Optimizing N3 parameters leads to better segmentation accuracy on 3T scanners

Introduction. Intensity non-uniformity can have detrimental effect on the accuracy of automated measurements of brain structures and can often be corrected only retrospectively. Among existing correction approaches, the nonparametric non-uniformity intensity normalization method N3 (1) is one of the most frequently used. However, at least one recent study (2) suggested that its performance on 3T scanners with multichannel phased-array receiver coils can be improved by optimizing a parameter that controls the smoothness of the estimated bias field. The present study not only confirms this finding but also demonstrates the beneficial effect of reduced to 30-50mm parameter values (default value is 200mm) on the quality of white matter surface estimation and reliability of cortical thickness and subcortical volume measurements.