You are in: eMedicine Specialties > Neurology > Computer Applications in Neurology Virtual Reality to Evaluate Motor Response During Seizure ActivityArticle Last Updated: Oct 13, 2008AUTHOR AND EDITOR INFORMATIONSection 1 of 11
  Author: Morris Steffin, MD, Chief Science Officer, Virtual Reality Neurotech Lab Editors: Anthony M Murro, MD, Laboratory Director, Professor, Department of Neurology, Medical College of Georgia; Francisco Talavera, PharmD, PhD, Senior Pharmacy Editor, eMedicine; Selim R Benbadis, MD, Professor, Director of Comprehensive Epilepsy Program, Departments of Neurology and Neurosurgery, University of South Florida School of Medicine, Tampa General Hospital; Matthew J Baker, MD, Consulting Staff, Collier Neurologic Specialists, Naples Community Hospital; Nicholas Y Lorenzo, MD, Chief Editor, eMedicine Neurology; Consulting Staff, Neurology Specialists and Consultants Author and Editor Disclosure Synonyms and related keywords: virtual reality in medicine, VR, virtual reality interaction, videospace analysis techniques, complex-seizure motor activity, complex partial seizures, movement patterns INTRODUCTIONSection 2 of 11
  Virtual reality (VR) in medicine may be defined broadly as a bidirectional interface between the patient and the computer. By extension of this concept, the computer can be used to recognize patterns of patient behavior in space and react to such behavior. Standard paradigms of VR interaction present a computer-generated environment for the patient with measurement of patient response, usually followed by stereotyped changes in the computer-generated presentation. The patient's behavior is thus molded to the computer's conception of reality. But in many medical applications, the reverse philosophy must be adopted. The computer must sense the environment generated by the patient and then react intelligently to the patient. The challenge in the design of such systems then becomes one of appropriate pattern recognition and generation of response. These principles come to the forefront in analysis of epileptic phenomena, particularly in the case of complex partial seizures. Adaptation of principles developed in VR approaches can enhance detection and diagnosis of partial seizures. BACKGROUNDSection 3 of 11
The partial epileptic seizure may be quite subtle in its external presentation, while the internal effect on the patient is usually profound. Patients frequently experience disruption of consciousness, total amnesia for events, and complete disorganization of purposeful behavior during complex partial seizures. Such seizures are often of mesial temporal or frontal lobe origin and EEG, especially with scalp electrode recordings, may be equivocal or even normal. Movement patterns in complex partial seizures may include head adversion, facial myoclonus, automatism, or simple semipurposeful movements. Even with present technology, the observing neurologist categorizes such movements by visual evaluation. Determining whether such manifestations look like seizures, especially when no EEG correlation is evident during monitoring, entails a large degree of subjectivity. In such cases, establishing stereotypy (ie, autocorrelation of the movement pattern) from one seizure to the next is crucial. Stereotypic behavior consists of more than a general type of movement. It embodies a quantitatively demonstrable repetition of the movement sequence. Movement patterns similar to those in complex partial seizures may occur during nonphysiologic seizures, which may be termed pseudoseizures or psychogenic or functional seizures. In some cases, such movements may appear automatic, or sometimes convulsive, even to the experienced observer. The movements occur at irregular intervals, and practical techniques have not been readily available to determine whether episodes are quantitatively superimposable. Analysis by visual inspection is very tedious and prone to inaccuracy in more ambiguous cases, with the concomitant pitfall of failing to diagnose or localize an atypical physiological seizure. These patients often represent a diagnostic enigma (and a dilemma). Treatment of nonphysiological seizures with antiepileptic drugs produces severe adverse physical, cognitive, and psychological effects and is also expensive. Moreover, the approach carries substantial social stigma for patients and results in severe limitations on day-to-day activities (eg, driving). On the other hand, misdiagnosis of complex partial seizures as nonphysiological spells results in erroneous attribution of behavioral aberration to a patient who cannot control the events leading to his intellectual and behavioral dysfunction, and also precludes the possibility of effective treatment. DIAGNOSISSection 4 of 11
