Hierarchical state space partitioning with a network self-organising map for the recognition of ST-T segment changes

Abstract

The problem of maximising the performance of ST-T segment automatic recognition for ischaemia detection is a difficult pattern classification problem. The paper proposes the network self-organising map (NetSOM) model as an enhancement to the Kohonen self-organised map (SOM) model. This model is capable of effectively decomposing complex large-scale pattern classification problems into a number of partitions, each of which is more manageable with a local classification device. The NetSOM attempts to generalize the regularization and ordering potential of the basic SOM from the space of vectors to the space of approximating functions. It becomes a device for the ordering of local experts (i.e. independent neural networks) over its lattice of neurons and for their selection and co-ordination. Each local expert is an independent neural network that is trained and activated under the control of the NetSOM. This method is evaluated with examples from the European ST-T database. The first results obtained after the application of NetSOM to ST-T segment change recognition show a significant improvement in the performance compared with that obtained with monolithic approaches, i.e. with single network types. The basic SOM model has attained an average ischaemic beat sensitivity of 73.6% and an average ischaemic beat predictivity of 68.3%. The work reports and discusses the improvements that have been obtained from the implementation of a NetSOM classification system with both multilayer perceptrons and radial basis function (RBF) networks as local experts for the ST-T segment change problem. Specifically, the NetSOM with multilayer perceptrons (radial basis functions) as local experts has improved the results over the basic SOM to an average ischaemic beat sensitivity of 75.9% (77.7%) and an average ischaemic beat predictivity of 72.5% (74.1%).

Ischemia detection using supervised learning for hierarchical neural networks based on kohonen-maps

“…The training set is extracted from 110 fifteen minutes ECG records, consisting of representative normal and abnormal ST-T waveforms. After R-peak detection (inside the QRS complex) using the amplitude and the first derivative of the signal, [9] and baseline wander rejection (based on cubic splines) we could precisely extract the ST-T patterns for PCA feature extraction [7], [8].…”

Computational intelligence methods on biomedical signal analysis and data mining in medical records

“…MI is the most common cause of death in the industrialized countries and, as a consequence, its early diagnosis and treatment is of great importance. That's why one of the main applications for biomedical signal processing that we tackled is myocardial ischemia detection, based on recognition of ST-T segment changes, [8], or the problem of detection of the ischemia episodes, by an intelligent design of a window that we fed at the inputs of sNet-SOM, see [58].…”

“…• A new approach for data mining of symbolic data was introduced, see [ • In [8] we describe an analysis for the optimum selection (based on energy analysis) of the number of Principal Components (PC) for the representation of ECG signals. We used multiresolution analysis (wavelets) for the denoising of the Principal Components time series.…”

“…We have tried both thresholding for the domain of principal components coefficients [8] and the soft-thresholding method outperformed hard-thresholding method, at a reasonable computational cost.…”

“…Figure 6.6 illustrates the shape of the five basis functions. Figure 3 in an early article of our group, [8] demonstrates that the ST-T Segment can be reconstructed effectively with the first five PCA projections that represent about 98.1 % of the total signal energy. The PCA projection coefficients are then fed to the SOM nonlinear device in order to perform the complex (and highly nonlinear) classification decision about the category pertaining to each analysis case (i.e.…”