Dual Decomposition of Convex Optimization Layers for Consistent Attention in Medical Images
This work addresses interpretability challenges for medical imaging applications, offering a domain-specific improvement over existing methods.
The paper tackled the problem of noisy and partial explanations in medical image analysis by proposing a multi-layer attention mechanism that enforces consistent interpretations between layers using convex optimization, resulting in complete and faithful explanations while preserving predictive performance with weakly annotated data.
A key concern in integrating machine learning models in medicine is the ability to interpret their reasoning. Popular explainability methods have demonstrated satisfactory results in natural image recognition, yet in medical image analysis, many of these approaches provide partial and noisy explanations. Recently, attention mechanisms have shown compelling results both in their predictive performance and in their interpretable qualities. A fundamental trait of attention is that it leverages salient parts of the input which contribute to the model's prediction. To this end, our work focuses on the explanatory value of attention weight distributions. We propose a multi-layer attention mechanism that enforces consistent interpretations between attended convolutional layers using convex optimization. We apply duality to decompose the consistency constraints between the layers by reparameterizing their attention probability distributions. We further suggest learning the dual witness by optimizing with respect to our objective; thus, our implementation uses standard back-propagation, hence it is highly efficient. While preserving predictive performance, our proposed method leverages weakly annotated medical imaging data and provides complete and faithful explanations to the model's prediction.