- Email: [email protected]
PS-33-3 Towards “objective” EMG-differential-diagnosis. Performance of discriminant — and clusterclassification in myositis
Postersession 33. Expertsystemsin EMG of motor unit potentials (MUPs) decrease with increasing distance from the motor unit. When quantifying MUPs, this must be taken into account in needle E M G and even more in SEMG. This volume conduction property can be used to estimate the motor unit location and a depth normalized motor unit size parameter from SEMG. Two rows of 18 electrodes are placed rectilinear to the muscle fibre direction. Using single fibre triggered averaging of voluntary MUPs, the potential distribution of single motor units are used to estimate the motor units location and size by fitting them with a power function model. For validation of the estimated location and size both scanning E M G and Macro E M G were used. It was shown that the power function fits the data well and that different components of the SEMG MUP decrease differently. In the SEMG recordings with a good signal-to-noise ratio the average difference between the estimated motor unit location from S E M G and scanning E M G was 3 mm and rarely exceeded 6 mm. Even without depth normalization, a good correlation (0.86) was found between Macro area and those parameters obtained from SEMG. After depth normalization, SEMG related even better to the Macro EMG, which indicates that SEMG parameters can be used to estimate the motor unit size.
PS-33. EXPERT SYSTEMS IN EMG
[ PS-33-1 ] The wavelet transform In MUAP analysis Constantinos S. Pattichis ], Marios S. Pattichis, Lefkos T. Middleton 3. 1Department of Computer Science, Universityof
Cyprus, Nicosia, Cyprus; 2Department of Electrical Engineering, University of Texas, TX, USA 3 The Cyprus Institute of Neurology and Genetics, Nicosia, Cyprus Features of MUAPs extracted in the time domain such as duration, amplitude, and phases proved to be very valuable in differentiating between muscle and nerve disorders. Although the duration parameter is considered to be the key parameter for quantitative MUAP analysis, the definition of widely accepted criteria that will allow the automated measurement of this parameter are still lacking. On the other hand, MUAP frequency features like the mean or median frequency provide additional information in the assessment of neuromuscular pathology. Although frequency analysis is useful in describing the frequencies present in the signal, information about their occurrence in time is missing. The objective of this communication is to examine how the wavelet transform (WT) can be used to extract MUAP features in the time-frequency domains. The WT for low frequency bands takes a small number of samples, whereas for high frequency bands it takes a large number of samples. Thus, the WT is appropriate for tracking signals with low frequency components changing slowly in time, and high frequency components changing rapidly in time such as in MUAP morphology. Furthermore, the WT was investigated in the context of providing a more effective diagnosis when combined with artificial neural network (ANN) models. ANN models are constructed to resemble some simple organisational principles of the human brain in the pattern recognition and classification of signal and image data. MUAP WT findings were supplied to ANN models trained to diagnose normals, MND and myopathic patients. A maximum diagnostic yield of 80% was obtained.
S167
t PS-33-2 ] Quantitative EMG in the diagnosis of neuromuscular disorders Sunil Wimalaratna z, M.A. Tooley, E. Churchill, A.W. Preece, M.H. Morgan. 1FRCP Derriford Hospital, U.K.; Plymouth and
UBH & Frenchay Trusts, Bristol, U.K. Introduction: Quantitative E M G is carried out by sampling multiple sites using needle electrodes and the technique is available for clinical purposes [1]. There has been limited success in applying similar analysis to the surface E M G [2]. Because of the advantages of non invasive techniques in clinical practice we set out to investigate the use of surface E M G in the diagnosis of neuromuscular diseases. This study examines whether simple parameters such as turns, zero-crossings and MPF derived from the interference pattern of the surface E M G can differentiate neuropathies from myopathies. Method: EMG activity was recorded from the tibialis anterior muscle simultaneously by a surface electrode and a concentric needle electrode inserted directly beneath it. Thirty consented individuals (neuropathies, n = 10 myopathies, n = 10 and controls, n = 10) were studied. The subjects sat in a special chair designed to provide force data from tibialis anterior by dorsiflexion of the foot using a strain gauge. Data was collected for sixty second period during full contraction. The signal was analysed in real time by a transputer controlled 386 PC. Variables used were: Turns frequency thresholds were 1% FSD, 0.5% FSD, 100/.tV and 50/zV. Zero crossings thresholds were: 0.5% FSD, 0.25% FSD, 50 g,V and 25/zV. MPF calculated as the frequency value in the FFT power spectrum 1.7 Hz-3 kHz where 0.5 of the power is below and 0.5 is above. Results: Non-parametric analysis (p > 0:05) of surface data showed that discrimination is possible between controlled subjects and pathological groups. It was not possible to differentiate neuropathies from myopathies; However separation between the three groups had been possible by combination of data from surface and needle electrodes. A multivariate technique has also been explored [3] and separation