Dyrba, Martin ORCID: 0000-0002-3353-3167, Hanzig, Moritz, Altenstein, Slawek, Bader, Sebastian, Ballarini, Tommaso, Brosseron, Frederic, Buerger, Katharina, Cantre, Daniel, Dechent, Peter, Dobisch, Laura, Duezel, Emrah, Ewers, Michael, Fliessbach, Klaus, Glanz, Wenzel, Haynes, John-Dylan, Heneka, Michael T., Janowitz, Daniel, Keles, Deniz B., Kilimann, Ingo, Laske, Christoph, Maier, Franziska, Metzger, Coraline D., Munk, Matthias H., Perneczky, Robert, Peters, Oliver, Preis, Lukas, Priller, Josef, Rauchmann, Boris, Roy, Nina, Scheffler, Klaus, Schneider, Anja, Schott, Bjoern H., Spottke, Annika, Spruth, Eike J., Weber, Marc-Andre, Ertl-Wagner, Birgit, Wagner, Michael ORCID: 0000-0003-2589-6440, Wiltfang, Jens, Jessen, Frank and Teipel, Stefan J. (2021). Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer's disease. Alzheimers Res. Ther., 13 (1). LONDON: BMC. ISSN 1758-9193

Full text not available from this repository.

Abstract

Background Although convolutional neural networks (CNNs) achieve high diagnostic accuracy for detecting Alzheimer's disease (AD) dementia based on magnetic resonance imaging (MRI) scans, they are not yet applied in clinical routine. One important reason for this is a lack of model comprehensibility. Recently developed visualization methods for deriving CNN relevance maps may help to fill this gap as they allow the visualization of key input image features that drive the decision of the model. We investigated whether models with higher accuracy also rely more on discriminative brain regions predefined by prior knowledge. Methods We trained a CNN for the detection of AD in N = 663 T1-weighted MRI scans of patients with dementia and amnestic mild cognitive impairment (MCI) and verified the accuracy of the models via cross-validation and in three independent samples including in total N = 1655 cases. We evaluated the association of relevance scores and hippocampus volume to validate the clinical utility of this approach. To improve model comprehensibility, we implemented an interactive visualization of 3D CNN relevance maps, thereby allowing intuitive model inspection. Results Across the three independent datasets, group separation showed high accuracy for AD dementia versus controls (AUC >= 0.91) and moderate accuracy for amnestic MCI versus controls (AUC approximate to 0.74). Relevance maps indicated that hippocampal atrophy was considered the most informative factor for AD detection, with additional contributions from atrophy in other cortical and subcortical regions. Relevance scores within the hippocampus were highly correlated with hippocampal volumes (Pearson's r approximate to -0.86, p < 0.001). Conclusion The relevance maps highlighted atrophy in regions that we had hypothesized a priori. This strengthens the comprehensibility of the CNN models, which were trained in a purely data-driven manner based on the scans and diagnosis labels. The high hippocampus relevance scores as well as the high performance achieved in independent samples support the validity of the CNN models in the detection of AD-related MRI abnormalities. The presented data-driven and hypothesis-free CNN modeling approach might provide a useful tool to automatically derive discriminative features for complex diagnostic tasks where clear clinical criteria are still missing, for instance for the differential diagnosis between various types of dementia.

Item Type: Journal Article
Creators:
CreatorsEmailORCIDORCID Put Code
Dyrba, MartinUNSPECIFIEDorcid.org/0000-0002-3353-3167UNSPECIFIED
Hanzig, MoritzUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Altenstein, SlawekUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Bader, SebastianUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Ballarini, TommasoUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Brosseron, FredericUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Buerger, KatharinaUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Cantre, DanielUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Dechent, PeterUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Dobisch, LauraUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Duezel, EmrahUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Ewers, MichaelUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Fliessbach, KlausUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Glanz, WenzelUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Haynes, John-DylanUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Heneka, Michael T.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Janowitz, DanielUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Keles, Deniz B.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Kilimann, IngoUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Laske, ChristophUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Maier, FranziskaUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Metzger, Coraline D.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Munk, Matthias H.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Perneczky, RobertUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Peters, OliverUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Preis, LukasUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Priller, JosefUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Rauchmann, BorisUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Roy, NinaUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Scheffler, KlausUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Schneider, AnjaUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Schott, Bjoern H.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Spottke, AnnikaUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Spruth, Eike J.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Weber, Marc-AndreUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Ertl-Wagner, BirgitUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Wagner, MichaelUNSPECIFIEDorcid.org/0000-0003-2589-6440UNSPECIFIED
Wiltfang, JensUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Jessen, FrankUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Teipel, Stefan J.UNSPECIFIEDUNSPECIFIEDUNSPECIFIED
URN: urn:nbn:de:hbz:38-605049
DOI: 10.1186/s13195-021-00924-2
Journal or Publication Title: Alzheimers Res. Ther.
Volume: 13
Number: 1
Date: 2021
Publisher: BMC
Place of Publication: LONDON
ISSN: 1758-9193
Language: English
Faculty: Unspecified
Divisions: Unspecified
Subjects: no entry
Uncontrolled Keywords:
KeywordsLanguage
ATROPHY; MRIMultiple languages
Clinical Neurology; NeurosciencesMultiple languages
URI: http://kups.ub.uni-koeln.de/id/eprint/60504

Downloads

Downloads per month over past year

Altmetric

Export

Actions (login required)

View Item View Item