EEG SIGNAL PROCESSING FOR EPILEPSY DETECTION
Epilepsy, a condition well-known from early periods and thought to be “given by Gods” and is now considered a gateway into the structure and function of the brain and is thus a progressively lively interdisciplinary study area. Epilepsy occurs most often in children and the elderly. It is attributed to genetic defects, developmental abnormalities, febrile seizures, brain craniofacial damage, diseases of the central nervous system, hypoxia, ischemia, and tumors. Epilepsy is a symptom of frequent attacks. The unexpected initiation of synchronous neuronal firing is the cause of attacks and is reported by the scalp by electrodes. epileptic brain indicates by recording Electroencephalographic (EEG), that in slices of the cerebral hemispheres these discharges will start locally. Seizures come and go, almost unpredictably. In certain patients seizures can happen hundreds of times a day; in other cases, seizures only occur once every few years.
This is estimated that about 1 percent of the world’s population is suffering from epilepsy. Epileptic is diagnosed using an EEG signal since it’s a brain-associated disorder. For this purpose, an outpatient recording of EEG is used, which comprises EEG data for up to one week for a long time. It requires the effort of an expert to examine all the data to find signs of epilepsy, as Zumsteg et al have reported (2000).
In recent years several digital epileptic EEG devices have arisen as the conventional diagnostic approaches are repetitive and time-consuming. The presence of Epilepticform action in the EEG supports the analysis of epilepsy, which has often been disordered with other conditions that produce an analogous seizure. Leon. D. Iasemidis et al ( 2003) addressed several problems related to the detection of epileptic seizures based on the complex nonlinear aspect of the EEG signals. Adlassnig (1986 ) reported occasional transient-spikes and rapid waves of epileptic activity between seizures are used for characterizing the ECG of a patient. Arthur C. Guyton (1996 ) states that various forms of epileptic seizures are distinguished by distinct forms in the EEG waveform.
The motivation of this research is induced by the Ph.D. thesis of Sukanesh (1997), which discussed the distinct Electrocardiogram (ECG) signal data patterns creation and analysis for the cardiac disorders using supervised, unsupervised, counter propagation and self-organizing Neural Networks and is a very successful one (as explained in Appendix 1). Therefore, we utilized the same data pattern analysis in this research using EEG signals instead of ECG signals.
The High amplifiers are coupled with scalp electrodes, the human electroencephalogram (EEG) is typically reported from rods connected with the scalp. The amplified signals are recorded on paper via a polygraph usually containing 8 to 16 channels. Normal subjects usually exhibit α, β, γ,∆ behaviors, while irregular behaviors can result in a slowing and decreasing EEG amplitude, an increase in EEG frequency, and unexpected EEG discharges (paroxysmal behavior) that vary from the background in both frequencies or magnitude content or form.EEG is an effective screening device for neurological disorders. After this innovation, to diagnose epilepsy, to determine trauma, to examine sleep, and to evaluate higher brain functions the EEG is used. The EEG is extremely dependent on the accessibility of first-class equipment, and automated signal qualification methods have been applied almost from the beginning. The main objective of this technique is to assist the encephalography in the inefficient task of quantifying gesture that seems to the eye as a small background data material mixed with any bursts of periodic action at various frequencies. Despite years of work to develop standardized, automatic systems of detection, progress has only been attained in limited regions. Sleep planning with a high degree of precision also requires achievements; tracking spikes and wave clusters, and checking in serious care units. But for clinical applications clinicians often rely on graphical examination.
In multichannel EEG recordings to identify theoretically specified patterns, the human eye and brain may be skilled. But ostensive meanings aren’t spreading readily. A representation by words of a mental picture is typically bad and lengthy. What medical training requires a way to explore the better shape recognition of a human visual system and improve the visual data processing performance. It is required in medical training is a method to explore the better pattern recognition of a human visual system and increasing visual data communication efficiency.