of neuropathies from myopathies had been possible by using surface E M G data alone. Conclusion: Simple parameters from surface E M G interference pattern is useful only in differentiating pathological muscle from the normal muscle. A combination of minimally invasive single needle examination used simultaneously with surface EMG, helps to differentiate the three groups. Multivariate analysis shows potential for future studies in discrimination of neuropathies from myopathies. Acknowledgements: St Jules Charitable trust for supporting the development of this research project and Basildon postgraduate research fund for assisting to produce a bedside E M G equipment [1] Stalberg E et al [1983] Electr Clin Neuro 30:397--405 [2] DeLuca CJ et al [1985] Muscles alive, 5th Ed. [3]. Duchene J et ai [1987] Pattern Recognition 20:437--442
[ PS-33-31 Towards "objective" EMG-dlfferentlaldiagnosis. Performance of discriminant - and clusterclassification in myoslUs Gustav Pfeiffer, Klaus Kunze. Department o/Neurology, University Hospital Hamburg-Eppendorf Germany Computers have facilitated quantitative EMG. Its interpretation, however, still remains subjective, Discriminant classification (DC) of motor unit potentials (MUPs) was as sensitive as other uni- and multivariate statistical approaches but had the advantage of immediate differentiation between normal, myopathic and neurogenic
S 168
Postersession 33. Expert systems in EMG
findings, without recourse to examiner judgement or consensus criteria. DC may show, to what degree the uncertainties of EMG diagnosis are intrinsic or due to imperfect interpretation. Myositis is a demanding field in this respect, because of its variable E M G aspects. DC misclassified 3 of 23 biceps brachii but none of 16 tibialis anterior muscles. This success of DC adds weight to EMG in neuromuscular differential diagnosis. Cluster classification was used to explore, whether MUP analysis might be discriminating beyond the major categories, neuropathic or myopathic. There were no separate clusters for muscles from. patients with myotonic dystrophy and the various myositic subgroups. This experience indicates, that valid discrimination between neuropathic and myogenic findings is in reach of MUP-analysis, but probably not much more. Discriminant classification improves interpretation of MUP data to a degree that is clinically relevant, as illustrated casuistically.
third part describes the typical neuromuscular abnormalities associated with different generalized disorders. Many systemic diseases are associated with various neuromuscular abnormalities that are quite typical of that disease. For instance patients with rheumatoid arthritis have carpal tunnel syndrome, axonal polyneuropathy, polymyositis and myasthenia gravis more often than others.
[ PS-33-61 Computer communication between EMG laboratories P.E.O. F~illmar, Erik St~lberg, Stefan St~dberg, Mats Astr6m, Per Ytterberg, Joze Trontelj, Mustafa Ertas. Departments of
Clinical Neurophysiology, University Hospital Uppsala, Sweden, Ljubljana, Slovenia and Izmir, Turkey
We have developed new Turn/Amplitude analysis that allows a global evaluation of interference pattern. Our goals are to make the exam easier to carry out, to obtain quantified parameters that are similar to those subjectively utilised by the examiner, to aquire epochs automatically, to eliminate silent periods and artefacts and to represent the temporal trend of the parameters. We examine biceps brachialis and tibialis anterior muscles of 20 normal subjectS, 10 males and 10 females, aged 18-65 years. Five seconds of E M G activity free of silent periods and artefacts are automatically recorded, divided in 200-ms long epochs. We examine the following parameters: Turn/Amplitude distribution, activity, interturn, mean and maximum amplitude, time and epochs distribution of the parameters. The results analysis, which is made on line, allows a global quantified evaluation of muscle contraction in normal and patological conditions in a little time.
All PCs in our department (about 40) are connected by means of a local area network, LAN, with dedicated PCs as servers. The network integrates the functions of the laboratory such as patient booking, investigation instruments (EMG, EP, EEG, force measurements), printers, administrative routines and research and development facilities. This net is connected to the hospital network. The department is providing neurophysiological service to 5 hospitals in the region. There technicians are performing nerve conduction studies, evoked potential recordings and EEG. The EEGs are presently sent on MO discs. Their EMG/EP equipment is identical to that in our department and equipped with telephone modem. At the end of the day, all nerve conduction studies and evoked potentials, are transmitted (50 sec per patient) to a server on our LAN for editing, analysis and reporting. The transmission consists of traces, tables and comments. This communication can also be used for direct consultation during the day. Recently we have established international contacts, via telephone modem (Munich, Germany) and via Internet (Ljubljana, Slovenia and Izmir, Turkey). These international connections have provided the participating groups possibilities to exchange information, second opinion dialogues on specific cases and to collaborate, e.g. in collecting various databases. We are aiming at further international collaboration, with anyone interested, by the development of a World Wide Web (WWW) in neurophysiology.
