Unsupervised Abnormality Detection through Mixed Structure Regularization (MSR) in Deep Sparse Autoencoders
This work addresses the problem of detecting abnormalities in medical imaging without needing expert annotations, which is incremental as it builds on autoencoder-based approaches.
The paper tackled unsupervised abnormality detection for coronary artery disease from CCTA images by proposing a mixed structure regularization (MSR) in deep sparse autoencoders, resulting in a 20-30% performance improvement over existing methods.
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.