Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Interpretable Medical Image Classification using Prototype Learning and Privileged Information

Luisa Gallee, Meinrad Beer, Michael Goetz

TL;DR

This work investigates whether additional information available during the training process can be used to create an understandable and powerful model and proposes an innovative solution called Proto-Caps that leverages the benefits of capsule networks, prototype learning and the use of privileged information.

Abstract

Interpretability is often an essential requirement in medical imaging. Advanced deep learning methods are required to address this need for explainability and high performance. In this work, we investigate whether additional information available during the training process can be used to create an understandable and powerful model. We propose an innovative solution called Proto-Caps that leverages the benefits of capsule networks, prototype learning and the use of privileged information. Evaluating the proposed solution on the LIDC-IDRI dataset shows that it combines increased interpretability with above state-of-the-art prediction performance. Compared to the explainable baseline model, our method achieves more than 6 % higher accuracy in predicting both malignancy (93.0 %) and mean characteristic features of lung nodules. Simultaneously, the model provides case-based reasoning with prototype representations that allow visual validation of radiologist-defined attributes.

Interpretable Medical Image Classification using Prototype Learning and Privileged Information

TL;DR

This work investigates whether additional information available during the training process can be used to create an understandable and powerful model and proposes an innovative solution called Proto-Caps that leverages the benefits of capsule networks, prototype learning and the use of privileged information.

Abstract

Interpretability is often an essential requirement in medical imaging. Advanced deep learning methods are required to address this need for explainability and high performance. In this work, we investigate whether additional information available during the training process can be used to create an understandable and powerful model. We propose an innovative solution called Proto-Caps that leverages the benefits of capsule networks, prototype learning and the use of privileged information. Evaluating the proposed solution on the LIDC-IDRI dataset shows that it combines increased interpretability with above state-of-the-art prediction performance. Compared to the explainable baseline model, our method achieves more than 6 % higher accuracy in predicting both malignancy (93.0 %) and mean characteristic features of lung nodules. Simultaneously, the model provides case-based reasoning with prototype representations that allow visual validation of radiologist-defined attributes.
Paper Structure (5 sections, 4 equations, 2 figures, 2 tables)

This paper contains 5 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Experiments
  4. Results
  5. Discussion and Conclusion

Figures (2)

  • Figure 1: Proposed Model Architecture The backbone capsule network results in capsules representing predefined attributes. For each capsule, a set of prototypes is trained. To fit the attribute scores, the capsule vectors are fed through individual dense layers. The latent vectors of all capsules are being accumulated for a dense layer to predict a target score and for a decoder network to reconstruct the region of interest.
  • Figure 2: One correct and two wrongly predicted examples with exemplary attribute prototypes. Prediction $\hat{\text{y}}$ and ground truth label y of malignancy and attribute respectively. Identifying false attribute predictions can help to identify misclassification in malignancy.