Our ultimate goal is to develop a clinically useful method for assigning risk of developing SWS brain involvement to babies born with facial PW birthmark, in order to target prophylactic interventions to babies at highest risk and to provide education for their parents, while providing reassurance to the parents of babies at lowest risk. At this time, the diagnosis of SWS is often made only when a child has already become symptomatic, however. MRI is the gold-standard neuroimaging technique to look for brain involvement, however the risks associated with repeated administration of sedation and contrast make serial MRI of the asymptomatic baby with PW birthmark impractical.
We hypothesize that a qEEG method may be a non-invasive, quick and relatively inexpensive method that can be repeated in order to permit optimal timing of MRI. This study builds on previous research by demonstrating that qEEG can discriminate between infants with SWS brain involvement and those who have a neurologically asymptomatic facial PW birthmark. Although this study does not (nor was designed to) demonstrate that qEEG is statistically better than a fellowship-trained pediatric electroencephalographer experienced in reading EEGs from individuals with SWS, we do note that the diagnostic accuracy of the clinical reads increased significantly from the Initial cohort to the Validation cohort, perhaps suggesting that accuracy in assessing these sorts of asymmetries by visual inspection comes with experience. An electroencephalographer with such qualifications and experience would not be available to most pediatricians who are seeing an infant with a PW birthmark, while this qEEG technique could be made widely available. In this technique, the only operator-dependent steps were the selection of epochs and the decision whether to exclude a channel with persistent artifact. This was accomplished by research assistants who were trained in recognition of artifact but who had no other formal EEG training. The use of research assistants helps demonstrate that this technique is not reliant on the same level of expertise that produced the standard clinical visual EEG interpretations in this study.
This study demonstrates the feasibility of a qEEG technique for discriminating reliably between infants with SWS vs. those with neurologically asymptomatic PW birthmarks. Our ultimate goal is to determine whether this technique can detect asymmetries before a patient becomes symptomatic, but the current study is limited by a dearth of infant subjects who were asymptomatic at the time of EEG but who later went on to develop either clinical or radiological signs of SWS. It remains to be seen whether qEEG changes can be detected before a patient becomes symptomatic and can be used to guide whom to image and when. The fact that 3 subjects had a positive qEEG prior to the development of symptoms provides at least a proof of principle that this may be the case.
Given that only a portion of at-risk babies with a PW birthmark develops the neurological involvement characteristic of SWS, it will require significantly larger numbers of subjects to power a study to look at the ability of qEEG to screen presymptomatically for brain involvement. To that end, our center has begun to collect EEGs performed at outside locations for analysis. In this way, we aim to have a sufficient sample to determine whether or not this technique will be helpful as a screening tool.
Although we currently conceive of qEEG’s potential as a “gatekeeper” to MRI, it is difficult to study these modalities together, e.g., with MRIs performed at predetermined ages. However, the performance of sedated MRIs with contrast is widely considered to be an unacceptable research risk in infants, and all MRIs in this study were preformed in the line of standard clinical care.
Given larger numbers of participants, we expect to be able to determine positive and negative predictive values for qEEG, as well as perhaps a graduated scale of risk (rather than a dichotomous scale). In order to make this technique widely available, we aim to automate the process as much as possible. We expect that the approach will be modified as we extend its utility to different age groups and clinical situations. We aim to modify the qEEG threshold to be used for following cerebral function for research and clinical purposes (e.g., to predict and evaluate response to neuroprotective interventions). Future work will address not only these issues, but will also look at serial data collected to chart the natural history of qEEG in SWS, to determine whether serial qEEG can predict clinical deterioration and to determine whether qEEG can be used as a marker of “brain health,” in situations such as the use of prophylactic interventions. This work may make use of statistical methods that can find efficient thresholds to differentiate between groups, such as Linear Discriminant Analysis, Bayesian Classifiers or Support Vector Machine.
In summary, quantitative EEG shows promise as an objective and measurable biomarker of neurological involvement in children with potential SWS brain involvement. A threshold that was developed in one cohort of infants with PW birthmark was validated on a second cohort of 9 infants. Further research involving infants who are negative at the time of qEEG but who later go on to develop signs SWS brain involvement will be needed to demonstrate qEEG as a valid predictor.