Poster (After Call)

FZJ-2023-01734

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png The Confound Continuum: A 2D confounder assessment for AI in precision medicine

Komeyer, V.FZJ* ; Eickhoff, S.FZJ* ; Grefkes, C. ; Raimondo, F.FZJ* ; Patil, K.FZJ*

2023

Please use a persistent id in citations: http://hdl.handle.net/2128/34262

Abstract: Confounding presents a major challenge in neuroimaging machine learning applications. Confounderscan influence both, brain-derived features and phenotypical targets1. Removing theirsignal from the data changes the feature-target relationship which ultimately affects the model interpretation.Additionally, confounders are not always straightforward to identify. To target this,we introduce the idea of a 2D Confound Continuum (CC). Its ordinate evaluates the strength ofthe statistical relationship between a confound and the feature(s)/target, thereby helping to betterunderstand its signal contributions to the data (statistical CC). Its abscissa defines the strength ofthe conceptual or biological relationship and hence the effects of removal on the model interpretation(conceptual CC). Sorting potential confounders within the CC can help to better understandtheir role and impact on building predictive models.

Note: This research was supported by the Joint Lab “Supercomputing and Modeling for the Human Brain”.

---

Contributing Institute(s):

1. Gehirn & Verhalten (INM-7)

Research Program(s):

1. 5251 - Multilevel Brain Organization and Variability (POF4-525) (POF4-525)
2. JL SMHB - Joint Lab Supercomputing and Modeling for the Human Brain (JL SMHB-2021-2027) (JL SMHB-2021-2027)

Appears in the scientific report 2023

---

Database coverage:

---