"The presence of confounding effects is inarguably one of the most critical challenges in medical applications. They influence both input (e.g., neuroimages) and the output (e.g., diagnosis or clinical score) variables and may cause spurious associations when not properly controlled for. Confounding effect removal is particularly difficult for a wide range of state-of-the-art prediction models, including deep learning methods. These methods operate directly on images and extract features in an end-to-end manner. This prohibits removing confounding effects by traditional statistical analysis, which often requires precomputed features (image measurements). In this talk, I will present methods to learn confounder-invariant discriminative features and novel normalization techniques to remove confounding and bias effects while training neural networks.

"