Towards harmonizing subtyping methods for neuroimaging studies in Alzheimer's disease [article]

Rosaleena Mohanty, Gustav Mårtensson, Konstantinos Poulakis, J-Sebastian Muehlboeck, Elena Rodriguez Vieitez, Konstantinos Chiotis, Michel J. Grothe, Agneta Nordberg, Daniel Ferreira, Eric Westman
2020 medRxiv   pre-print
Biological subtypes in Alzheimer's disease (AD), originally identified on neuropathological data, have been translated to in vivo biomarkers such as structural magnetic resonance imaging (sMRI) and positron emission tomography (PET), to disentangle the heterogeneity within AD. Although there is methodological variability across studies, comparable characteristics of subtypes are reported at the group level. In this study, we investigated whether group-level similarities translate to
more » ... evel agreement across subtyping methods, in a head-to-head context. Methods: We compared five previously published subtyping methods. Firstly, we validated the subtyping methods in 89 amyloid-beta positive (Aβ+) AD dementia patients (reference group: 70 Aβ- healthy individuals; HC) using sMRI. Secondly, we extended and applied the subtyping methods to 53 Aβ+ prodromal AD and 30 Aβ+ AD dementia patients (reference group: 200 Aβ- HC) using both sMRI and tau PET. Subtyping methods were implemented as outlined in each original study. Group-level and individual-level comparisons across methods were performed. Results: Each individual method was replicated and the proof-of-concept was established. All methods captured subtypes with similar patterns of demographic and clinical characteristics, and with similar maps of cortical thinning and tau PET uptake, at the group level. However, large disagreements were found at the individual level. Conclusions: Although characteristics of subtypes may be comparable at the group level, there is a large disagreement at the individual level across subtyping methods. Therefore, there is an urgent need for consensus and harmonization across subtyping methods. We call for establishment of an open benchmarking framework to overcome this problem.
doi:10.1101/2020.04.19.20064881 fatcat:tzn47yajmzgh5m6lku72j53zgm