A Machine Learning Perspective Christian Ward Dr Iyad Obeid and Dr Joseph Picone Neural Engineering Data Consortium College of Engineering Temple University Philadelphia Pennsylvania USA ID: 916955
Download Presentation The PPT/PDF document "Manual Interpretation of EEGs:" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Manual Interpretation of EEGs:A Machine Learning Perspective
Christian Ward, Dr. Iyad Obeid and Dr. Joseph PiconeNeural Engineering Data ConsortiumCollege of EngineeringTemple UniversityPhiladelphia, Pennsylvania, USA
Slide2AbstractThe goal of this presentation is to
describe, from a machine learning perspective, how an electroencephalogram (EEG) is manually interpreted. This presentation is not meant to be a comprehensive tutorial on diagnosing illnesses from EEG data. Instead, it attempts to document what physicians look for in the signals, and to describe these in terms that can be translated into signal processing and machine learning code.An extracranial EEG measures brain function indirectly through electrodes placed in an array on the scalp. Events of interest to physicians originate as nerve impulses deep in the brain. By the time these signals reach the scalp they are heavily filtered and spatially dispersed. Interpreting the underlying pathology or determining the diagnosis is an art because (1) the transduced signals at the scalp are very noisy due to their relatively low voltage (microvolts) and (2) the artifacts in the waveform that correlate with a pathology are context dependent (knowledge of the surrounding behavior is important).In this presentation, we review what diseases are diagnosed using EEGs, what artifacts are used to identify these diseases, and what properties of the signal correlate with these artifacts. Both time domain (e.g., waveform shapes) and frequency domain behavior (e.g., fundamental frequency) play an important role in identifying these artifacts.
Slide3A technician administers a 30−minute recording session.
An EEG specialist (neurologist) interprets the EEG.
An EEG report is generated with the diagnosis.
Patient is billed once the report is coded and signed off.
Manual Interpretation of EEGs
Slide4329 positions are defined as potential electrode locations:
24 to 36 channels are commonly used; 64 and 128-channel EEGs are used in research10/20 configuration (black): yields 19 channels plus several reference channels (typically the ears and forehead)10/10 configuration (gray): typically doubles the number of channelsDifferential voltages are typically measured between a sensor and an ear (e.g., F3-T9)
Common reference points include the ears (T9 and T10), the nose (e.g., ???) and the heart/respiratory system (e.g., EKG1 and EKG2)
Signal Transduction
Slide5Epilepsy
: a neurological disorder marked by sudden recurrent episodes of sensory disturbance, loss of consciousness, or convulsions.Stroke: a sudden disabling attack or loss of consciousness caused by an interruption in the flow of blood to the brain.Posterior Reversible Encephalopathy Syndrome (PRES): characterized
by headache
, confusion, seizures and visual loss. It may occur due to a number of causes, predominantly malignant hypertension, eclampsia and some medical treatments.
Middle Cerebral Artery (MCA) I
nfarct:
obstruction of one
of the three major paired arteries that supply blood to the cerebrum
.
Other uses
of EEGs include
diagnosis of Alzheimer's disease, certain psychoses, and sleep disorders (narcolepsy).
The EEG may also be used to determine the overall electrical activity of the
brain, which is used to
evaluate trauma, drug intoxication, or
brain damage.
The
EEG may also be used to monitor blood flow in the brain during surgical procedures
.
The EEG
cannot be used to measure
intelligence
Common Diseases Diagnosed With An EEG
Slide6Background:
models all other artifactsnot included in the above categories.Though there are some variations in these signals for infants whose brains are not fully developed, the basic composition of these signals is relatively age and subject invariant.
Spike and Wave:
… describe… … .. .. .. . . . . . . . …………………. . . . . . .
GPED
Triphasic
:
… describe …. . . . . . . . ... . . . … . . . . . . . . . . . . . . .
PLED:
… describe …. . . . . . . .. . . . . . . . . . . . . … . . . . . . . . . . . .
Eye Blink:
… describe why …… . . . . . . . . . . . . . . . . ….. . . . . . . . . .
Most Significant Primitives
Slide7Other Primitives
NameAcronymDefinitionElectrographic Seizure
N/A
Rhythmic
discharge or spike and wave pattern with definite evolution in frequency, location or morphology lasting at least 10 sec
Periodic
Epileptiform
discharges
PED
Repetitive
sharp waves, spikes, or sharply contoured waves at regular or nearly regular intervals and without clear evolution in frequency or location
Perioidic
Lateralized
Epileptiform
discharges
PLED
Consistently lateralized
PED
Generalized
PED
GPED
Bilateral
and synchronous PED with no consistent lateralization
Bilateral
PLED
BiPLED
PLED occurring bilaterally, but independently and
sychronously
Triphasic
Waves
N/A
Generalized periodic
sharp waves or sharply contoured delta waves with
triphasic
morphology, at 1-3 Hz, with/without anterior-posterior or posterior-anterior lag
Frontal Intermittent Rhythmic Delta Activity
FIRDA
Moderate- to high-voltage
monorhythmic
and sinusoidal 1-3 Hz activity seen bilaterally, maximal in anterior leads, no evolution
Slide8Classification of Diseases
NamePrimary Marker(s)Secondary Marker(s)Epilepsy
Stroke
PRES
MCA
Infarct
Slide9Summary and Future Work
Summary:Deep…For…NEDC…Future Work:Three…Two…
Slide10Brief
Bibliography of Relevant Documentation[1] Tatum, W., Husain, A., Benbadis, S., & Kaplan, P. (2007). Handbook of EEG Interpretation. (Kirsch, Ed.) (p. 276). New York City, New York, USA: Demos Medical Publishing.[2] Wulsin, D. F., Gupta, J. R., Mani, R., Blanco, J. A., & Litt, B. (2011). Modeling electroencephalography waveforms with semi-supervised deep belief nets: fast classification and anomaly measurement. Journal of Neural Engineering, 8(3), 36015.[3] Jurcak, V., Tsuzuki, D., & Dan, I. (2007). 10/20, 10/10, and 10/5 systems revisited: Their validity as relative head-surface-based positioning systems. NeuroImage, 34(4), 1600–1611.[4] Claassen
, J., Mayer, S. A., Kowalski, R. G., Emerson, R. G., & Hirsch, L. J. (2004). Detection of electrographic seizures with continuous EEG monitoring in critically ill patients.
Neurology, 62(10), 1743–1748.