Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ProtoAL: Interpretable Deep Active Learning with prototypes for medical imaging

Iury B. de A. Santos, André C. P. L. F. de Carvalho

TL;DR

The paper addresses the dual challenges of interpretability and data scarcity in AI-CAD for medical imaging. It introduces ProtoAL, a method that embeds an interpretable ProtoPNet into a Deep Active Learning framework, using MC Dropout for uncertainty estimation and prototype-based explanations aligned with clinical reasoning. On the Messidor dataset, ProtoAL achieves an AUPRC of 0.79 while labeling only about 76.5% of the data, offering a practical balance between performance and labeling cost. The work demonstrates that interpretable, prototype-guided learning can be effectively integrated into active learning to enhance trust and data efficiency in medical image analysis.

Abstract

The adoption of Deep Learning algorithms in the medical imaging field is a prominent area of research, with high potential for advancing AI-based Computer-aided diagnosis (AI-CAD) solutions. However, current solutions face challenges due to a lack of interpretability features and high data demands, prompting recent efforts to address these issues. In this study, we propose the ProtoAL method, where we integrate an interpretable DL model into the Deep Active Learning (DAL) framework. This approach aims to address both challenges by focusing on the medical imaging context and utilizing an inherently interpretable model based on prototypes. We evaluated ProtoAL on the Messidor dataset, achieving an area under the precision-recall curve of 0.79 while utilizing only 76.54\% of the available labeled data. These capabilities can enhances the practical usability of a DL model in the medical field, providing a means of trust calibration in domain experts and a suitable solution for learning in the data scarcity context often found.

ProtoAL: Interpretable Deep Active Learning with prototypes for medical imaging

TL;DR

The paper addresses the dual challenges of interpretability and data scarcity in AI-CAD for medical imaging. It introduces ProtoAL, a method that embeds an interpretable ProtoPNet into a Deep Active Learning framework, using MC Dropout for uncertainty estimation and prototype-based explanations aligned with clinical reasoning. On the Messidor dataset, ProtoAL achieves an AUPRC of 0.79 while labeling only about 76.5% of the data, offering a practical balance between performance and labeling cost. The work demonstrates that interpretable, prototype-guided learning can be effectively integrated into active learning to enhance trust and data efficiency in medical image analysis.

Abstract

The adoption of Deep Learning algorithms in the medical imaging field is a prominent area of research, with high potential for advancing AI-based Computer-aided diagnosis (AI-CAD) solutions. However, current solutions face challenges due to a lack of interpretability features and high data demands, prompting recent efforts to address these issues. In this study, we propose the ProtoAL method, where we integrate an interpretable DL model into the Deep Active Learning (DAL) framework. This approach aims to address both challenges by focusing on the medical imaging context and utilizing an inherently interpretable model based on prototypes. We evaluated ProtoAL on the Messidor dataset, achieving an area under the precision-recall curve of 0.79 while utilizing only 76.54\% of the available labeled data. These capabilities can enhances the practical usability of a DL model in the medical field, providing a means of trust calibration in domain experts and a suitable solution for learning in the data scarcity context often found.
Paper Structure (16 sections, 2 figures, 2 tables)

This paper contains 16 sections, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Deep Active Learning
  4. Uncertainty estimation with MC Dropout
  5. Prototypical Neural Network
  6. Experiments
  7. Dataset
  8. Baselines
  9. Vanilla ResNet-18 model.
  10. ProtoPNet standalone baseline.
  11. ProtoAL with random search query.
  12. Implementation details
  13. DAL framework implementation.
  14. ProtoPNet.
  15. Results
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Schematic view illustrating the DAL and model training cycles. In the DAL cycle, labeled instances are added to $\mathcal{L}$ by selecting unlabeled instances from $\mathcal{U}$ using a search strategy. Meanwhile, the learning model M undergoes training iterations within each DAL cycle.
  • Figure 2: Comparisons of ProtoAL MC method and the baselines, evaluated on the validation set