ساخت شبكه ي پيچيده از روي سري‌زماني براي تشخيص اختلالات عصبي ‌عضلاني ميوپاتي و نوروپاتي

چكيده انگليسي :

Creating a complex network of time series to diagnose myopathy and neuropathy neuromuscular disorders Samaneh Samiei s samiei@ec iut ac ir 2020 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisor Dr Nasser Ghadiri nghadiri@iut ac ir Advisor Dr Behnaz Ansari ansaribehnaz@yahoo com Abstract Electromyography EMG is a biomedical signal that shows information about the neuromuscular activity as well asmuscular morphology This signal is used to diagnose neuromuscular disorders Improved analysis of electromyographicdata will help the experts to correctly diagnose neuromuscular disorders seizures and related diseases and to acceleratethe healing process Recently several techniques are introduced fro mapping from the time series to the complex network The time series is analyzed with the characteristics of a complex network These methods are used in many scientific andresearch fields and could be exploited in understanding the dynamics or predicting how the system evolves The resultingnetworks create a completely different visual that can be used by the physician to complement what is being taken fromEMG signals As a result medical errors are reduced and the treatment process is carried out more accurately and quickly following the correct identification of the disease and examining its various dimensions The characteristics of a time seriesare mapped to summarized criteria This summarization might lead to missing important information and prevent the modelfrom preserving all the properties of a time series Therefore it is still challenging to find an approach that can maintain allthe features of the time series and have a good representation of it In this study a new approach to building a network from an electromyographic time series is proposed using the visibil ity graph algorithm The proposed method fills the shortcoming of previous approaches in terms of insufficient accuracy interms of maintaining all the features of the time series and low classification accuracy In the proposed approach first theelectromyographic signals are pre processed The pre processing step involves the windowing of epochs and finding thesignal s linear envelope Then using the visibility graph algorithm the resulting linear envelope is mapped to a complexnetwork The networks obtained from the proposed approach can distinguish between healthy and patient samples and byexamining their structure useful information can be obtained about the patient or the health of the inputs In the next step the features are calculated through selective statistical measurements and the feature matrix is given to the classifiers Theperformance evaluation experiments on the proposed approach with a deep neural network show an accuracy of 99 30 for training data and 99 18 for test data Therefore the proposed method while adequately representing the network andincluding time series properties for three groups of healthy myopathy and neuropathy adequately covers the weaknessesof previous techniques It performs better than previous work in terms of accuracy precision recall specificity and F score Key Words Classification Complex network Electromyography Myopathy Neuromuscular dis orders Neuropathy Time series Visibility graph

سميعي، سمانه

ساخت شبكه ي پيچيده از روي سري‌زماني براي تشخيص اختلالات عصبي ‌عضلاني ميوپاتي و نوروپاتي

الكتروميوگرافي , دسته‌بندي , اختلالات عصبي‌عضلاني , سري زماني , شبكه پيجيده , گراف پديداري , ميوپاتي , نوروپاتي , EMG