A Unified Deep Learning Framework For Simultaneous Segmentation And Multiclass Classification Of Chronic Wounds *

Author

Mustafa Alhababi, Wayne State University

Access Type

WSU Access

Date of Award

January 2025

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Electrical and Computer Engineering

First Advisor

Gregory Auner

Second Advisor

Hafiz Malik

Abstract

Chronic wounds affect millions worldwide, posing significant challenges for healthcare systems and a heavy economic burden globally. The segmentation and classification (S&C) of chronic wounds are critical for wound care management and diagnosis, aiding clinicians in selecting appropriate treatments. Existing approaches have utilized either traditional machine learning or deep learning methods for S&C. However, most focus on binary classification, with few addressing multi-class classification, often showing degraded performance for pressure and diabetic wounds. Wound segmentation has been largely limited to foot ulcer images, and there is no unified diagnostic tool for both S&C tasks. To address these gaps, we developed a unified approach that performs S&C simultaneously. For segmentation, we proposed Attention-Dense-UNet (Att-d-UNet), and for classification, we introduced a feature concatenation-based method. Our framework segments wound images using Att-d-UNet, followed by classification into one of the wound types using our proposed method. We evaluated our models on publicly available wound classification datasets (AZH and Medetec) and segmentation datasets (FUSeg and AZH). To test our unified approach, we extended wound classification datasets by generating segmentation masks for Medetec and AZH images. The proposed unified approach achieved 90% accuracy and an 86.55% dice score on the Medetec dataset and 81% accuracy and an 86.53% dice score on the AZH dataset These results demonstrate the effectiveness of our separate models and unified approach for wound S&C.

Wound segmentation aids in the measurement of wound area which further assists in analyzing the wound healing progress. In this research work, we have also presented a deep learning-based segmentation approach namely Dual-UNet for precisely segmenting the diabetic foot ulcers images. The foot ulcer images are passed to Dual-UNet to generate the mask image having the segmented wound area. In our proposed approach, two UNets are utilized each having encoder, atrous spatial pyramid pooling (ASPP) block, decoder with skip connections, and output block to improve the segmentation results. We assessed our framework by performing the experimentation on two benchmark datasets named foot ulcer segmentation (FUSeg) challenge and AZH segmentation dataset. Additionally, we have also evaluated our Dual-UNet for cross-dataset validation to demonstrate the generalization capability of the proposed approach. Both the quantitative and visual results demonstrate the effectiveness of the proposed framework for the segmentation of chronic diabetic wounds.

We also propose a novel lightweight fused-densenet method capable of reliable classification of multiple types of chronic wounds. Our method comprises a fully trained and a partially trained densenet model, which is fused to develop an effective multiclass wound classification approach. We introduce the GeLU activation function to tackle the dying ReLU problem, enhanced performance, better learning, and efficient training. Further, we add the dense and dropout layers along with the L2 regularization approach to counter the model overfitting. We assessed the performance of our lightweight model on the standard Medetec and AZH datasets, as well as their augmented versions. We employed multiple augmentation techniques to increase the number of samples and diversity of these datasets to tackle the overfitting and class imbalance issues. Experimental evaluation on AZH, Medetec, and augmented versions of both datasets signifies the efficacy of our proposed method for multiclass wound classification.

Recommended Citation

Alhababi, Mustafa, "A Unified Deep Learning Framework For Simultaneous Segmentation And Multiclass Classification Of Chronic Wounds *" (2025). Wayne State University Dissertations. 4174.
https://digitalcommons.wayne.edu/oa_dissertations/4174