I PS-33-51 Strategies in electrodlagnostic medicine - an Interactive EMG planner
I PS,-33-71 Quantitative motor unit potential shape analysis - - application to real signals
B. Falck, E. St~lberg, L. Korpinen. Department of Clinical Neurophysiology, University Hospital Uppsala, Sweden; Department of Power Engineering~ Tampers University of Technology, Tampers, Finland
Ewa Zalewska 1, Katarzyna Rowinska-Marcinska 1.2, Irena Hausmanowa-Petrusewicz I. l Neuromuscular Unit, Medical
IPS-33-41
Automatic analysis of the EMG interference pattern
Domenico De Grandis, Lispi Ludovico ]. Divisione Neurologica Arcispedale S. Anna, Ferrara, Italy; 11stituto di Clinica delle Malattie Nervose e Mentali, Roma, Italy
The strategy of an electrodiagnostic consultation is the plan for the investigation which aims at a correct diagnosis with the available methods and skills. With an optimal strategy the goal is reached with a minimum number of muscles and nerves tested; This is to save the patient from unnecessary discomfort and to reduce the time the doctor and technician need for the investigation. We have designed a prototype multimedia program that will guide the examiner in different clinical problems: The program consists of three parts. The first part deals with individual diseases such as carpal tunnel syndrome and myotonic dystrophy. The essential findings, both expected abnormal findings and expected normal findings are given. The procedure is the recommended tests that should be performed in each disorder. The second part describes differential diagnostic alternatives of commonly encountered symptoms. For instance paresthesias in digits 4 and 5. This may be due to an ulnar neuropathy at the elbow or wrist and also a lesion in the plexus brachialis or a C8 radiculopathy. The
Research Centre, Polish Academy of Sciences, Warsaw, Poland; 2 Department of Neurology, Warsaw Medical School, Warsaw, Poland The motor unit potential's shape depends on the motor architecture, i.e. on the number of fibres, their diameters and density. The changes of the shape are manifested by an increase in the number of phases and/or turns i.e. by signals of a more complicated shape than the normal motor unit potential. Such potentials are called polyphasic, polyturn or complex ones. The potential's parameters, e.g. amplitude, duration and area are not sufficiently sensitive to the detection of the signal's shape details. Therefore, a method for the quantitative evaluation of the motor unit potential's irregularity by means of an appropriately defined coefficient has been developed. The assumption was that this coefficient should resemble the visual impression of examiner. Visual evaluation is based mainly on what may be called a "curve length" i.e. visual estimate of the degree of complication of the potential. Hence, the irregularity coefficient has been defined as the "length"
PS-33. EXPERT SYSTEMS IN EMG
[ PS-33-1 ] The wavelet transform In MUAP analysis Constantinos S. Pattichis ], Marios S. Pattichis, Lefkos T. Middleton 3. 1Department of Computer Science, Universityof
Cyprus, Nicosia, Cyprus; 2Department of Electrical Engineering, University of Texas, TX, USA 3 The Cyprus Institute of Neurology and Genetics, Nicosia, Cyprus Features of MUAPs extracted in the time domain such as duration, amplitude, and phases proved to be very valuable in differentiating between muscle and nerve disorders. Although the duration parameter is considered to be the key parameter for quantitative MUAP analysis, the definition of widely accepted criteria that will allow the automated measurement of this parameter are still lacking. On the other hand, MUAP frequency features like the mean or median frequency provide additional information in the assessment of neuromuscular pathology. Although frequency analysis is useful in describing the frequencies present in the signal, information about their occurrence in time is missing. The objective of this communication is to examine how the wavelet transform (WT) can be used to extract MUAP features in the time-frequency domains. The WT for low frequency bands takes a small number of samples, whereas for high frequency bands it takes a large number of samples. Thus, the WT is appropriate for tracking signals with low frequency components changing slowly in time, and high frequency components changing rapidly in time such as in MUAP morphology. Furthermore, the WT was investigated in the context of providing a more effective diagnosis when combined with artificial neural network (ANN) models. ANN models are constructed to resemble some simple organisational principles of the human brain in the pattern recognition and classification of signal and image data. MUAP WT findings were supplied to ANN models trained to diagnose normals, MND and myopathic patients. A maximum diagnostic yield of 80% was obtained.
