Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Interpretable Automatic Rosacea Detection with Whitened Cosine Similarity

Chengyu Yang, Chengjun Liu

TL;DR

Problem addressed: rosacea detection from facial images in low-data settings requiring interpretability. Approach: an interpretable detector based on whitened cosine similarity delta_WC using a whitening transform derived from the data covariance; decisions compare delta_WC to rosacea vs normal class means. Contributions: automatic rosacea detector with high unseen-data accuracy and recall, interpretability via class-mean similarity, and privacy-aware GAN-based training data; code available on GitHub. Findings: on a constrained dataset with real test images, the method achieves 100% validation accuracy and 99% test accuracy, with recall around 0.96, surpassing a ResNet-18 baseline and PCA baselines and providing a transparent diagnostic rationale.

Abstract

According to the National Rosacea Society, approximately sixteen million Americans suffer from rosacea, a common skin condition that causes flushing or long-term redness on a person's face. To increase rosacea awareness and to better assist physicians to make diagnosis on this disease, we propose an interpretable automatic rosacea detection method based on whitened cosine similarity in this paper. The contributions of the proposed methods are three-fold. First, the proposed method can automatically distinguish patients suffering from rosacea from people who are clean of this disease with a significantly higher accuracy than other methods in unseen test data, including both classical deep learning and statistical methods. Second, the proposed method addresses the interpretability issue by measuring the similarity between the test sample and the means of two classes, namely the rosacea class versus the normal class, which allows both medical professionals and patients to understand and trust the results. And finally, the proposed methods will not only help increase awareness of rosacea in the general population, but will also help remind patients who suffer from this disease of possible early treatment, as rosacea is more treatable in its early stages. The code and data are available at https://github.com/chengyuyang-njit/ICCRD-2025. The code and data are available at https://github.com/chengyuyang-njit/ICCRD-2025.

Interpretable Automatic Rosacea Detection with Whitened Cosine Similarity

TL;DR

Problem addressed: rosacea detection from facial images in low-data settings requiring interpretability. Approach: an interpretable detector based on whitened cosine similarity delta_WC using a whitening transform derived from the data covariance; decisions compare delta_WC to rosacea vs normal class means. Contributions: automatic rosacea detector with high unseen-data accuracy and recall, interpretability via class-mean similarity, and privacy-aware GAN-based training data; code available on GitHub. Findings: on a constrained dataset with real test images, the method achieves 100% validation accuracy and 99% test accuracy, with recall around 0.96, surpassing a ResNet-18 baseline and PCA baselines and providing a transparent diagnostic rationale.

Abstract

According to the National Rosacea Society, approximately sixteen million Americans suffer from rosacea, a common skin condition that causes flushing or long-term redness on a person's face. To increase rosacea awareness and to better assist physicians to make diagnosis on this disease, we propose an interpretable automatic rosacea detection method based on whitened cosine similarity in this paper. The contributions of the proposed methods are three-fold. First, the proposed method can automatically distinguish patients suffering from rosacea from people who are clean of this disease with a significantly higher accuracy than other methods in unseen test data, including both classical deep learning and statistical methods. Second, the proposed method addresses the interpretability issue by measuring the similarity between the test sample and the means of two classes, namely the rosacea class versus the normal class, which allows both medical professionals and patients to understand and trust the results. And finally, the proposed methods will not only help increase awareness of rosacea in the general population, but will also help remind patients who suffer from this disease of possible early treatment, as rosacea is more treatable in its early stages. The code and data are available at https://github.com/chengyuyang-njit/ICCRD-2025. The code and data are available at https://github.com/chengyuyang-njit/ICCRD-2025.

Paper Structure

This paper contains 12 sections, 12 equations, 2 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Experiment
  4. Dataset
  5. Training and Validation Dataset
  6. Test Dataset
  7. Technical Details Disclosure
  8. Performance Evaluation and Comparison
  9. Evaluation Metrics
  10. Performance Analysis
  11. Insights
  12. Conclusion

Figures (2)

  • Figure 1: The first column shows the mean images of the two training datasets from the rosacea negative and positive classes, respectively. The remaining six columns display six pair of random example images from the two training datasets corresponding to the normal people and those with rosacea, respectively.
  • Figure 2: The two test cases that our method misclassified.