Talk:Seizure prediction
Contents |
Reviewer B
This is an excellent short review which should be updated regularly to cover recent developments.
Reply to Reviewer B
Modifications by Reviewer B were accepted.
Reviewer C
This review provides a concise and up-to-date review of the field of seizure prediction. Text and figures are, in general, well suited for the target audience of Scholarpedia. I have only some suggestions and amendments the author might want to consider in order to further improve his contribution.
(1) This contribution gives the impression of being a little bit too close to a previous review published by the author in Brain in 2007. This could be avoided by re-wording some parts of the manuscript and by adding some more up-to-date references (e.g. in section Motivation: Stacey & Litt‘s recent review in Nature Clinical Practice Neurology 2008 within the context of on-demand therapy; Hoppe et al Arch Neurol 2007 within the context of self-reported data)
(2) Usage of term EEG: in order to avoid confusion, it should be mentioned in the text that the majority of studies used intracranially recorded EEG.
(3) The terms “retrospective prediction algorithm” and “prospective prediction algorithm” should be explained more thoroughly. Currently they appear kind of mixed up. It is not clear to what type of algorithm the author refers to.
(4) The caption to Fig 2 needs to be expanded. What is the difference between upper and lower part of the figure? What exactly is meant by “electrophysiological measure”?
(5) Terms like “channels with optimum performance” or “optimal channels” need to be explained more thoroughly.
(6) Terms from the medical community (such as “pre-ictal”) should be explained.
(7) I would suggest subdividing the section “Understanding mechanisms of seizure generation” into: (a) studies in humans, (b) studies in animal models of epilepsy, and (c) computational models and would suggest to add the following references for part (a) Schindler et al., Brain 2007; for part (b) McCormick & Contreras Rev Physiol 2001; Avoli et al., Prog Neurobiol 2002; Jefferys, Epilepsia 2003; Beck and Yaari Nature Rev Neurosci 2008; and for part (c) Lytton Nature Rev Neurosci 2008; Feldt et al, PRE 2007.
(8) current reference list Review by Hughes: despite being published in 2008 this review appears to be out-of-date. Moreover, it appears to be written from someone who does not appear to be involved in the field. The author might also want to cite the book chapter by Lehnertz K, Le Van Quyen M, Litt B. (2007) Seizure Prediction. In: J Engel & T Pedley (eds.) Epilepsy: A Comprehensive Textbook, 2nd Ed. Lippincott Williams & Wilkins, pp.1011-1024
Reply to Reviewer C
(1) Added the suggested references.
(2) Added the sentence ‘These and later studies predominantly analyzed EEG signals from patients undergoing video-EEG monitoring, with chronic electrodes implanted directly inside or on the surface of the brain to localize the seizure focus for possible surgical resection.’
(3) I am not sure what a “retrospective prediction algorithm” is, but this term is not used anywhere in the text. The term “prospective prediction algorithm” is defined in the text as an algorithm whose “output at a given time should be a function of the information available at this time.”
(4) The upper and lower part of Figure 2 show various types of true and false warnings. They are conceptually the same. I replaced ‘electrophysiological’ measure with ‘characterizing measure’.
(5) Replaced ‘channels with optimum performance’ with ‘channels with particularly good performance’.
(6) The terms ‘ictal’ and ‘pre-ictal’ are now defined when first used in the text.
(7) I have followed the reviewer’s suggestion to distinguish between different types of studies on seizure generation and added the suggested references.
(8) Replaced the review by Hughes with the book chapter by Lehnertz et al.
Reviewer D
Comments on Scholarpedia Chapter “Seizure Prediction” by Florian Mormann
My general impression is that the Chapter is well balanced and clearly written. Furthermore it is accessible to the non-specialist. In my opinion a few points could be made more precise. 1. On page 3 the term “pre-ictal” or “pre-seizure state" is introduced. Here it would be appropriate to note that such a state exists always per definition; the point of all these analyses is to identify a “pro-ictal” state, i.e. a state that leads to a seizure within a relatively short time. 2. On page 4 and Fig. 2 it would be useful to be more precise in defining quantitatively the “warning time” or “prediction horizon”, since the length of the latter is essential for an evaluation of the performance of any “prediction algorithm”. 3. On page 3 (top) there are “Clinical findings” mentioned that support, not the existence of a pre-seizure state, as written, since this always exists, but the fact that a state with distinct properties can be identified in the period preceding a seizure. 4. On page 5 the term “learning set” may be added to “.…part of the data” after “this optimization must be performed on a .…”. 5. On page 6 it should be stressed that the large majority of studies carried out during the 1990’s were performed using electrodes implanted in the brain while video-EEG monitoring was carried out in the framework of a pre-surgical evaluation. 6. It should be noted that most algorithms used for seizure prediction make use of EEG signals that occur spontaneously. An exception is the approach proposed by Kalitzin et al (2005) that used a stimulation paradigm in order to forecast the probability of a seizure occurring within a given time period. This difference between the two approaches – (a) one based on features extracted from spontaneous EEG signals, and (b) another based on local EEG responses elicited by electrical stimulation - should be emphasized. In general this is very good review.
