Chronic Wound Image Augmentation and Assessment Using Semi-Supervised Progressive Multi-Granularity EfficientNet

Chronic Wound Image Augmentation and Assessment Using Semi-Supervised Progressive Multi-Granularity EfficientNet

<italic>Goal:</italic> Augment a small, imbalanced, wound dataset by using semi-supervised learning with a secondary dataset. Then utilize the augmented wound dataset for deep learning-based wound assessment. <italic>Methods:</italic> The clinically-validated Photographic Wou...

Full description

Saved in:
Bibliographic Details
Main Authors: Ziyang Liu, Emmanuel Agu, Peder Pedersen, Clifford Lindsay, Bengisu Tulu, Diane Strong
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Open Journal of Engineering in Medicine and Biology
Subjects:
Chronic wounds
data imbalance
data augmentation
neural networks
smartphone assessment
Online Access:https://ieeexplore.ieee.org/document/10050724/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:<italic>Goal:</italic> Augment a small, imbalanced, wound dataset by using semi-supervised learning with a secondary dataset. Then utilize the augmented wound dataset for deep learning-based wound assessment. <italic>Methods:</italic> The clinically-validated Photographic Wound Assessment Tool (PWAT) scores eight wound attributes: Size, Depth, Necrotic Tissue Type, Necrotic Tissue Amount, Granulation Tissue type, Granulation Tissue Amount, Edges, Periulcer Skin Viability to comprehensively assess chronic wound images. A small corpus of 1639 wound images labeled with ground truth PWAT scores was used as reference. A Semi-Supervised learning and Progressive Multi-Granularity training mechanism were used to leverage a secondary corpus of 9870 unlabeled wound images. Wound scoring utilized the EfficientNet Convolutional Neural Network on the augmented wound corpus. <italic>Results:</italic> Our proposed Semi-Supervised PMG EfficientNet (SS-PMG-EfficientNet) approach estimated all 8 PWAT sub-scores with classification accuracies and F1 scores of about 90&#x0025; on average, and outperformed a comprehensive list of baseline models and had a 7&#x0025; improvement over the prior state-of-the-art (without data augmentation). We also demonstrate that synthetic wound image generation using Generative Adversarial Networks (GANs) did not improve wound assessment. <italic>Conclusions:</italic> Semi-supervised learning on unlabeled wound images in a secondary dataset achieved impressive performance for deep learning-based wound grading.
ISSN:2644-1276

Similar Items