Genetics-based machine learning for the assessment of certain neuromuscular disorders
Ημερομηνία
1996ISSN
1045-9227Source
IEEE Transactions on Neural NetworksVolume
7Issue
2Pages
427-439Google Scholar check
Keyword(s):
Metadata
Εμφάνιση πλήρους εγγραφήςΕπιτομή
Clinical electromyography (EMG) provides useful information for the diagnosis of neuromuscular disorders. The utility of artificial neural networks (ANN's) in classifying EMG data trained with backpropagation or Kohonen's self-organizing feature maps algorithm has recently been demonstrated. The objective of this study is to investigate how genetics-based machine learning (GBML) can be applied for diagnosing certain neuromuscular disorders based on EMG data. The effect of GBML control parameters on diagnostic performance is also examined. A hybrid diagnostic system is introduced that combines both neural network and GBML models. Such a hybrid system provides the end-user with a robust and reliable system, as its diagnostic performance relies on more than one learning principle. In the clinical EMG laboratory, 680 motor unit action potentials (MUAP's) were collected from 12 normal, 11 motor neuron disease, and 11 myopathy subjects. Eight subjects from each group formed the training set, and the other 10 subjects formed the evaluation set. Each subject was described by a 14-element feature vector consisting of the mean and the standard deviation of each of the following MUAP parameters: duration, spike duration, amplitude, area, spike area, phases, and turns. More than a thousand GBML models were developed by varying the following parameters: message length size (49, 74), number of classifiers (100, 150, 200, 250, 300, 500), lifetax (0.000, 0.002, 0.005, 0.010), period of genetic algorithm (GA) introduced, which is expressed in iterations, showing how often the classifier system calls the GA (50, 100, 200, 500), crossover probability (0.5, 1.0), and mutation probability (0.00, 0.01, 0.02). A total of 28 models were selected that achieved a diagnostic yield better than 95% and 70% for the training and evaluation sets, respectively. This criterion, suggested by two expert neurophysiologists, has formed the bases for classifying a GBML model as "successful" and worthy of further consideration in a clinical environment. The performance of GBML models as affected by varying the above parameters can be summarized as follows: 1) 49-bit MUAP parameters decoding scheme were sufficient for accommodating the complexity of the feature vector 2) the number of classifiers for selected models trained with 74-bit data strings were 300 and 500, whereas the number of classifiers for most selected models with 49-bit data strings were 200 and 500 3) by increasing lifetax, training performance is reduced, whereas evaluation performance remains at the same levels 4) the genetic algorithm should not be called upon very frequently because it causes drastic changes to the status of the classifiers and 5) models with crossover probability equal to one yielded better overall performance. GBML models demonstrated similar performance to neural-network models, but with less computation. The diagnostic performance of neural network and GBML models is enhanced by the hybrid system. © 1996 IEEE.