In many cases, diagnosis of complex partial seizures on an outpatient basis "in the field" is impossible. Ambulatory EEG may be helpful in selected cases, but is of little help when scalp EEG findings are normal or equivocal (which occurs quite frequently). In such cases, the patient must be admitted for video-EEG telemetry. This procedure is time-consuming and costly. Observation of the patient's behavior is primarily subjective and requires review of extensive data by a trained neurologist to determine whether the behavior in question does, or does not, look like a seizure. METHODSSection 5 of 11
Consider the computer response to an epileptic patient. Here "reality" is defined by measurable patient functions. The EEG can be used to track both generalized and many partial seizures. In these cases, EEG might serve as an early warning to the computer that a seizure is imminent (ie, the lightning before the thunder), though with major caveats. During the convulsive episode, patient movement in a virtual videospace may be tracked and related to the EEG by use of the virtual reality (VR) video intensity analytic technique used in the visual-haptic realm (see Visual-Haptic Interfaces: Modification of Motor and Cognitive Performance). This approach allows detailed quantitative analysis of motor activity during both convulsive and nonconvulsive seizures. The raw-gated analog video intensity signal represents composite depiction of the head adversive movements and the following jaw movement spikes as shown in Images 4-5, respectively. The suitably gated intensity signal provides a precise indication of movement patterns in definable patient videospace regions. A wave in the video signal can be identified that represents the head adversive movement into a portion of videospace that has been mapped as a region of interest (ROI), followed by an intensity change in the ROI that corresponds to chin myoclonic activity represented in Image 5. The reference ROI is shown as a blue overlay in Images 4 and 5. With this physical basis, the technique of spatial analysis called the video kinetogram (VKG) is demonstrated in Image 1. The patient in Image 1 and in the animation is experiencing a convulsive seizure. In the video-EEG monitoring setup, the patient has 3 rectangular ROIs, as shown overlaid in blue, arranged above the head and the right and left upper extremities (RUE and LUE, respectively). The placements of these ROIs are determined by the body regions involved most prominently in the seizure. As the head and limb movements occur, the video intensities averaged over each ROI for each frame of video are plotted (as labeled in Image 1) in the time domain at the lower portion of the screen. As the seizure builds, the convulsive activity is tracked accurately with regard to frequency and amplitude. These quantitative data also are presented in the frequency domain at the left of the video display. The traces to the left of the video are fast Fourier transforms (FFTs), calculated on the fly and completed at the end of each epoch, color coded (from the top) to the red, green, and blue time domain–graphed ROIs. The lower (black) FFT trace is coherent with the upper 3. All epochs are set in these demonstration cases at 18 seconds. In the displayed epoch (Image 1), the buildup process of rhythmic clonic activity in the 3 ROIs, along with frequency domain analysis at the left of Image 1 for each of the ROIs and their coherence, can be seen. A major challenge in the evaluation of the epileptic video "reality space" is the quantification of activity in partial seizures. Unlike the primarily generalized convulsive event, partial seizures may feature only subtle movement characteristics that often must provide the analytic direction because EEG onset may well be initially indiscernible. An indication of a seizure might be quite subtle: a simple facial or hand movement that, even to the trained eye, appears quite adventitious. RESULTSSection 6 of 11
The videospace analysis techniques implicit in the virtual reality (VR) approaches presented here can be brought to bear even more effectively in these subtle partial seizures. The patient's movements in selected regions of interest within the patient's videospace can be quantified. Consider the patient shown in Images 2-3. An animation of this activity shows a quick, unified, head adversive movement and handclasp. With observation of only a single event, stating unequivocally whether activity suggesting an ictal event has occurred is difficult. This determination is especially difficult by visual inspection alone. From the movement graphs in Images 2-3, the red, green, and blue graphs depict intensity changes within the blue rectangular regions of interest in the patient's space (marked on the video): head, RUE, and LUE, respectively. These graphs are generated in real time (as seen in the animation). Very soon after the head adversive movement (indicated in the upper, red trace), the RUE and then the LUE movements begin. Just before the end of the epoch is a relaxation phase, with partial return of the head, RUE, and LUE toward their initial positions. The fully displayed