Attribute Regularized Soft Introspective Variational Autoencoder for Interpretable Cardiac Disease Classification
This work addresses interpretability for clinicians in cardiac disease diagnosis, but it is incremental as it builds on existing variational autoencoder methods with attribute regularization.
The paper tackled the problem of interpretability in medical imaging by proposing an attribute-regularized soft introspective variational autoencoder, which improved latent space interpretability and addressed blurry reconstructions in cardiac MRI data, with analysis showing that disease classification relied on regularized dimensions aligning with clinical observations.
Interpretability is essential in medical imaging to ensure that clinicians can comprehend and trust artificial intelligence models. In this paper, we propose a novel interpretable approach that combines attribute regularization of the latent space within the framework of an adversarially trained variational autoencoder. Comparative experiments on a cardiac MRI dataset demonstrate the ability of the proposed method to address blurry reconstruction issues of variational autoencoder methods and improve latent space interpretability. Additionally, our analysis of a downstream task reveals that the classification of cardiac disease using the regularized latent space heavily relies on attribute regularized dimensions, demonstrating great interpretability by connecting the used attributes for prediction with clinical observations.