Reply to Reviewer D
1. The term ‘pre-ictal state’ I refers to a dynamical state that is presumed to be different from the inter-ictal state and that can therefore be distinguished and detected. Alternatively, the brain could continue to be in the inter-ictal state throughout the pre-ictal period and then abruptly transition into a seizure. To make this clearer, I added the sentence ‘The crucial question thus is whether a pre-seizure state exists that can be distinguished from the inter-ictal state. Since such a state would reflect an increased probability of seizure occurrence, it is also sometimes referred to as a pro-ictal state.’
2. Added a sentence describing the range of prediction horizons found in the literature. Rephrased the sentence stating that a false alarm can be recognized as such only after the full duration of the prediction horizon.
3. See point 1.
4. Added the terms ‘training set’ and ‘test set’.
5. Added the sentence ‘These and later studies predominantly analyzed EEG signals from patients undergoing video-EEG monitoring, with chronic electrodes implanted directly inside or on the surface of the brain to localize the seizure focus for possible surgical resection.’
6. The study by Kalitzin et al. was already mentioned in the section ‘Understanding the Mechanisms of Seizure Generation’. I added the sentence ‘Remarkably, this study used active external perturbation of ongoing EEG activity rather than passive extraction of features from spontaneous EEG signals to detect impending seizures.’
Another review
Review of Mormann, Scholarpedia
This is a very good short review of the field and it is well written to be accessible to the non-specialist. I have only a few minor comments.
Last sentence of introduction: I suggest removing “in people who never had a seizure” because the same applies in people who had a seizure. Seizure prediction has nothing to do with diagnostic or prognostic value of pathological EEG in anybody.
Figure 1: there should be some sort of time scale, with an explanation that this is an example of time. Or some indication of ranges of time scales.
The following sentence is not correct: “In principle there are two types of scenarios of how a seizure could evolve”. The word “evolve” implies the evolution of the seizure from its onset to its end. I suggest “occur”. The same applies to the first sentence of the section on Mechanisms.
I do not agree that the first Lopes da Silva scenario is more likely in generalized epilepsy and the second more likely in focal epilepsy. Generalized seizures themselves usually have an abrupt onset; focal seizures have sometimes an abrupt onset and sometimes a gradual onset. This should not be confused with the preictal state, and I do not think there is evidence for this dichotomy. If there is, it would be nice to quote it.
Figure 2: the words “warning” and ‘alarms” are a bit mixed up. I am not sure of the difference between them; the title talks about true and false warning but the figure shows true and false alarms. Consistent terminology would help. I was also not sure from the variations of the characterizing measure what triggers an alarm. It seems to be a crossing, up or down, of the threshold, but then the last alarm does not correspond to any crossing. I have the feeling it would be clearer if there was an alarm only on crossing the threshold going up, and if this was stated in the caption.
The sentence: “…reported findings remained on an exemplary level” must be modified as ‘exemplary” means “remarkable”, “extraordinarily good” and does not refer to the fact that results mostly came from a few examples.
Reply to 'Another review'
Last sentence of introduction: I suggest removing “in people who never had a seizure” because the same applies in people who had a seizure. Seizure prediction has nothing to do with diagnostic or prognostic value of pathological EEG in anybody.
Agreed and modified.
Figure 1: there should be some sort of time scale, with an explanation that this is an example of time. Or some indication of ranges of time scales.
I added an example for a time scale in the caption of Figure 2.
The following sentence is not correct: “In principle there are two types of scenarios of how a seizure could evolve”. The word “evolve” implies the evolution of the seizure from its onset to its end. I suggest “occur”. The same applies to the first sentence of the section on Mechanisms.
Agreed and modified.
I do not agree that the first Lopes da Silva scenario is more likely in generalized epilepsy and the second more likely in focal epilepsy. Generalized seizures themselves usually have an abrupt onset; focal seizures have sometimes an abrupt onset and sometimes a gradual onset. This should not be confused with the preictal state, and I do not think there is evidence for this dichotomy. If there is, it would be nice to quote it.
The sentence about the two different scenarios reflects the view held by many people in the field that seizures in primary generalized, hereditary epilepsy syndromes are inherently unpredictable whereas focal seizures (including those with an abrupt beginning and secondary generalized seizures) may follow a stereotypic 'dynamical route' that leads to their occurrence, which makes them in principle detectable. To avoid the impression of a strict dichotomy I changed 'more likely in focal epilepsy' to 'more common in focal epilepsy'. Other than that I believe the sentence is phrased weakly enough to be valid.
Figure 2: the words “warning” and ‘alarms” are a bit mixed up. I am not sure of the difference between them; the title talks about true and false warning but the figure shows true and false alarms. Consistent terminology would help. I was also not sure from the variations of the characterizing measure what triggers an alarm. It seems to be a crossing, up or down, of the threshold, but then the last alarm does not correspond to any crossing. I have the feeling it would be clearer if there was an alarm only on crossing the threshold going up, and if this was stated in the caption.
I modified the figure caption to clarify that alarms are issued continuously for as long as the measure profile remains above the threshold. 'warnings' and 'alarms' are used synonymously in this article.
The sentence: “…reported findings remained on an exemplary level” must be modified as ‘exemplary” means “remarkable”, “extraordinarily good” and does not refer to the fact that results mostly came from a few examples.
Agreed and modified.
