Intra-Session Reliability Metrics for Quality Assurance in Pre-Surgical Mapping with Magnetoencephalography

Abstract

Magnetoencephalography (MEG) functional maps can localize brain activity for pre-surgical mapping, but their quality is difficult to quantify. Clinical standard metrics cannot be used when multiple sources of activity are distributed across the brain. This thesis validates the use of reliable fraction, a novel intra-session reliability metric, for focal maps. Scans were acquired in 'good' and 'poor' conditions, in which common MEG quality issues were simulated. Clinical standard methods and reliable fraction, along with two other possible metrics (the Dice and Pearson coefficients), were used to assess data quality. High quality data proved difficult to achieve, highlighting the need for robust quality assurance procedures. Reliable fraction was more sensitive to data quality issues than the Dice or Pearson coefficients. Comparison of reliable fraction with clinical standard metrics showed comparable sensitivity to changes in data quality and suggests reliable fraction may be a useful metric for cases of distributed brain activity.