Unsupervised Abnormality Detection through Mixed Structure Regularization (MSR) in Deep Sparse Autoencoders

TL;DR

This paper proposes a novel deep autoencoder regularization method that enhances unsupervised abnormality detection, particularly in medical imaging, by improving performance through mixed structure regularization and noise robustness.

Contribution

It introduces mixed structure regularization (MSR) in deep sparse autoencoders, improving unsupervised abnormality detection without requiring abnormal training data.

Findings

MSR improves detection performance by 20-30%.

Enhanced robustness to input noise.

Effective in medical imaging applications.

Abstract

Deep sparse auto-encoders with mixed structure regularization (MSR) in addition to explicit sparsity regularization term and stochastic corruption of the input data with Gaussian noise have the potential to improve unsupervised abnormality detection. Unsupervised abnormality detection based on identifying outliers using deep sparse auto-encoders is a very appealing approach for medical computer aided detection systems as it requires only healthy data for training rather than expert annotated abnormality. In the task of detecting coronary artery disease from Coronary Computed Tomography Angiography (CCTA), our results suggests that the MSR has the potential to improve overall performance by 20-30% compared to deep sparse and denoising auto-encoders.