical models and fitting methods available, many ecologists seem to believe this is best left to specialists. Here, we describe the issues that need consideration when analysing spatial data and illustrate these using simulation studies. Our comparative analysis involves using methods including generalized least squares, spatial filters, wavelet revised models, conditional autoregressive models and generalized additive mixed models to estimate regression coefficients from synthetic but realistic data sets, including some which violate standard regression assumptions. We assess the performance of each method using two measures and using statistical error rates for model selection. Methods that performed well included generalized least squares family of models and a Bayesian implementation of the conditional auto-regressive model. Ordinary least squares also performed adequately in the absence of model selection, but had poorly controlled Type I error rates and so did not show the improvements in performance under model selection when using the above methods. Removing large-scale spatial trends in the response led to poor performance. These are empirical results; hence extrapolation of these findings to other situations should be performed cautiously. Nevertheless, our simulation-based approach provides much stronger evidence for comparative analysis than assessments based on single or small numbers of data sets, and should be considered a necessary foundation for statements of this type in future.