• Bayesian assessment of newborn brain maturity from two-channel sleep electroencephalograms

      Jakaite, Livija; Schetinin, Vitaly; Maple, Carsten; University of Bedfordshire (2012)
      Newborn brain maturity can be assessed by expert analysis of maturity-related patterns recognizable in polysomnograms. Since 36 weeks most of these patterns become recognizable in EEG exclusively, particularly, in EEG recorded via the two central-temporal channels. The use of such EEG recordings enables experts to minimize the disturbance of sleep, preparation time as well as the movement artifacts. We assume that the brain maturity of newborns aged 36 weeks and older can be automatically assessed from the 2-channel sleep EEG as accurately as by expert analysis of the full polysomnographic information. We use Bayesian inference to test this assumption and assist experts to obtain the full probabilistic information on the EEG assessments. The Bayesian methodology is feasibly implemented with Monte Carlo integration over areas of high posterior probability density, however the existing techniques tend to provide biased assessments in the absence of prior information required to explore a model space in detail within a reasonable time. In this paper we aim to use the posterior information about EEG features to reduce possible bias in the assessments. The performance of the proposed method is tested on a set of EEG recordings.
    • Bayesian decision trees for predicting survival of patients: a study on the US National Trauma Data Bank

      Schetinin, Vitaly; Jakaite, Livija; Jakaitis, Janis; Krzanowski, Wojtek; University of Bedfordshire; University of Exeter (2013)
      Trauma and Injury Severity Score (TRISS) models have been developed for predicting the survival probability of injured patients the majority of which obtain up to three injuries in six body regions. Practitioners have noted that the accuracy of TRISS predictions is unacceptable for patients with a larger number of injuries. Moreover, the TRISS method is incapable of providing accurate estimates of predictive density of survival, that are required for calculating confidence intervals. In this paper we propose Bayesian in ference for estimating the desired predictive density. The inference is based on decision tree models which split data along explanatory variables, that makes these models interpretable. The proposed method has outperformed the TRISS method in terms of accuracy of prediction on the cases recorded in the US National Trauma Data Bank. The developed method has been made available for evaluation purposes as a stand-alone application.
    • Classification of newborn EEG maturity with Bayesian averaging over decision trees

      Schetinin, Vitaly; Jakaite, Livija (Elsevier, 2012-08)
      EEG experts can assess a newborn’s brain maturity by visual analysis of age-related patterns in sleep EEG. It is highly desirable to make the results of assessment most accurate and reliable. However, the expert analysis is limited in capability to provide the estimate of uncertainty in assessments. Bayesian inference has been shown providing the most accurate estimates of uncertainty by using Markov Chain Monte Carlo (MCMC) integration over the posterior distribution. The use of MCMC enables to approximate the desired distribution by sampling the areas of interests in which the density of distribution is high. In practice, the posterior distribution can be multimodal, and so that the existing MCMC techniques cannot provide the proportional sampling from the areas of interest. The lack of prior information makes MCMC integration more difficult when a model parameter space is large and cannot be explored in detail within a reasonable time. In particular, the lack of information about EEG feature importance can affect the results of Bayesian assessment of EEG maturity. In this paper we explore how the posterior information about EEG feature importance can be used to reduce a negative influence of disproportional sampling on the results of Bayesian assessment. We found that the MCMC integration tends to oversample the areas in which a model parameter space includes one or more features, the importance of which counted in terms of their posterior use is low. Using this finding, we proposed to cure the results of MCMC integration and then described the results of testing the proposed method on a set of sleep EEG recordings.
    • Comparing robustness of pairwise and multiclass neural-network systems for face recognition

      Uglov, J.; Jakaite, Livija; Schetinin, Vitaly; Maple, Carsten (2008)
      Noise, corruptions, and variations in face images can seriously hurt the performance of face-recognition systems. To make these systems robust to noise and corruptions in image data, multiclass neural networks capable of learning from noisy data have been suggested. However on large face datasets such systems cannot provide the robustness at a high level. In this paper, we explore a pairwise neural-network system as an alternative approach to improve the robustness of face recognition. In our experiments, the pairwise recognition system is shown to outperform the multiclass-recognition system in terms of the predictive accuracy on the test face images.
    • Feature extraction from electroencephalograms for Bayesian assessment of newborn brain maturity

      Jakaite, Livija; Schetinin, Vitaly; Schult, Joachim (IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC, 2011)
      We explored the feature extraction techniques for Bayesian assessment of EEG maturity of newborns in the context that the continuity of EEG is the most important feature for assessment of the brain development. The continuity is associated with EEG “stationarity” which we propose to evaluate with adaptive segmentation of EEG into pseudo-stationary intervals. The histograms of these intervals are then used as new features for the assessment of EEG maturity. In our experiments, we used Bayesian model averaging over decision trees to differentiate two age groups, each included 110 EEG recordings. The use of the proposed EEG features has shown, on average, a 6% increase in the accuracy of age differentiation.
    • Informativeness of sleep cycle features in Bayesian assessment of newborn electroencephalographic maturation

      Schetinin, Vitaly; Jakaite, Livija; Schult, Joachim (IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC, 2011)
      Clinical experts assess the newborn brain development by analyzing and interpreting maturity-related features in sleep EEGs. Typically, these features widely vary during the sleep hours, and their informativeness can be different in different sleep stages. Normally, the level of muscle and electrode artifacts during the active sleep stage is higher than that during the quiet sleep that could reduce the informative-ness of features extracted from the active stage. In this paper, we use the methodology of Bayesian averaging over Decision Trees (DTs) to assess the newborn brain maturity and explore the informativeness of EEG features extracted from different sleep stages. This methodology has been shown providing the most accurate inference and estimates of uncertainty, while the use of DT models enables to find the EEG features most important for the brain maturity assessment.
    • Prediction of survival probabilities with Bayesian Decision Trees

      Schetinin, Vitaly; Jakaite, Livija; Krzanowski, Wojtek; University of Bedfordshire; University of Exeter (Elsevier, 2013)
      Practitioners use Trauma and Injury Severity Score (TRISS) models for predicting the survival probability of an injured patient. The accuracy of TRISS predictions is acceptable for patients with up to three typical injuries, but unacceptable for patients with a larger number of injuries or with atypical injuries. Based on a regression model, the TRISS methodology does not provide the predictive density required for accurate assessment of risk. Moreover, the regression model is difficult to interpret. We therefore consider Bayesian inference for estimating the predictive distribution of survival. The inference is based on decision tree models which recursively split data along explanatory variables, and so practitioners can understand these models. We propose the Bayesian method for estimating the predictive density and show that it outperforms the TRISS method in terms of both goodness-of-fit and classification accuracy. The developed method has been made available for evaluation purposes as a stand-alone application.