Data-driven treatment selection for seamless phase II/III trials incorporating early-outcome data

Seamless phase II/III clinical trials are conducted in two stages with treatment selection at the first stage. In the first stage, patients are randomized to a control or one of k > 1 experimental treatments. At the end of this stage, interim data are analysed, and a decision is made concerning which experimental treatment should continue to the second stage. If the primary endpoint is observable only after some period of follow-up, at the interim analysis data may be available on some early outcome on a larger number of patients than those for whom the primary endpoint is available. These early endpoint data can thus be used for treatment selection. For two previously proposed approaches, the power has been shown to be greater for one or other method depending on the true treatment effects and correlations. We propose a new approach that builds on the previously proposed approaches and uses data available at the interim analysis to estimate these parameters and then, on the basis of these estimates, chooses the treatment selection method with the highest probability of correctly selecting the most effective treatment. This method is shown to perform well compared with the two previously described methods for a wide range of true parameter values. In most cases, the performance of the new method is either similar to or, in some cases, better than either of the two previously proposed methods.

The development of a data-driven application benchmarking approach to performance modelling

Performance modelling is a useful tool in the lifeycle of high performance scientific software, such as weather and climate models, especially as a means of ensuring efficient use of available computing resources. In particular, sufficiently accurate performance prediction could reduce the effort and experimental computer time required when porting and optimising a climate model to a new machine. In this paper, traditional techniques are used to predict the computation time of a simple shallow water model which is illustrative of the computation (and communication) involved in climate models. These models are compared with real execution data gathered on AMD Opteron-based systems, including several phases of the U.K. academic community HPC resource, HECToR. Some success is had in relating source code to achieved performance for the K10 series of Opterons, but the method is found to be inadequate for the next-generation Interlagos processor. The experience leads to the investigation of a data-driven application benchmarking approach to performance modelling. Results for an early version of the approach are presented using the shallow model as an example.

First-principle data-driven models for assessment of motor disorders in Parkinson’s disease

Parkinson’s disease (PD) is an increasing neurological disorder in an aging society. The motor and non-motor symptoms of PD advance with the disease progression and occur in varying frequency and duration. In order to affirm the full extent of a patient’s condition, repeated assessments are necessary to adjust medical prescription. In clinical studies, symptoms are assessed using the unified Parkinson’s disease rating scale (UPDRS). On one hand, the subjective rating using UPDRS relies on clinical expertise. On the other hand, it requires the physical presence of patients in clinics which implies high logistical costs. Another limitation of clinical assessment is that the observation in hospital may not accurately represent a patient’s situation at home. For such reasons, the practical frequency of tracking PD symptoms may under-represent the true time scale of PD fluctuations and may result in an overall inaccurate assessment. Current technologies for at-home PD treatment are based on data-driven approaches for which the interpretation and reproduction of results are problematic.  The overall objective of this thesis is to develop and evaluate unobtrusive computer methods for enabling remote monitoring of patients with PD. It investigates first-principle data-driven model based novel signal and image processing techniques for extraction of clinically useful information from audio recordings of speech (in texts read aloud) and video recordings of gait and finger-tapping motor examinations. The aim is to map between PD symptoms severities estimated using novel computer methods and the clinical ratings based on UPDRS part-III (motor examination). A web-based test battery system consisting of self-assessment of symptoms and motor function tests was previously constructed for a touch screen mobile device. A comprehensive speech framework has been developed for this device to analyze text-dependent running speech by: (1) extracting novel signal features that are able to represent PD deficits in each individual component of the speech system, (2) mapping between clinical ratings and feature estimates of speech symptom severity, and (3) classifying between UPDRS part-III severity levels using speech features and statistical machine learning tools. A novel speech processing method called cepstral separation difference showed stronger ability to classify between speech symptom severities as compared to existing features of PD speech. In the case of finger tapping, the recorded videos of rapid finger tapping examination were processed using a novel computer-vision (CV) algorithm that extracts symptom information from video-based tapping signals using motion analysis of the index-finger which incorporates a face detection module for signal calibration. This algorithm was able to discriminate between UPDRS part III severity levels of finger tapping with high classification rates. Further analysis was performed on novel CV based gait features constructed using a standard human model to discriminate between a healthy gait and a Parkinsonian gait. The findings of this study suggest that the symptom severity levels in PD can be discriminated with high accuracies by involving a combination of first-principle (features) and data-driven (classification) approaches. The processing of audio and video recordings on one hand allows remote monitoring of speech, gait and finger-tapping examinations by the clinical staff. On the other hand, the first-principles approach eases the understanding of symptom estimates for clinicians. We have demonstrated that the selected features of speech, gait and finger tapping were able to discriminate between symptom severity levels, as well as, between healthy controls and PD patients with high classification rates. The findings support suitability of these methods to be used as decision support tools in the context of PD assessment.

