Catrin O. Plumpton

Random Subspace Ensembles for fMRI Classification

Classification of brain images obtained through functional magnetic resonance imaging (fMRI) poses a serious challenge to pattern recognition and machine learning due to the extremely large feature-to-instance ratio. This calls for revision and adaptation of the current state-of-the-art classification methods. We investigate the suitability of the random subspace (RS) ensemble method for fMRI classification. RS samples from the original feature set and builds one (base) classifier on each subset. The ensemble assigns a class label by either majority voting or averaging of output probabilities. Looking for guidelines for setting the two parameters of the method-ensemble size and feature sample size-we introduce three criteria calculated through these parameters: usability of the selected feature sets, coverage of the set of ¿important¿ features, and feature set diversity. Optimized together, these criteria work toward producing accurate and diverse individual classifiers. RS was tested on three fMRI datasets from single-subject experiments: the Haxby data (Haxby, 2001.) and two datasets collected in-house. We found that RS with support vector machines (SVM) as the base classifier outperformed single classifiers as well as some of the most widely used classifier ensembles such as bagging, AdaBoost, random forest, and rotation forest. The closest rivals were the single SVM and bagging of SVM classifiers. We use kappa-error diagrams to understand the success of RS.

A Systematic Review of Economic Evaluations of Pharmacogenetic Testing for Prevention of Adverse Drug Reactions

BackgroundPharmacogenetics offers the potential to improve health outcomes by identifying individuals who are at greater risk of harm from certain medicines. Routine adoption of pharmacogenetic tests requires evidence of their cost effectiveness.ObjectiveThe present review aims to systematically review published economic evaluations of pharmacogenetic tests that aim to prevent or reduce the incidence of ADRs.MethodsWe conducted a systematic literature review of economic evaluations of pharmacogenetic tests aimed to reduce the incidence of adverse drug reactions. Literature was searched using Embase, MEDLINE and the NHS Economic Evaluation Database with search terms relating to pharmacogenetic testing, adverse drug reactions, economic evaluations and pharmaceuticals. Titles were screened independently by two reviewers. Articles deemed to meet the inclusion criteria were screened independently on abstract, and full texts reviewed.ResultsWe identified 852 articles, of which 47 met the inclusion criteria. There was evidence supporting the cost effectiveness of testing for HLA-B*57:01 (prior to abacavir), HLA-B*15:02 and HLA-A*31:01 (prior to carbamazepine), HLA-B*58:01 (prior to allopurinol) and CYP2C19 (prior to clopidogrel treatment). Economic evidence was inconclusive with respect to TPMT (prior to 6-mercaptoputine, azathioprine and cisplatin therapy), CYP2C9 and VKORC1 (to inform genotype-guided dosing of coumarin derivatives), MTHFR (prior to methotrexate treatment) and factor V Leiden testing (prior to oral contraception). Testing for A1555G is not cost effective before prescribing aminoglycosides.ConclusionsOur systematic review identified robust evidence of the cost effectiveness of genotyping prior to treatment with a number of common drugs. However, further analyses and (or) availability of robust clinical evidence is necessary to make recommendations for others.

Adaptive Learning Rate for Online Linear Discriminant Classifiers

We propose a strategy for updating the learning rate parameter of online linear classifiers for streaming data with concept drift. The change in the learning rate is guided by the change in a running estimate of the classification error. In addition, we propose an online version of the standard linear discriminant classifier (O-LDC) in which the inverse of the common covariance matrix is updated using the Sherman-Morrison-Woodbury formula. The adaptive learning rate was applied to four online linear classifier models on generated and real streaming data with concept drift. O-LDC was found to be better than balanced Winnow, the perceptron and a recently proposed online linear discriminant analysis.

Cost‐effectiveness of screening for HLA‐A*31:01 prior to initiation of carbamazepine in epilepsy

Carbamazepine causes severe cutaneous adverse drug reactions that may be predicted by the presence of the HLA‐A*31:01 allele in northern European populations. There is uncertainty as to whether routine testing of patients with epilepsy is cost‐effective. We conducted an economic evaluation of HLA‐A*31:01 testing from the perspective of the National Health Service (NHS) in the United Kingdom.

