A statistical method (cross-validation) for bone loss region detection after spaceflight

Abstract

Astronauts experience bone loss after the longspaceflight missions. Identifying specific regions thatundergo the greatest losses (e.g. the proximal femur) couldreveal information about the processes of bone loss in disuseand disease. Methods for detecting such regions, however,remains an open problem. This paper focuses on statisticalmethods to detect such regions. We perform statisticalparametric mapping to get t-maps of changes in images, andpropose a new cross-validation method to select an optimumsuprathreshold for forming clusters of pixels. Once thesecandidate clusters are formed, we use permutation testing oflongitudinal labels to derive significant changes. © 2010 The Author(s).