Selection bias in the reported performances of AD classification pipelines - Archive ouverte HAL Access content directly
Journal Articles Neuroimage-Clinical Year : 2017

Selection bias in the reported performances of AD classification pipelines

(1) , (1) , (2) , (3) , (1)
1
2
3

Abstract

The last decade has seen a great proliferation of supervised learning pipelines for individual diagnosis and prognosis in Alzheimer's disease. As more pipelines are developed and evaluated in the search for greater performance, only those results that are relatively impressive will be selected for publication. We present an empirical study to evaluate the potential for optimistic bias in classification performance results as a result of this selection. This is achieved using a novel, resampling-based experiment design that effectively simulates the optimisation of pipeline specifications by individuals or collectives of researchers using cross validation with limited data. Our findings indicate that bias can plausibly account for an appreciable fraction (often greater than half) of the apparent performance improvement associated with the pipeline optimisation, particularly in small samples. We discuss the consistency of our findings with patterns observed in the literature and consider strategies for bias reduction and mitigation.
Fichier principal
Vignette du fichier
1-s2.0-S221315821630256X-main.pdf (1.59 Mo) Télécharger le fichier
Origin : Publication funded by an institution
Loading...

Dates and versions

hal-01843390 , version 1 (27-07-2018)

Licence

Attribution - CC BY 4.0

Identifiers

Cite

Alex F. Mendelson, Maria A. Zuluaga, Marco Lorenzi, Brian F. Hutton, Sébastien Ourselin. Selection bias in the reported performances of AD classification pipelines. Neuroimage-Clinical, 2017, 14, pp.400 - 416. ⟨10.1016/j.nicl.2016.12.018⟩. ⟨hal-01843390⟩

Collections

INRIA INRIA2
151 View
80 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More