Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Interpretable Machine Learning for Oral Lesion Diagnosis through Prototypical Instances Identification

Alessio Cascione, Mattia Setzu, Federico A. Galatolo, Mario G. C. A. Cimino, Riccardo Guidotti

TL;DR

This work tackles the need for interpretable AI in clinical decision support by introducing PivotTree, a data-agnostic, hierarchical prototype-based model that selects a small set of pivots and grounds predictions in their similarities. By mapping data to a similarity space and using an interpretable predictor (Decision Tree or kNN) on pivot similarities, PivotTree provides transparent, rule-based reasoning even for image data via embeddings. The method is evaluated on the DoctOral-AI oral lesion dataset, showing that PivotTree with a modest number of pivots achieves competitive performance while delivering interpretable explanations and aligning partially with expert prototypes. The study highlights PivotTree’s potential to extend interpretability across medical modalities and suggests future work on alternative splits, other data types, and human-subject interpretability assessments.

Abstract

Decision-making processes in healthcare can be highly complex and challenging. Machine Learning tools offer significant potential to assist in these processes. However, many current methodologies rely on complex models that are not easily interpretable by experts. This underscores the need to develop interpretable models that can provide meaningful support in clinical decision-making. When approaching such tasks, humans typically compare the situation at hand to a few key examples and representative cases imprinted in their memory. Using an approach which selects such exemplary cases and grounds its predictions on them could contribute to obtaining high-performing interpretable solutions to such problems. To this end, we evaluate PivotTree, an interpretable prototype selection model, on an oral lesion detection problem, specifically trying to detect the presence of neoplastic, aphthous and traumatic ulcerated lesions from oral cavity images. We demonstrate the efficacy of using such method in terms of performance and offer a qualitative and quantitative comparison between exemplary cases and ground-truth prototypes selected by experts.

Interpretable Machine Learning for Oral Lesion Diagnosis through Prototypical Instances Identification

TL;DR

This work tackles the need for interpretable AI in clinical decision support by introducing PivotTree, a data-agnostic, hierarchical prototype-based model that selects a small set of pivots and grounds predictions in their similarities. By mapping data to a similarity space and using an interpretable predictor (Decision Tree or kNN) on pivot similarities, PivotTree provides transparent, rule-based reasoning even for image data via embeddings. The method is evaluated on the DoctOral-AI oral lesion dataset, showing that PivotTree with a modest number of pivots achieves competitive performance while delivering interpretable explanations and aligning partially with expert prototypes. The study highlights PivotTree’s potential to extend interpretability across medical modalities and suggests future work on alternative splits, other data types, and human-subject interpretability assessments.

Abstract

Decision-making processes in healthcare can be highly complex and challenging. Machine Learning tools offer significant potential to assist in these processes. However, many current methodologies rely on complex models that are not easily interpretable by experts. This underscores the need to develop interpretable models that can provide meaningful support in clinical decision-making. When approaching such tasks, humans typically compare the situation at hand to a few key examples and representative cases imprinted in their memory. Using an approach which selects such exemplary cases and grounds its predictions on them could contribute to obtaining high-performing interpretable solutions to such problems. To this end, we evaluate PivotTree, an interpretable prototype selection model, on an oral lesion detection problem, specifically trying to detect the presence of neoplastic, aphthous and traumatic ulcerated lesions from oral cavity images. We demonstrate the efficacy of using such method in terms of performance and offer a qualitative and quantitative comparison between exemplary cases and ground-truth prototypes selected by experts.

Paper Structure

This paper contains 7 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Pivot Tree in a Nutshell
  4. Experiments
  5. Conclusion
  6. Acknowledgments.
  7. Disclosure of Interests.

Figures (4)

  • Figure 1: PivotTree as (a) selector, (b) interpretable model, (c) Decision Tree.
  • Figure 2: Partial visual depiction of best PTC configuration on the test set. Branches are labeled with similarity threshold values used for prediction.
  • Figure 3: PivotTree pivots (rows) and ground-truth prototypes (columns) comparison as Euclidean distances on D2 embedding. The darker the color the more similar are a pivot and a ground truth prototype. The first letter identifies the class of the instances: neoplastic, aphthous, and traumatic.
  • Figure 4: PivotTree pivots (rows) and ground-truth prototypes (columns) comparison as SSIM on raw regions of interest. Same rules from Fig. \ref{['fig:heatmap_euclid']} apply.