Abstract #3349

Robust Intensity Normalisation and Automatic Intensity Window Selection for the Calculation of the Boundary Shift Integral

Kelvin Ka-fai Leung¹, Matthew Clarkson², Nick C. Fox², Sebastien Ourselin^1,2

¹Centre for Medical Image Computing, University College London, London, UK; ²Dementia Research Centre, University College London, London, UK

The boundary shift integral (BSI) is a robust measure of regional and global cerebral atrophy rates. We describe a new and robust intensity normalisation method to calculate the BSI and applied it to a subset of images from the ADNI database. Initial results show that the new method improves group separation between AD and control groups, with effect size increased from 0.51 to 0.97 as measured by Cohens d. This implies that the new method may be more robust to changes in image quality between baseline and repeat scans (e.g. due to scanner upgrade).