MEG localization of language-specific cortex utilizing MR-FOCUSS

To the Editor: Bowyer et al.1 present a novel mathematical approach for determining patterns of brain activity associated with performance of language tasks. The software developed by the authors has two main advantages, which could, in theory, ensure a level of confidence sufficient for presurgical planning applications. First, it takes into account the anatomy of the brain, constraining potential “sources” of magnetic activity based on physiologic and physical considerations. Second, and perhaps most important, the software is designed to operate unsupervised, significantly reducing the impact of subjective input from the user, a limitation characteristic of the “standard” method of the iterative application of the single equivalent current dipole (ECD) that the Bowyer et al. method claims to surpass in efficiency. Incidentally, the ECD method remains the “gold standard” in magnetoencephalography (MEG) studies on the cerebral mechanisms of basic sensory functions and of language, because it is the only one that has been validated against invasive brain mapping techniques and postoperative outcome,2–5 yet Bowyer et al. do not acknowledge this. As with previous attempts to outline brain activity profiles using alternative mathematical approaches, the results reported by Bowyer et al. are both interesting and promising. Clinical applications of MEG will be greatly facilitated by the use of userindependent analysis techniques, provided, however, that the key requirement of external validation of activation maps that applies to every functional imaging method is fully met. External validation is particularly crucial for techniques that model activity in terms of spatially extensive sources (like the Bowyer et al. technique does), given the inherent uncertainty surrounding threshold selection for displaying activation images. Adopting different image thresholding criteria may modify the extent of cortical regions that appear active, thereby seriously affecting the utility of the technique for presurgical mapping applications. Hopefully, the method that Bowyer et al. advocate will soon meet the validation requirement and emerge as a “more sensitive and useful technique” than the “standard” one which has been validated. Yet until this transpires, we believe it is advisable not to confound hope with fact. Incidentally, we also believe it is advisable not to confound literary genres (i.e. epic) with segments or durations of time-series (i.e. epochs) which the authors repeatedly do in their text.