Rotate to view desktop version ...

Spatio-Temporal Image Analysis

Longitudinal FreeSurfer
FreeSurfer is an open source software suite for the automatic processing and analysis of human brain MRI data with currently more than 23.000 licenses (> 3800 new in 2015). Our lab has contributed several modules to this package, most prominently the dedicated longitudinal processing pipeline. Longitudinal, or spatio-temporal, image analysis is the analysis of a sequence of images from the same subject to quantify changes of the brain across time. Instead of processing images completely independent, joint analysis of multiple time points can significantly reduce noise, and increase reliability and sensitivity of the methods. We demonstrate in [1,4] that using a within-subject template for information transfer and common initialization of segmentation and surface reconstruction algorithms increases statistical power and reduces required sample size by approximately 50% - or it allows detection of much smaller changes at the same sample size, as, e.g., required in drug trials.
An important aspect in longitudinal image analysis is the accidental introduction of a processing bias. For example, when follow-up images get registered and mapped to a baseline image, interpolation artifacts are introduced only in the follow-up images. This creates spurious effects (for interpolation artifacts usually spurious volume loss due to blurry edges) and can bias the whole study [2]. Our longitudinal framework in FreeSurfer avoids these and other types of processing bias by treating all input images exactly the same. E.g, they are registered to a joint mid-space and information, such as preliminary surfaces, are transferred only from the common within-subject template, but not directly across time points.
Selected Publications (full list):
Martin Reuter - MIT - Cambridge, MA, USA - EMail:
Last modified 14.01.2022 -- [xhtml] -- [css] -- [welcome]