Naive random subspace ensemble with linear classifiers for real-time classification of fMRI data

Functional magnetic resonance imaging (fMRI) provides a spatially accurate measure of brain activity. Real-time classification allows the use of fMRI in neurofeedback experiments. With limited labelled data available, a fixed pre-trained classifier may be inaccurate. We propose that streaming fMRI data may be classified using a classifier ensemble which is updated through naive labelling. Naive labelling is a protocol where in the absence of ground truth, updates are carried out using the label assigned by the classifier. We perform experiments on three fMRI datasets to demonstrate that naive labelling is able to improve upon a pre-trained initial classifier.

Conducting Economic Evaluations Alongside Randomised Trials: Current Methodological Issues and Novel Approaches

Trial-based economic evaluations are an important aspect of health technology assessment. The availability of patient-level data coupled with unbiased estimates of clinical outcomes means that randomised controlled trials are effective vehicles for the generation of economic data. However there are methodological challenges to trial-based evaluations, including the collection of reliable data on resource use and cost, choice of health outcome measure, calculating minimally important differences, dealing with missing data, extrapolating outcomes and costs over time and the analysis of multinational trials. This review focuses on the state of the art of selective elements regarding the design, conduct, analysis and reporting of trial-based economic evaluations. The limitations of existing approaches are detailed and novel methods introduced. The review is internationally relevant but with a focus towards practice in the UK.

On-Line fMRI Data Classification Using Linear and Ensemble Classifiers

The advent of real-time fMRI pattern classification opens many avenues for interactive self-regulation where the brain’s response is better modelled by multivariate, rather than univariate techniques. Here we test three on-line linear classifiers, applied to a real fMRI dataset, collected as part of an experiment on the cortical response to emotional stimuli. We propose a random subspace ensemble as a fast and more accurate alternative to component classifiers. The on-line linear discriminant classifier (O-LDC) was found to be a better base classifier than the on-line versions of the perceptron and the balanced winnow.

Choosing parameters for random subspace ensembles for fMRI classification

Functional magnetic resonance imaging (fMRI) is a non-invasive and powerful method for analysis of the operational mechanisms of the brain. fMRI classification poses a severe challenge because of the extremely large feature-to-instance ratio. Random Subspace ensembles (RS) have been found to work well for such data. To enable a theoretical analysis of RS ensembles, we assume that only a small (known) proportion of the features are important to the classification, and the remaining features are noise. Three properties of RS ensembles are defined: usability, coverage and feature-set diversity. Their expected values are derived for a range of RS ensemble sizes (L) and cardinalities of the sampled feature subsets (M). Our hypothesis that larger values of the three properties are beneficial for RS ensembles was supported by a simulation study and an experiment with a real fMRI data set. The analyses suggested that RS ensembles benefit from medium M and relatively small L.

Economic evaluation of a behavior-modifying intervention to enhance antiepileptic drug adherence

Between 35% and 50% of patients with epilepsy are reported to be not fully adherent to their medication schedule. We aimed to conduct an economic evaluation of strategies for improving adherence to antiepileptic drugs. Based on the findings of a systematic review, we identified an implementation intention intervention (specifying when, where, and how to act) which was tested in a trial that closely resembled current clinical management of patients with epilepsy and which measured adherence with an objective and least biased method. Using patient-level data, trial patients were matched with those recruited for the Standard and New Antiepileptic Drugs trial according to their clinical characteristics and adherence. Generalized linear models were used to adjust cost and utility in order to estimate the incremental cost per quality-adjusted life-year (QALY) gained from the perspective of the National Health Service in the UK. The mean cost of the intervention group, £1340 (95% CI: £1132, £1688), was marginally lower than that of the control group representing standard care, £1352 (95% CI: £1132, £1727). Quality-adjusted life-year values in the intervention group were higher than those in the control group, i.e., 0.75 (95% CI: 0.70, 0.79) compared with 0.74 (95% CI: 0.68, 0.79), resulting in a cost saving of £12 (€15, US

Cost effectiveness analysis of HLA-B*58:01 genotyping prior to initiation of allopurinol for gout

19) and with the intervention being dominant. The probability that the intervention is cost-effective at a threshold of £20,000 per QALY is 94%. Our analysis lends support to the cost-effectiveness of a self-directed, implementation intention intervention for improving adherence to antiepileptic drugs. However, as with any modeling dependent on limited data on efficacy, there is considerable uncertainty surrounding the clinical effectiveness of the intervention which would require a substantive trial for a more definitive conclusion.