Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease

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 comprehen...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Martin Dyrba, Moritz Hanzig, Slawek Altenstein, Sebastian Bader, Tommaso Ballarini, Frederic Brosseron, Katharina Buerger, Daniel Cantré, Peter Dechent, Laura Dobisch, Emrah Düzel, Michael Ewers, Klaus Fliessbach, Wenzel Glanz, John-Dylan Haynes, Michael T. Heneka, Daniel Janowitz, Deniz B. Keles, Ingo Kilimann, Christoph Laske, Franziska Maier, Coraline D. Metzger, Matthias H. Munk, Robert Perneczky, Oliver Peters, Lukas Preis, Josef Priller, Boris Rauchmann, Nina Roy, Klaus Scheffler, Anja Schneider, Björn H. Schott, Annika Spottke, Eike J. Spruth, Marc-André Weber, Birgit Ertl-Wagner, Michael Wagner, Jens Wiltfang, Frank Jessen, Stefan J. Teipel, for the ADNI, AIBL, DELCODE study groups
Formato:	article
Lenguaje:	EN
Publicado:	BMC 2021
Materias:	Alzheimer’s disease Deep learning Convolutional neural network MRI Layer-wise relevance propagation Neurosciences. Biological psychiatry. Neuropsychiatry RC321-571 Neurology. Diseases of the nervous system RC346-429
Acceso en línea:	https://doaj.org/article/4ea07e2a7a2f4319ab977e9fcdcd6e3d
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:4ea07e2a7a2f4319ab977e9fcdcd6e3d
record_format	dspace
spelling	oai:doaj.org-article:4ea07e2a7a2f4319ab977e9fcdcd6e3d2021-11-28T12:38:24ZImproving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease10.1186/s13195-021-00924-21758-9193https://doaj.org/article/4ea07e2a7a2f4319ab977e9fcdcd6e3d2021-11-01T00:00:00Zhttps://doi.org/10.1186/s13195-021-00924-2https://doaj.org/toc/1758-9193Abstract 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 ≈ 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 ≈ −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.Martin DyrbaMoritz HanzigSlawek AltensteinSebastian BaderTommaso BallariniFrederic BrosseronKatharina BuergerDaniel CantréPeter DechentLaura DobischEmrah DüzelMichael EwersKlaus FliessbachWenzel GlanzJohn-Dylan HaynesMichael T. HenekaDaniel JanowitzDeniz B. KelesIngo KilimannChristoph LaskeFranziska MaierCoraline D. MetzgerMatthias H. MunkRobert PerneczkyOliver PetersLukas PreisJosef PrillerBoris RauchmannNina RoyKlaus SchefflerAnja SchneiderBjörn H. SchottAnnika SpottkeEike J. SpruthMarc-André WeberBirgit Ertl-WagnerMichael WagnerJens WiltfangFrank JessenStefan J. Teipelfor the ADNI, AIBL, DELCODE study groupsBMCarticleAlzheimer’s diseaseDeep learningConvolutional neural networkMRILayer-wise relevance propagationNeurosciences. Biological psychiatry. NeuropsychiatryRC321-571Neurology. Diseases of the nervous systemRC346-429ENAlzheimer’s Research & Therapy, Vol 13, Iss 1, Pp 1-18 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Alzheimer’s disease Deep learning Convolutional neural network MRI Layer-wise relevance propagation Neurosciences. Biological psychiatry. Neuropsychiatry RC321-571 Neurology. Diseases of the nervous system RC346-429
spellingShingle	Alzheimer’s disease Deep learning Convolutional neural network MRI Layer-wise relevance propagation Neurosciences. Biological psychiatry. Neuropsychiatry RC321-571 Neurology. Diseases of the nervous system RC346-429 Martin Dyrba Moritz Hanzig Slawek Altenstein Sebastian Bader Tommaso Ballarini Frederic Brosseron Katharina Buerger Daniel Cantré Peter Dechent Laura Dobisch Emrah Düzel Michael Ewers Klaus Fliessbach Wenzel Glanz John-Dylan Haynes Michael T. Heneka Daniel Janowitz Deniz B. Keles Ingo Kilimann Christoph Laske Franziska Maier Coraline D. Metzger Matthias H. Munk Robert Perneczky Oliver Peters Lukas Preis Josef Priller Boris Rauchmann Nina Roy Klaus Scheffler Anja Schneider Björn H. Schott Annika Spottke Eike J. Spruth Marc-André Weber Birgit Ertl-Wagner Michael Wagner Jens Wiltfang Frank Jessen Stefan J. Teipel for the ADNI, AIBL, DELCODE study groups Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
description	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 ≈ 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 ≈ −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.
format	article
author	Martin Dyrba Moritz Hanzig Slawek Altenstein Sebastian Bader Tommaso Ballarini Frederic Brosseron Katharina Buerger Daniel Cantré Peter Dechent Laura Dobisch Emrah Düzel Michael Ewers Klaus Fliessbach Wenzel Glanz John-Dylan Haynes Michael T. Heneka Daniel Janowitz Deniz B. Keles Ingo Kilimann Christoph Laske Franziska Maier Coraline D. Metzger Matthias H. Munk Robert Perneczky Oliver Peters Lukas Preis Josef Priller Boris Rauchmann Nina Roy Klaus Scheffler Anja Schneider Björn H. Schott Annika Spottke Eike J. Spruth Marc-André Weber Birgit Ertl-Wagner Michael Wagner Jens Wiltfang Frank Jessen Stefan J. Teipel for the ADNI, AIBL, DELCODE study groups