S167
t PS-33-2 ] Quantitative EMG in the diagnosis of neuromuscular disorders Sunil Wimalaratna z, M.A. Tooley, E. Churchill, A.W. Preece, M.H. Morgan. 1FRCP Derriford Hospital, U.K.; Plymouth and
UBH & Frenchay Trusts, Bristol, U.K. Introduction: Quantitative E M G is carried out by sampling multiple sites using needle electrodes and the technique is available for clinical purposes [1]. There has been limited success in applying similar analysis to the surface E M G [2]. Because of the advantages of non invasive techniques in clinical practice we set out to investigate the use of surface E M G in the diagnosis of neuromuscular diseases. This study examines whether simple parameters such as turns, zero-crossings and MPF derived from the interference pattern of the surface E M G can differentiate neuropathies from myopathies. Method: EMG activity was recorded from the tibialis anterior muscle simultaneously by a surface electrode and a concentric needle electrode inserted directly beneath it. Thirty consented individuals (neuropathies, n = 10 myopathies, n = 10 and controls, n = 10) were studied. The subjects sat in a special chair designed to provide force data from tibialis anterior by dorsiflexion of the foot using a strain gauge. Data was collected for sixty second period during full contraction. The signal was analysed in real time by a transputer controlled 386 PC. Variables used were: Turns frequency thresholds were 1% FSD, 0.5% FSD, 100/.tV and 50/zV. Zero crossings thresholds were: 0.5% FSD, 0.25% FSD, 50 g,V and 25/zV. MPF calculated as the frequency value in the FFT power spectrum 1.7 Hz-3 kHz where 0.5 of the power is below and 0.5 is above. Results: Non-parametric analysis (p > 0:05) of surface data showed that discrimination is possible between controlled subjects and pathological groups. It was not possible to differentiate neuropathies from myopathies; However separation between the three groups had been possible by combination of data from surface and needle electrodes. A multivariate technique has also been explored [3] and separation of neuropathies from myopathies had been possible by using surface E M G data alone. Conclusion: Simple parameters from surface E M G interference pattern is useful only in differentiating pathological muscle from the normal muscle. A combination of minimally invasive single needle examination used simultaneously with surface EMG, helps to differentiate the three groups. Multivariate analysis shows potential for future studies in discrimination of neuropathies from myopathies. Acknowledgements: St Jules Charitable trust for supporting the development of this research project and Basildon postgraduate research fund for assisting to produce a bedside E M G equipment [1] Stalberg E et al [1983] Electr Clin Neuro 30:397--405 [2] DeLuca CJ et al [1985] Muscles alive, 5th Ed. [3]. Duchene J et ai [1987] Pattern Recognition 20:437--442
[ PS-33-31 Towards "objective" EMG-dlfferentlaldiagnosis. Performance of discriminant - and clusterclassification in myoslUs Gustav Pfeiffer, Klaus Kunze. Department o/Neurology, University Hospital Hamburg-Eppendorf Germany Computers have facilitated quantitative EMG. Its interpretation, however, still remains subjective, Discriminant classification (DC) of motor unit potentials (MUPs) was as sensitive as other uni- and multivariate statistical approaches but had the advantage of immediate differentiation between normal, myopathic and neurogenic
S 168
Postersession 33. Expert systems in EMG
findings, without recourse to examiner judgement or consensus criteria. DC may show, to what degree the uncertainties of EMG diagnosis are intrinsic or due to imperfect interpretation. Myositis is a demanding field in this respect, because of its variable E M G aspects. DC misclassified 3 of 23 biceps brachii but none of 16 tibialis anterior muscles. This success of DC adds weight to EMG in neuromuscular differential diagnosis. Cluster classification was used to explore, whether MUP analysis might be discriminating beyond the major categories, neuropathic or myopathic. There were no separate clusters for muscles from. patients with myotonic dystrophy and the various myositic subgroups. This experience indicates, that valid discrimination between neuropathic and myogenic findings is in reach of MUP-analysis, but probably not much more. Discriminant classification improves interpretation of MUP data to a degree that is clinically relevant, as illustrated casuistically.
third part describes the typical neuromuscular abnormalities associated with different generalized disorders. Many systemic diseases are associated with various neuromuscular abnormalities that are quite typical of that disease. For instance patients with rheumatoid arthritis have carpal tunnel syndrome, axonal polyneuropathy, polymyositis and myasthenia gravis more often than others.