Like our Indian adage, “Let Noble Thoughts Flow from All Directions,” Each of these considerations lead us to approach the use of computers from a slightly different angle in EEG research. Alison: Alison. A. Dingle et al (1993 ) described work in computer-based environments design that will be used to assist the physician in the visual medical evaluation of EEG multi-channel recordings. It is generally recognized that the information available to the doctor is somewhat unclear about his patient and about medical relationships in general. Nevertheless, the doctor is still quite able to draw conclusions from this knowledge, while approximate. The novel attempt in this research is to include a structured model of this mechanism in the form of a computerized diagnostic framework, using a mathematical approach in implementing the model. The two opposing and complementary approaches provide an indication of the emergence of epilepsy risk through non-linear models, AI, NN, and GA.
1.2 EEG SIGNALS FOR EPILEPSY RISK LEVEL DETECTION
“Electroencephalography” is a well-known medical technique that can deliver useful data on a variety of brain conditions (e.g., autism or brain cancers) being treated. Nonetheless, given its widespread use, fully automated diagnosis is one of the past predictable medical measures. Electroencephalogram (EEG) analysis requires the identification of patterns and characteristics that are typical of pathological conditions. Irregularities in the magnitude or frequency of normal action, for example, indicate a scratch while epileptiform action supports a medical analysis of epilepsy.
The EEG recording for the duration of a seizure is predominantly useful for deciding if a patient has epilepsy. Since seizures typically occur occasionally and randomly, finding those recordings may involve a multi-day EEG extension. Methods for the automatic discovery of small mal attacks and grand mal attacks have been developed, which have proved fairly effective. A patient with epilepsy may have EEG marked by occasional epileptiform transients (spikes and sharp waves) between seizures and thus relatively short recording can still be useful in epilepsy diagnostics. A routine recording normally takes 20-30 minutes, during which some 4 meters of paper recording is done. Epileptiform transients are detected by an electroencephalographic (EEG) by visual inspection of the image, which requires significant ability and it will take more time. Computerization of this method could also save time to improve objectivity and accuracy, and allow for quantification of investigation studies.
Some of these devices are in the experimental stage, and those in clinical use are limited to long-term EEG checking, with an EEGer checking entire detections. During regular EEG environment, such devices cannot work satisfactorily due to a big amount of wrong recognitions. It is widely agreed that using a spatial and temporal sense is the only way to distinguish the Epileptiform from non-Epileptiform waves. This method is being applied by different organizations in an attempt to eliminate false detections Glover et al. ( 1989) developed a method focused on a specific spatial framework, analyzing 12 EEG channels and additional contextual information given by the EKG, Electro oculogram& Electromyogram ( EMG) signals. Equally, the method introduced by Gotman and Wang (1991) applies as an extra-temporal framework, where EEG parts are categorized into one of the five situations (active wakefulness, quiet sleeplessness, synchronized EEG, phasic EEG, or slow-wave EEG) until separate guidelines are implemented to exclude non-Epileptiform behavior.
A new framework was developed, which creates substantial usage of contextual spatial and time-based data. This method has been shown to be especially effective in rejecting non-Epileptiform activity during awake EEGs and rest. This uses a mimetic approach to identify transient applicants, who are eventually trimmed or rejected by an expert program as Epileptiform. The device incorporates contextual information, both spatial and temporal, to detect definite and probable Epileptiform activities and reject non-Epileptiform waves. Preliminary findings suggest this device should be able to conduct a routine EEG clinical environment.
Fig: Overall process of predicting the onset of a seizure
Jing Wang1,2, GuanghuaXu “Some Highlights on Epileptic EEG Processing”, Recent Patents on Biomedical Engineering, Vol. 2, pp.48-57, 2009
By Li Qin, Yuedong Wang “Nonparametric spectral analysis with applications to seizure characterization using EEG time series”, Institute of Mathematical Statistics- The Annals of Applied Statistics, Vol. 2, No. 4, 1432–1451, 2008.