A data driven approach for automating vehicle activated signs

Vehicle activated signs (VAS) display a warning message when drivers exceed a particular threshold. VAS are often installed on local roads to display a warning message depending on the speed of the approaching vehicles. VAS are usually powered by electricity; however, battery and solar powered VAS are also commonplace. This thesis investigated devel-opment of an automatic trigger speed of vehicle activated signs in order to influence driver behaviour, the effect of which has been measured in terms of reduced mean speed and low standard deviation. A comprehen-sive understanding of the effectiveness of the trigger speed of the VAS on driver behaviour was established by systematically collecting data. Specif-ically, data on time of day, speed, length and direction of the vehicle have been collected for the purpose, using Doppler radar installed at the road. A data driven calibration method for the radar used in the experiment has also been developed and evaluated. Results indicate that trigger speed of the VAS had variable effect on driv-ers’ speed at different sites and at different times of the day. It is evident that the optimal trigger speed should be set near the 85th percentile speed, to be able to lower the standard deviation. In the case of battery and solar powered VAS, trigger speeds between the 50th and 85th per-centile offered the best compromise between safety and power consump-tion. Results also indicate that different classes of vehicles report differ-ences in mean speed and standard deviation; on a highway, the mean speed of cars differs slightly from the mean speed of trucks, whereas a significant difference was observed between the classes of vehicles on lo-cal roads. A differential trigger speed was therefore investigated for the sake of completion. A data driven approach using Random forest was found to be appropriate in predicting trigger speeds respective to types of vehicles and traffic conditions. The fact that the predicted trigger speed was found to be consistently around the 85th percentile speed justifies the choice of the automatic model.

A comparative study of supervised learning techniques for data-driven haptic simulation

This paper focuses on the choice of a supervised learning algorithm and possible data preprocessing in the domain of data-driven haptic simulation. This is done through a comparison of the performance of different supervised learning techniques with and without data preprocessing. The simulation of haptic interactions with deformable objects using data-driven methods has emerged as an alternative to parametric methods. The accuracy of the simulation depends on the empirical data and the learning method. Several methods were suggested in the literature and here we provide a comparison between their performance and applicability to this domain. We selected four examples to be compared: singular learning mechanism which is artificial neural networks (ANN), attribute selection followed by ANN learning process, ensemble of multiple learning techniques, and attribute selection followed by the learning ensemble. These methods performance was compared in the domain of simulating multiple interactions with a deformable object with nonlinear material behavior.

Flood forecasting in large rivers with data-driven models

Results from the application of adaptive neuro-fuzzy inference system (ANFIS) to forecast water levels at 3 stations along the mainstream of the Lower Mekong River are reported in this paper. The study investigated the effects of including water levels from upstream stations and tributaries, and rainfall as inputs to ANFIS models developed for the 3 stations. When upstream water levels in the mainstream were used as input, improvements to forecasts were realized only when the water levels from 1 or at most 2 upstream stations were included. This is because when there are significant contributions of flow from the tributaries, the correlation between the water levels in the upstream stations and stations of interest decreases, limiting the effectiveness of including water levels from upstream stations as inputs. In addition, only improvements at short lead times were achieved. Including the water level from the tributaries did not significantly improve forecast results. This is attributed mainly to the fact that the flow contributions represented by the tributaries may not be significant enough, given that there could be large volume of flow discharging directly from the catchments which are ungauged, into the mainstream. The largest improvement for 1-day forecasts was obtained for Kratie station where lateral flow contribution was 17 %, the highest for the 3 stations considered. The inclusion of rainfall as input resulted in significant improvements to long-term forecasts. For Thakhek, where rainfall is most significant, the persistence index and coefficient of efficiency for 5-lead-day forecasts improved from 0.17 to 0.44 and 0.89 to 0.93, respectively, whereas the root mean square error decreased from 0.83 to 0.69 m.

3D Intervertebral disc localization and segmentation from MR images by data-driven regression and classification

In this paper we propose a new fully-automatic method for localizing and segmenting 3D intervertebral discs from MR images, where the two problems are solved in a unified data-driven regression and classification framework. We estimate the output (image displacements for localization, or fg/bg labels for segmentation) of image points by exploiting both training data and geometric constraints simultaneously. The problem is formulated in a unified objective function which is then solved globally and efficiently. We validate our method on MR images of 25 patients. Taking manually labeled data as the ground truth, our method achieves a mean localization error of 1.3 mm, a mean Dice metric of 87%, and a mean surface distance of 1.3 mm. Our method can be applied to other localization and segmentation tasks.

On the determination of the number of factors using information criteria with data-driven penalty

As is well known, when using an information criterion to select the number of common factors in factor models the appropriate penalty is generally indetermine in the sense that it can be scaled by an arbitrary constant, c say, without affecting consistency. In an influential paper, Hallin and Liška (J Am Stat Assoc102:603–617, 2007) proposes a data-driven procedure for selecting the appropriate value of c. However, by removing one source of indeterminacy, the new procedure simultaneously creates several new ones, which make for rather complicated implementation, a problem that has been largely overlooked in the literature. By providing an extensive analysis using both simulated and real data, the current paper fills this gap.

Localization and degmentation of 3D intervertebral discs in MR Images by data driven estimation

This paper addresses the problem of fully-automatic localization and segmentation of 3D intervertebral discs (IVDs) from MR images. Our method contains two steps, where we first localize the center of each IVD, and then segment IVDs by classifying image pixels around each disc center as foreground (disc) or background. The disc localization is done by estimating the image displacements from a set of randomly sampled 3D image patches to the disc center. The image displacements are estimated by jointly optimizing the training and test displacement values in a data-driven way, where we take into consideration both the training data and the geometric constraint on the test image. After the disc centers are localized, we segment the discs by classifying image pixels around disc centers as background or foreground. The classification is done in a similar data-driven approach as we used for localization, but in this segmentation case we are aiming to estimate the foreground/background probability of each pixel instead of the image displacements. In addition, an extra neighborhood smooth constraint is introduced to enforce the local smoothness of the label field. Our method is validated on 3D T2-weighted turbo spin echo MR images of 35 patients from two different studies. Experiments show that compared to state of the art, our method achieves better or comparable results. Specifically, we achieve for localization a mean error of 1.6-2.0 mm, and for segmentation a mean Dice metric of 85%-88% and a mean surface distance of 1.3-1.4 mm.