The presence of confounding effects (or biases) is one of the most critical challenges in using deep learning for medical imaging studies. Confounders influence the relationship between input data (e.g., brain MRIs) and output variables (e.g., diagnosis), often leading to spurious and biased associations. Traditional machine learning and statistical models mitigate confounders by techniques such as matching datasets, stratification, or residualizing imaging measurements. However, deep learning models, which rely on end-to-end training to extract informative features automatically, require alternative strategies. To address this, we work on a series of approaches that could remove confounding effects while learning features in modern architectures.