TY  - JOUR
AU  - Singh, Devesh
AU  - Brima, Yusuf
AU  - Levin, Fedor
AU  - Becker, Martin
AU  - Hiller, Bjarne
AU  - Hermann, Andreas
AU  - Villar-Munoz, Irene
AU  - Beichert, Lukas
AU  - Bernhardt, Alexander
AU  - Buerger, Katharina
AU  - Butryn, Michaela
AU  - Dechent, Peter
AU  - Düzel, Emrah
AU  - Ewers, Michael
AU  - Fliessbach, Klaus
AU  - Freiesleben, Silka D
AU  - Glanz, Wenzel
AU  - Hetzer, Stefan
AU  - Janowitz, Daniel
AU  - Görß, Doreen
AU  - Kilimann, Ingo
AU  - Kimmich, Okka
AU  - Laske, Christoph
AU  - Levin, Johannes
AU  - Lohse, Andrea
AU  - Luesebrink, Falk
AU  - Munk, Matthias
AU  - Perneczky, Robert
AU  - Peters, Oliver
AU  - Preis, Lukas
AU  - Priller, Josef
AU  - Prudlo, Johannes
AU  - Prychynenko, Diana
AU  - Rauchmann, Boris Stephan
AU  - Rostamzadeh, Ayda
AU  - Roy-Kluth, Nina
AU  - Scheffler, Klaus
AU  - Schneider, Anja
AU  - Droste Zu Senden, Louise
AU  - Schott, Björn H
AU  - Spottke, Annika
AU  - Synofzik, Matthis
AU  - Wiltfang, Jens
AU  - Jessen, Frank
AU  - Weber, Marc-André
AU  - Teipel, Stefan J
AU  - Dyrba, Martin
TI  - An unsupervised XAI framework for dementia detection with context enrichment.
JO  - Scientific reports
VL  - 15
IS  - 1
SN  - 2045-2322
CY  - [London]
PB  - Springer Nature
M1  - DZNE-2025-01261
SP  - 39554
PY  - 2025
AB  - Explainable Artificial Intelligence (XAI) methods enhance the diagnostic efficiency of clinical decision support systems by making the predictions of a convolutional neural network's (CNN) on brain imaging more transparent and trustworthy. However, their clinical adoption is limited due to limited validation of the explanation quality. Our study introduces a framework that evaluates XAI methods by integrating neuroanatomical morphological features with CNN-generated relevance maps for disease classification. We trained a CNN using brain MRI scans from six cohorts: ADNI, AIBL, DELCODE, DESCRIBE, EDSD, and NIFD (N = 3253), including participants that were cognitively normal, with amnestic mild cognitive impairment, dementia due to Alzheimer's disease and frontotemporal dementia. Clustering analysis benchmarked different explanation space configurations by using morphological features as proxy-ground truth. We implemented three post-hoc explanations methods: (i) by simplifying model decisions, (ii) explanation-by-example, and (iii) textual explanations. A qualitative evaluation by clinicians (N = 6) was performed to assess their clinical validity. Clustering performance improved in morphology enriched explanation spaces, improving both homogeneity and completeness of the clusters. Post hoc explanations by model simplification largely delineated converters and stable participants, while explanation-by-example presented possible cognition trajectories. Textual explanations gave rule-based summarization of pathological findings. Clinicians' qualitative evaluation highlighted challenges and opportunities of XAI for different clinical applications. Our study refines XAI explanation spaces and applies various approaches for generating explanations. Within the context of AI-based decision support system in dementia research we found the explanations methods to be promising towards enhancing diagnostic efficiency, backed up by the clinical assessments.
KW  - Humans
KW  - Dementia: diagnosis
KW  - Dementia: diagnostic imaging
KW  - Magnetic Resonance Imaging: methods
KW  - Male
KW  - Female
KW  - Aged
KW  - Cognitive Dysfunction: diagnostic imaging
KW  - Cognitive Dysfunction: diagnosis
KW  - Alzheimer Disease: diagnostic imaging
KW  - Alzheimer Disease: diagnosis
KW  - Neural Networks, Computer
KW  - Artificial Intelligence
KW  - Brain: diagnostic imaging
KW  - Brain: pathology
KW  - Aged, 80 and over
KW  - Alzheimer’s disease (Other)
KW  - Brain volumetry (Other)
KW  - Explainable artificial intelligence (XAI) (Other)
KW  - Frontotemporal dementia (Other)
KW  - Magnetic resonance imaging (Other)
KW  - Neurodegenerative diseases (Other)
KW  - Qualitative evaluation (Other)
LB  - PUB:(DE-HGF)16
C6  - pmid:41224940
C2  - pmc:PMC12612133
DO  - DOI:10.1038/s41598-025-26227-2
UR  - https://pub.dzne.de/record/282290
ER  -