Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CountXplain: Interpretable Cell Counting with Prototype-Based Density Map Estimation

Abdurahman Ali Mohammed, Wallapak Tavanapong, Catherine Fonder, Donald S. Sakaguchi

TL;DR

CountXplain addresses the interpretability gap in cell counting by integrating a prototype layer into density map estimation. It learns representative cell and background prototypes and generates explanations by highlighting regions similar to each prototype. The model optimizes a multi-objective loss including a prototype-to-feature term and a diversity term to maintain distinct prototypes while preserving counting accuracy. Empirical results on IDCIA and DCC show comparable MAE to CSRNet and provide concrete, visual explanations, validated by a biologist survey, suggesting practical utility in biomedical settings.

Abstract

Cell counting in biomedical imaging is pivotal for various clinical applications, yet the interpretability of deep learning models in this domain remains a significant challenge. We propose a novel prototype-based method for interpretable cell counting via density map estimation. Our approach integrates a prototype layer into the density estimation network, enabling the model to learn representative visual patterns for both cells and background artifacts. The learned prototypes were evaluated through a survey of biologists, who confirmed the relevance of the visual patterns identified, further validating the interpretability of the model. By generating interpretations that highlight regions in the input image most similar to each prototype, our method offers a clear understanding of how the model identifies and counts cells. Extensive experiments on two public datasets demonstrate that our method achieves interpretability without compromising counting effectiveness. This work provides researchers and clinicians with a transparent and reliable tool for cell counting, potentially increasing trust and accelerating the adoption of deep learning in critical biomedical applications. Code is available at https://github.com/NRT-D4/CountXplain.

CountXplain: Interpretable Cell Counting with Prototype-Based Density Map Estimation

TL;DR

CountXplain addresses the interpretability gap in cell counting by integrating a prototype layer into density map estimation. It learns representative cell and background prototypes and generates explanations by highlighting regions similar to each prototype. The model optimizes a multi-objective loss including a prototype-to-feature term and a diversity term to maintain distinct prototypes while preserving counting accuracy. Empirical results on IDCIA and DCC show comparable MAE to CSRNet and provide concrete, visual explanations, validated by a biologist survey, suggesting practical utility in biomedical settings.

Abstract

Cell counting in biomedical imaging is pivotal for various clinical applications, yet the interpretability of deep learning models in this domain remains a significant challenge. We propose a novel prototype-based method for interpretable cell counting via density map estimation. Our approach integrates a prototype layer into the density estimation network, enabling the model to learn representative visual patterns for both cells and background artifacts. The learned prototypes were evaluated through a survey of biologists, who confirmed the relevance of the visual patterns identified, further validating the interpretability of the model. By generating interpretations that highlight regions in the input image most similar to each prototype, our method offers a clear understanding of how the model identifies and counts cells. Extensive experiments on two public datasets demonstrate that our method achieves interpretability without compromising counting effectiveness. This work provides researchers and clinicians with a transparent and reliable tool for cell counting, potentially increasing trust and accelerating the adoption of deep learning in critical biomedical applications. Code is available at https://github.com/NRT-D4/CountXplain.

Paper Structure

This paper contains 11 sections, 10 equations, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Proposed Interpretable Prototype-based DME
  3. Model Architecture
  4. Training Procedure and Loss Function
  5. Loss Function
  6. Experiments and Results
  7. Results
  8. Conclusion and Future Work
  9. Additional Details
  10. Additional ablation studies
  11. Expert Evaluations

Figures (10)

  • Figure 1: Architecture of CountXplain. See Section \ref{['arch']} for details.
  • Figure 2: Examples of predicted density maps by CountXplain
  • Figure 3: Global knowledge of patterns recognized by individual CountXplain models
  • Figure 4: Analysis on the impact of the value of $K$ on MAE. Lower MAEs are prefered.
  • Figure 5: Model's global knowledge when the prototype to feature loss is not used.
  • ...and 5 more figures