Manual segmentation qualification platform for the EADC‐ADNI harmonized protocol for hippocampal segmentation project

Abstract

This platform is an adequate infrastructure allowing standardized training and evaluation of tracers' compliance with the HarP. This is a necessary step allowing the use of hippocampal volumetry as a biomarker for AD in clinical and research centers.

The EADC‐ADNI Harmonized Protocol for manual hippocampal segmentation on magnetic resonance: Evidence of validity

“…The platform provided color-coded visual feedback on compliance (or departure) from benchmark labels' segmentation, by color-coding the extent of departure on a red-to-green scale, where red denoted departure and green compliance. The platform also provided Dice and Jaccard measures of accuracy for each segmented slice and for the overall hippocampus [14]. At the end of the first two rounds, tracers received detailed written feedback from the project manager (M. Boccardi) illustrating HarP violations according to the color-coded visual feedback.…”

“…Segmentation inaccuracies were checked by the project manager, tracers were asked to edit segmentation inaccuracies if identified, re-upload edited images, and segment new scans. After 3 such training rounds, tracers were asked to segment and upload 10 new scans (qualification phase [14]). In the Qualification phase, tracers received only feedback regarding general performance, and corrections were not allowed.…”

“…Thirteen tracers successfully completed the qualification procedure and Phase I; one tracer was trained and had served as master tracer and did not need to undergo the qualification procedure. At the end of the qualification procedure tracers had very good reliability indices, with Dice values ranging from 0.88 to 0.92 for 3T images, and 0.87 to 0.91 for 1.5 T images [14]. These 14 tracers resegmented the same ADNI cases following the HarP (mean delay following local protocol segmentation of one year).…”

“…These master segmentations were checked, corrected and certified as the “benchmark” labels to be used for the qualification of any future human tracer or automated segmentation procedure [13]. An online platform, freely available at www.hippocampal-protocol.net, was developed to qualify new (“naïve”) tracers to the use of the HarP based on the benchmark labels [14]. …”

Associations between hippocampal morphometry and neuropathologic markers of Alzheimer's disease using 7 T MRI

“…Recently, neuroimaging techniques have allowed for 3D modeling of hippocampal changes and have been valuable for the longitudinal study of neuropathologic changes in AD (Apostolova et al, 2006c, Apostolova et al, 2010a, Duchesne et al, 2015). The present study uses a computational anatomy shape-based method called hippocampal radial distance in order to compare cross-sectional 3D hippocampal maps of subjects diagnosed with dementia and normal controls.…”