[ PS-33-61 Computer communication between EMG laboratories P.E.O. F~illmar, Erik St~lberg, Stefan St~dberg, Mats Astr6m, Per Ytterberg, Joze Trontelj, Mustafa Ertas. Departments of
Clinical Neurophysiology, University Hospital Uppsala, Sweden, Ljubljana, Slovenia and Izmir, Turkey
We have developed new Turn/Amplitude analysis that allows a global evaluation of interference pattern. Our goals are to make the exam easier to carry out, to obtain quantified parameters that are similar to those subjectively utilised by the examiner, to aquire epochs automatically, to eliminate silent periods and artefacts and to represent the temporal trend of the parameters. We examine biceps brachialis and tibialis anterior muscles of 20 normal subjectS, 10 males and 10 females, aged 18-65 years. Five seconds of E M G activity free of silent periods and artefacts are automatically recorded, divided in 200-ms long epochs. We examine the following parameters: Turn/Amplitude distribution, activity, interturn, mean and maximum amplitude, time and epochs distribution of the parameters. The results analysis, which is made on line, allows a global quantified evaluation of muscle contraction in normal and patological conditions in a little time.
All PCs in our department (about 40) are connected by means of a local area network, LAN, with dedicated PCs as servers. The network integrates the functions of the laboratory such as patient booking, investigation instruments (EMG, EP, EEG, force measurements), printers, administrative routines and research and development facilities. This net is connected to the hospital network. The department is providing neurophysiological service to 5 hospitals in the region. There technicians are performing nerve conduction studies, evoked potential recordings and EEG. The EEGs are presently sent on MO discs. Their EMG/EP equipment is identical to that in our department and equipped with telephone modem. At the end of the day, all nerve conduction studies and evoked potentials, are transmitted (50 sec per patient) to a server on our LAN for editing, analysis and reporting. The transmission consists of traces, tables and comments. This communication can also be used for direct consultation during the day. Recently we have established international contacts, via telephone modem (Munich, Germany) and via Internet (Ljubljana, Slovenia and Izmir, Turkey). These international connections have provided the participating groups possibilities to exchange information, second opinion dialogues on specific cases and to collaborate, e.g. in collecting various databases. We are aiming at further international collaboration, with anyone interested, by the development of a World Wide Web (WWW) in neurophysiology.
I PS-33-51 Strategies in electrodlagnostic medicine - an Interactive EMG planner
I PS,-33-71 Quantitative motor unit potential shape analysis - - application to real signals
B. Falck, E. St~lberg, L. Korpinen. Department of Clinical Neurophysiology, University Hospital Uppsala, Sweden; Department of Power Engineering~ Tampers University of Technology, Tampers, Finland
Ewa Zalewska 1, Katarzyna Rowinska-Marcinska 1.2, Irena Hausmanowa-Petrusewicz I. l Neuromuscular Unit, Medical
IPS-33-41
Automatic analysis of the EMG interference pattern
Domenico De Grandis, Lispi Ludovico ]. Divisione Neurologica Arcispedale S. Anna, Ferrara, Italy; 11stituto di Clinica delle Malattie Nervose e Mentali, Roma, Italy
The strategy of an electrodiagnostic consultation is the plan for the investigation which aims at a correct diagnosis with the available methods and skills. With an optimal strategy the goal is reached with a minimum number of muscles and nerves tested; This is to save the patient from unnecessary discomfort and to reduce the time the doctor and technician need for the investigation. We have designed a prototype multimedia program that will guide the examiner in different clinical problems: The program consists of three parts. The first part deals with individual diseases such as carpal tunnel syndrome and myotonic dystrophy. The essential findings, both expected abnormal findings and expected normal findings are given. The procedure is the recommended tests that should be performed in each disorder. The second part describes differential diagnostic alternatives of commonly encountered symptoms. For instance paresthesias in digits 4 and 5. This may be due to an ulnar neuropathy at the elbow or wrist and also a lesion in the plexus brachialis or a C8 radiculopathy. The
Research Centre, Polish Academy of Sciences, Warsaw, Poland; 2 Department of Neurology, Warsaw Medical School, Warsaw, Poland The motor unit potential's shape depends on the motor architecture, i.e. on the number of fibres, their diameters and density. The changes of the shape are manifested by an increase in the number of phases and/or turns i.e. by signals of a more complicated shape than the normal motor unit potential. Such potentials are called polyphasic, polyturn or complex ones. The potential's parameters, e.g. amplitude, duration and area are not sufficiently sensitive to the detection of the signal's shape details. Therefore, a method for the quantitative evaluation of the motor unit potential's irregularity by means of an appropriately defined coefficient has been developed. The assumption was that this coefficient should resemble the visual impression of examiner. Visual evaluation is based mainly on what may be called a "curve length" i.e. visual estimate of the degree of complication of the potential. Hence, the irregularity coefficient has been defined as the "length"