Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-modal wound classification using wound image and location by Xception and Gaussian Mixture Recurrent Neural Network (GMRNN)

Ramin Mousa, Ehsan Matbooe, Hakimeh Khojasteh, Amirali Bengari, Mohammadmahdi Vahediahmar

TL;DR

This work tackles multi-class wound classification by fusing wound images with anatomical location data. It introduces a multi-modal network that uses Xception for image feature extraction, a Capsule network with self-attention for robust representation, and a Gaussian Mixture Recurrent Neural Network (GMRNN) to encode location information, with the outputs merged into a final softmax classifier over $N$ classes. The approach achieves state-of-the-art-like performance on the AZH dataset across four-, five-, six-class problems and ten binary tasks, demonstrating the value of incorporating location data alongside images. The study shows that this integration, along with transfer learning and data augmentation, yields strong generalization and practical potential for rapid, accurate wound diagnosis in clinical settings, with future work pointing to GAN-based augmentation and cost-sensitive training.

Abstract

The effective diagnosis of acute and hard-to-heal wounds is crucial for wound care practitioners to provide effective patient care. Poor clinical outcomes are often linked to infection, peripheral vascular disease, and increasing wound depth, which collectively exacerbate these comorbidities. However, diagnostic tools based on Artificial Intelligence (AI) speed up the interpretation of medical images and improve early detection of disease. In this article, we propose a multi-modal AI model based on transfer learning (TL), which combines two state-of-the-art architectures, Xception and GMRNN, for wound classification. The multi-modal network is developed by concatenating the features extracted by a transfer learning algorithm and location features to classify the wound types of diabetic, pressure, surgical, and venous ulcers. The proposed method is comprehensively compared with deep neural networks (DNN) for medical image analysis. The experimental results demonstrate a notable wound-class classifications (containing only diabetic, pressure, surgical, and venous) vary from 78.77 to 100\% in various experiments. The results presented in this study showcase the exceptional accuracy of the proposed methodology in accurately classifying the most commonly occurring wound types using wound images and their corresponding locations.

Multi-modal wound classification using wound image and location by Xception and Gaussian Mixture Recurrent Neural Network (GMRNN)

TL;DR

This work tackles multi-class wound classification by fusing wound images with anatomical location data. It introduces a multi-modal network that uses Xception for image feature extraction, a Capsule network with self-attention for robust representation, and a Gaussian Mixture Recurrent Neural Network (GMRNN) to encode location information, with the outputs merged into a final softmax classifier over classes. The approach achieves state-of-the-art-like performance on the AZH dataset across four-, five-, six-class problems and ten binary tasks, demonstrating the value of incorporating location data alongside images. The study shows that this integration, along with transfer learning and data augmentation, yields strong generalization and practical potential for rapid, accurate wound diagnosis in clinical settings, with future work pointing to GAN-based augmentation and cost-sensitive training.

Abstract

The effective diagnosis of acute and hard-to-heal wounds is crucial for wound care practitioners to provide effective patient care. Poor clinical outcomes are often linked to infection, peripheral vascular disease, and increasing wound depth, which collectively exacerbate these comorbidities. However, diagnostic tools based on Artificial Intelligence (AI) speed up the interpretation of medical images and improve early detection of disease. In this article, we propose a multi-modal AI model based on transfer learning (TL), which combines two state-of-the-art architectures, Xception and GMRNN, for wound classification. The multi-modal network is developed by concatenating the features extracted by a transfer learning algorithm and location features to classify the wound types of diabetic, pressure, surgical, and venous ulcers. The proposed method is comprehensively compared with deep neural networks (DNN) for medical image analysis. The experimental results demonstrate a notable wound-class classifications (containing only diabetic, pressure, surgical, and venous) vary from 78.77 to 100\% in various experiments. The results presented in this study showcase the exceptional accuracy of the proposed methodology in accurately classifying the most commonly occurring wound types using wound images and their corresponding locations.
Paper Structure (10 sections, 34 equations, 17 figures, 7 tables)

This paper contains 10 sections, 34 equations, 17 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Gaussian Mixture Recurrent Neural Network (GMRNN) Cell
  4. Evaluation tools
  5. Dataset
  6. Deep learning library
  7. Evaluation Metrics
  8. Result
  9. Sensitivity analysis
  10. Conclusion

Figures (17)

  • Figure 1: An overview of proposed model.
  • Figure 2: Body map for location selection(image tacked from BASE).
  • Figure 3: Dataset processing steps(image tacked from BASE).
  • Figure 4: Some examples of data sets.
  • Figure 5: Bar plot for four wound class classification (D vs. P vs. S vs. V) on AZH dataset with original body map.
  • ...and 12 more figures