Confounder-Aware Visualization of ConvNets

With recent advances in deep learning, neuroimaging studies increasingly rely on convolutional networks (ConvNets) to predict diagnosis based on MR images. To gain a better understanding of how a disease impacts the brain, the studies visualize the salience maps of the ConvNet highlighting voxels within the brain majorly contributing to the prediction. However, these salience maps are generally confounded, i.e., some salient regions are more predictive of confounding variables (such as age) than the diagnosis. To avoid such misinterpretation, we propose in this paper an approach that aims to visualize confounder-free saliency maps that only highlight voxels predictive of the diagnosis. The approach incorporates univariate statistical tests to identify confounding effects within the intermediate features learned by ConvNet. The influence from the subset of confounded features is then removed by a novel partial back-propagation procedure. We use this two-step approach to visualize confounder-free saliency maps extracted from synthetic and two real datasets. These experiments reveal the potential of our visualization in producing unbiased model-interpretation.

[1]  A. Dobson An introduction to generalized linear models , 1990 .

[2]  J. Kaye,et al.  The Significance of Age‐Related Enlargement of the Cerebral Ventricles in Healthy Men and Women Measured by Quantitative Computed X‐Ray Tomography , 1992, Journal of the American Geriatrics Society.

[3]  T. Jernigan,et al.  Development of cortical and subcortical brain structures in childhood and adolescence: a structural MRI study , 2002, Developmental medicine and child neurology.

[4]  D. Mackinnon Introduction to Statistical Mediation Analysis , 2008 .

[5]  A. Baghestani,et al.  How to control confounding effects by statistical analysis , 2012, Gastroenterology and hepatology from bed to bench.

[6]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[7]  Torsten Rohlfing,et al.  The National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA): A Multisite Study of Adolescent Development and Substance Use. , 2015, Journal of studies on alcohol and drugs.

[8]  David J. Sharp,et al.  Increased brain-predicted aging in treated HIV disease , 2017, Neurology.

[9]  Kilian M. Pohl,et al.  End-To-End Alzheimer's Disease Diagnosis and Biomarker Identification , 2018, MLMI@MICCAI.

[10]  Kilian M. Pohl,et al.  Chained regularization for identifying brain patterns specific to HIV infection , 2018, NeuroImage.

[11]  Sang Hyun Park,et al.  Alcohol use effects on adolescent brain development revealed by simultaneously removing confounding factors, identifying morphometric patterns, and classifying individuals , 2018, Scientific Reports.

[12]  Jie Xu,et al.  The practical implementation of artificial intelligence technologies in medicine , 2019, Nature Medicine.

[13]  Eric J Topol,et al.  High-performance medicine: the convergence of human and artificial intelligence , 2019, Nature Medicine.