Muscle artifact removal in an Epilepsy Monitoring Unit

1
Muscle artifact removal in an Epilepsy Monitoring
Unit
Highlighted application
2
Introduction

• Muscle artifact in EEG recordings is a common
  problem we found that muscle artifact
  interferred with the interpretation of ictal EEG
  recordings in around 90 of cases
• Ictal EEGs are often unreadable due to muscle
  artifact1
• Focal ictal beta discharges localize the ictal
  onset zone accurately and are highly predictive
  of excellent postsurgical outcome2.
  
  A low-pass filter with
  cut-off frequency of 15 Hz often removes this
  ictal beta activity and does not completely
  remove muscle artifact

1S.S. Spencer et al. Neurology 1985 35
1567-1575. 2G.A. Worrell et al. Epilepsia 2002
43 277-282
3
Aims

• To study the impact of muscle artifact on the
  readability of ictal EEG
• To study the impact of our new muscle artifact
  removal algorithm on the readability of ictal
  EEG.
• To study the improvement of the new method
  compared to the existing software available for
  muscle artifact removal

4
Methods

• We have developed a new algorithm to remove
  muscle artifact from ictal recordings. The method
  is semi-automatic and user-dependent.
• The technique is illustrated in the next slides
• The original EEG was an ictal recording of a
  patient with temporal lobe epilepsy. The
  original EEG was unreadable due to muscle
  artifact. You will have to click 15 times, and at
  each click, muscle artifact will be removed.
  After 15 clicks, we thought that all muscle
  artifact was removed. Left temporal lobe ictal
  activity is now obviously present.

A fully automated muscle artifact removal with
the method is possible but still under research.
5
The cursor is at the bottom of the stack. At each
click, it will move upward and a part of the
muscle artifact will be removed.
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
Muscle artifact-filtered EEG
20
Methods

• We selected one ictal EEG of 26 patients with
  refractory partial epilepsy, who underwent a
  presurgical evaluation at UZ Gasthuisberg,
  Leuven.
• All patients had concordant data (clinic, EEG,
  MRI, ictal SPECT, FDG-PET and neuropsychology).
• We selected the ictal EEG of the seizure during
  which an informative ictal SPECT injection was
  given, in order to have another functional gold
  standard in cases where ictal EEG was not
  informative (but not discordant).
• The muscle artifact-filtered EEG was compared to
  the original band pass filtered (0.3-35 Hz
  standard clinical settings) EEG.
• study the improvement of our method compared to
  the existing available software.
• We present our preliminary findings of an
  unblinded neurologist.1
• The same study with two blinded neurologist is
  planned in the near future.

1 These results were submitted for presentation
at the 26th International Epilepsy
Congress in Paris, August 2005.
21
Results
by the new method compared to currently
available software for muscle artifact removal
22
Example 1

• Patient was a 31 year old woman with epilepsy
  since age 16 years.
• Seizure frequency 20 per month.
• Aura scotomata and blindness.
• SISCOM cfr figure area of hyperperfusion right
  posterior
• At the site of hyperperfusion, we suspected a
  small focal cortical dysplasia on 1.5 T MRI. A 3T
  MRI is planned to confirm this.
• The ictal EEG as obtained with current available
  software and the muscle artifact filtered of this
  patient are presented in the next two slides.

23
EEG as obtained with current available software
24
Muscle artifact-filtered EEG
25
Example 2

• This 38 year old woman had refractory partial
  epilepsy since the age of 5 years after cerebral
  trauma affecting the left hemisphere
• Her right hand was functional and her language
  centers were on the left. Therefore, we did not
  consider a hemispherotomy
• In view of the sclerotic hippocampus on the left
  (arrow), we considered the possibility of a left
  temporal lobe resection if we could establish
  that all her seizures started in the left
  temporal lobe.
• The ictal EEG was unreadable due to muscle
  artifact.
• The EEG after removal of muscle artifact clearly
  showed ictal onset in frontocentral regions with
  spread towards the temporal lobe.
• She was not offered surgery.

26
EEG as obtained with current available software
27
Muscle artifact-filtered EEG
28
EEG as obtained with current available software
(next 10 sec)
29
Muscle artifact-filtered EEG
(next 10 sec)
Later spread towards left temporal lobe
30
Example 3

• This 36 year old woman suffered from refractory
  mesial temporal lobe epilepsy associated with
  left hippocampal sclerosis (white arrow).
• Ictal EEG was contaminated with muscle artifact
  and did not show obvious epileptic activity
• After muscle artifact removal, low voltage
  semirythmic activity over the left temporal lobe
  was evident.
• She underwent a left temporal lobe resection and
  has been seizure free for more than two years

31
EEG as obtained with current available software
32
Muscle artifact-filtered EEG
33
Conclusion

• Our new algorithm to remove muscle artifact
• Is fast
• Is user-friendly
• Can be implemented on any digital EEG workstation
• Makes interpretation of 90 of the ictal EEGs
  much easier
• Allows to detect seizure onset earlier, low
  voltage fast activity more frequent, and to
  pinpoint a more focal seizure onset in around 40
  of cases