Automated computation of the Gyrification Index in prefrontal lobes: methods and comparison with manual implementation

In this paper, we introduce an automated method of calculating Gyrification Index (GI), a measure of cortical folding. Automated GI (A-GI) is an in vivo GI implementation applied to MRI T1 weighted scans and is designed as an extension to the SPM analysis package. The A-GI tool is unbiased in its application, and is unlimited in the size of test cohort to which it can be applied. In comparison to manual methods, A-GI substantially reduces the time costs and improves repeatability. The current A-GI implementation is limited to analysis of prefrontal lobes, but an extension to provide whole brain A-GI is under consideration. In determination of the GI inner contour, A-GI traces high spatial frequencies typically missed in manual tracing, and thus, A-GI reports a high GI value. We examine the operation of this tool in two scan cohorts. We establish that the tool has good repeatability through its application to a cohort where 5 well individuals were scanned 5 times over a period of 6 months. This indicates that A-GI has low susceptibility to scanner noise and is not affected by the variability in brain representation given by repeat scans. We demonstrate replication of hand tracing results by comparisons with a manual GI study that has shown differences between high risk subjects who go on to develop schizophrenia and those who are at high risk but remain well. Direct scan by scan comparisons are carried out between manual and A-GI methods. In respect of scan orientation and coronal sampling, the methods differ, and these considerations contribute to a between methods right prefrontal ICC of 0.67 and left prefrontal ICC of 0.63. The replication results demonstrate that A-GI has discriminatory power equivalent to manual methods. A-GI is therefore a reliable measure of cortical folding that could be usefully applied to a number of MRI data sets of the brain in health and disease.