Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Wavelet Guided Attention Module for Skin Cancer Classification with Gradient-based Feature Fusion

Ayush Roy, Sujan Sarkar, Sohom Ghosal, Dmitrii Kaplun, Asya Lyanova, Ram Sarkar

TL;DR

A novel model is proposed, which uses a novel attention mechanism to pinpoint the differences in features across the spatial dimensions and symmetry of the lesion, thereby focusing on the dissimilarities of various classes based on symmetry, uniformity in texture and color, etc.

Abstract

Skin cancer is a highly dangerous type of cancer that requires an accurate diagnosis from experienced physicians. To help physicians diagnose skin cancer more efficiently, a computer-aided diagnosis (CAD) system can be very helpful. In this paper, we propose a novel model, which uses a novel attention mechanism to pinpoint the differences in features across the spatial dimensions and symmetry of the lesion, thereby focusing on the dissimilarities of various classes based on symmetry, uniformity in texture and color, etc. Additionally, to take into account the variations in the boundaries of the lesions for different classes, we employ a gradient-based fusion of wavelet and soft attention-aided features to extract boundary information of skin lesions. We have tested our model on the multi-class and highly class-imbalanced dataset, called HAM10000, and achieved promising results, with a 91.17\% F1-score and 90.75\% accuracy. The code is made available at: https://github.com/AyushRoy2001/WAGF-Fusion.

A Wavelet Guided Attention Module for Skin Cancer Classification with Gradient-based Feature Fusion

TL;DR

A novel model is proposed, which uses a novel attention mechanism to pinpoint the differences in features across the spatial dimensions and symmetry of the lesion, thereby focusing on the dissimilarities of various classes based on symmetry, uniformity in texture and color, etc.

Abstract

Skin cancer is a highly dangerous type of cancer that requires an accurate diagnosis from experienced physicians. To help physicians diagnose skin cancer more efficiently, a computer-aided diagnosis (CAD) system can be very helpful. In this paper, we propose a novel model, which uses a novel attention mechanism to pinpoint the differences in features across the spatial dimensions and symmetry of the lesion, thereby focusing on the dissimilarities of various classes based on symmetry, uniformity in texture and color, etc. Additionally, to take into account the variations in the boundaries of the lesions for different classes, we employ a gradient-based fusion of wavelet and soft attention-aided features to extract boundary information of skin lesions. We have tested our model on the multi-class and highly class-imbalanced dataset, called HAM10000, and achieved promising results, with a 91.17\% F1-score and 90.75\% accuracy. The code is made available at: https://github.com/AyushRoy2001/WAGF-Fusion.
Paper Structure (9 sections, 2 equations, 4 figures, 2 tables)

This paper contains 9 sections, 2 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Gradient-based Feature Fusion
  4. Symmetry-aware Feature Attention Module
  5. Results
  6. Experimental Setup
  7. Ablation study
  8. Comparison with SOTA
  9. Conclusion

Figures (4)

  • Figure 1: The proposed classification model utilizes the DenseNet-121 as the backbone for feature extraction. The extracted feature $F_{enc}$ is treated with Soft Attention and the SaFA module. A gradient-based fusion of the wavelet and attention-aided features reinforces the boundary information of the lesions.
  • Figure 2: Symmetry-aware Feature Attention (SaFA) module.
  • Figure 3: Heatmap of the SaFA module for all seven classes.
  • Figure 4: Confusion matrix and feature representation of the proposed model.