epoch, including the relaxation phase, is 18 seconds. Activity is graphed for the full epoch in Image 2. Here, in the 3 regions of interest (ie, head, right hand, and left hand), the head adversive movement, the right hand movement, and the left hand movement are observed to be closely timed. That could be a chance occurrence, but the graphed results in Image 3 show a different episode in which the timing is very close, and with good coherence from one episode to the next; a high degree of autocorrelation exists between the 2 episodes. Even the relaxation phase toward the end of the episode is very consistent from one episode to the next. In many cases, the automatic behavior of the complex partial seizure can be quite elaborate. In Image 4, the seizure is seen to begin with an adversive head movement to the right. This is graphed in the tracing beneath the video. The time course of the head adversion is demonstrated clearly and shown as a complete graph of the evolution in Image 4. The next phase of this seizure is the development of facial circumoral myoclonic activity, which is detected within the ROI (blue region) as shown in Image 5, and the movement spiking is demonstrated in the graph below the video. The change in facial configuration from relaxed state to myoclonus is, in fact, quite subtle on visual inspection, but it can be quantified quite accurately in terms of frequency and amplitude variation and by the video kinetogram (VKG) technique. From the head adversive and circumoral myoclonic phases, the seizure evolves to a major motor automatism, characterized by body turning and development of a complex wingbeating movement involving primarily the RUE, but also extending to the LUE. With suitable ROI placement (Image 6), the characteristics of this movement can be tracked accurately. The seizure shown in Image 7 shows a different sequence. Here, the seizure begins with a disorganized thrashing movement that appears semipurposeful, followed by a relatively quiet period. Then, a chin myoclonic phase occurs as the only prominent movement. Finally, more disorganized activity occurs toward the end of the seizure. Image 7 (A) shows the onset of the thrashing, and Image 7 (B) demonstrates the relatively quiet phase, while Image 7 (C) shows the period of isolated oral movements. These tend to be repetitive from one seizure to the next. CONCLUSIONSection 7 of 11
Video kinetogram is a new technique that shows promise as an analytical tool in assessing complex-seizure motor activity. Major benefits of this approach include reliable quantification of motor activity across multiple seizures and the potential for enhanced EEG-motor correlation and analysis. EMG, accelerometer data, and patient audio vocalization data also may be included for correlation with movement patterns; development is proceeding in the author's laboratory to include these types of data. Enhancing the accuracy of seizure flagging in the video-EEG telemetry environment will facilitate more rapid diagnosis and seizure localization. As the technique is perfected, the major issue will be allocating a greater discriminatory role to the computer in processing movement data. Currently, ROIs are placed manually by the operator. In this respect, the system, contrary to strict VR principles, is still in an open-loop phase. Enhancements are being developed to allow online computer analysis of movement patterns so that the operator will see the quantitative relationship in a more intuitive form. This phase of development will also require a more traditional VR presentation to the clinician, with intuitive mapping functions and a more immersive presentation of the data. In concert with increasingly automated intelligence in ROI placement and tracking on the part of the computer, such a computer-clinician interaction will facilitate the analytic process, especially as correlation with EEG is included in the process. FFTs are helpful in assessing repeatability of seizure pattern, particularly the presence of fairly constant movement types across seizures, although optimal display parameters and methods for inclusion in the clinician-oriented VR display are still under investigation. Early results indicate that these techniques will be useful in improving flagging and diagnostic accuracy of seizures during video-EEG telemetry. ADDITIONAL INFORMATIONSection 8 of 11
See eMedicine's articles Virtual Reality: Overview of its Application to Neurology and Visual-Haptic Interfaces: Modification of Motor and Cognitive Performance. ACKNOWLEDGMENTSSection 9 of 11
Patient material was kindly provided for the studies described here by Dr. Robert Fisher, Epilepsy Section, Division of Neurology, Barrow Neurological Institute, Phoenix, Arizona. MULTIMEDIASection 10 of 11
REFERENCESSection 11 of 11
Virtual Reality to Evaluate Motor Response During Seizure Activity excerpt Article Last Updated: Oct 13, 2008 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||