author_facet	Martin Dyrba Moritz Hanzig Slawek Altenstein Sebastian Bader Tommaso Ballarini Frederic Brosseron Katharina Buerger Daniel Cantré Peter Dechent Laura Dobisch Emrah Düzel Michael Ewers Klaus Fliessbach Wenzel Glanz John-Dylan Haynes Michael T. Heneka Daniel Janowitz Deniz B. Keles Ingo Kilimann Christoph Laske Franziska Maier Coraline D. Metzger Matthias H. Munk Robert Perneczky Oliver Peters Lukas Preis Josef Priller Boris Rauchmann Nina Roy Klaus Scheffler Anja Schneider Björn H. Schott Annika Spottke Eike J. Spruth Marc-André Weber Birgit Ertl-Wagner Michael Wagner Jens Wiltfang Frank Jessen Stefan J. Teipel for the ADNI, AIBL, DELCODE study groups
author_sort	Martin Dyrba
title	Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
title_short	Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
title_full	Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
title_fullStr	Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
title_full_unstemmed	Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease
title_sort	improving 3d convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in alzheimer’s disease
publisher	BMC
publishDate	2021
url	https://doaj.org/article/4ea07e2a7a2f4319ab977e9fcdcd6e3d
work_keys_str_mv	AT martindyrba improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT moritzhanzig improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT slawekaltenstein improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT sebastianbader improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT tommasoballarini improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT fredericbrosseron improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT katharinabuerger improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT danielcantre improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT peterdechent improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT lauradobisch improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT emrahduzel improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT michaelewers improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT klausfliessbach improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT wenzelglanz improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT johndylanhaynes improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT michaeltheneka improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT danieljanowitz improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT denizbkeles improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT ingokilimann improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT christophlaske improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT franziskamaier improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT coralinedmetzger improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT matthiashmunk improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT robertperneczky improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT oliverpeters improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT lukaspreis improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT josefpriller improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT borisrauchmann improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT ninaroy improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT klausscheffler improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT anjaschneider improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT bjornhschott improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT annikaspottke improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT eikejspruth improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT marcandreweber improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT birgitertlwagner improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT michaelwagner improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT jenswiltfang improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT frankjessen improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT stefanjteipel improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease AT fortheadniaibldelcodestudygroups improving3dconvolutionalneuralnetworkcomprehensibilityviainteractivevisualizationofrelevancemapsevaluationinalzheimersdisease
_version_	1718407888543678464

Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer’s disease

Ejemplares similares