Automatic Foot Ulcer Segmentation Using an Ensemble of Convolutional Neural Networks

Abstract

Foot ulcer is a common complication of diabetes mellitus; it is associated with substantial morbidity and mortality and remains a major risk factor for lower leg amputation. Extracting accurate morphological features from the foot wounds is crucial for proper treatment. Although visual and manual inspection by medical professionals is the common approach to extract the features, this method is subjective and errorprone. Computer-mediated approaches are the alternative solutions to segment the lesions and extract related morphological features. Among various proposed computer-based approaches for image segmentation, deep learning-based methods and more specifically convolutional neural networks (CNN) have shown excellent performances for various image segmentation tasks including medical image segmentation. In this work, we proposed an ensemble approach based on two encoder-decoder-based CNN models, namely LinkNet and UNet, to perform foot ulcer segmentation. To deal with limited training samples, we used pre-trained weights (Ef-ficientNetB1 for the LinkNet model and EfficientNetB2 for the UNet model) and further pre-training by the Medetec dataset. We also applied a number of morphological-based and colour-based augmentation techniques to train the models. We integrated five-fold cross-validation, test time augmentation and result fusion in our proposed ensemble approach to boost the segmentation performance. Applied on a publicly available foot ulcer segmentation dataset and the MICCAI 2021 Foot Ulcer Segmentation (FUSeg) Challenge, our method achieved state-of-the-art data-based Dice scores of 92.07% and 88.80%, respectively. Our developed method achieved the first rank in the FUSeg challenge leaderboard. The Dockerised guideline, inference codes and saved trained models are publicly available in the published GitHub repository:https://github.com/masih4/Foot_Ulcer_ Segmentation.

A Survey of Wound Image Analysis Using Deep Learning: Classification, Detection, and Segmentation

“…Medetec dataset [24] Mixed -No [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35] BIP_US database [36] Burn 94 Yes [37], [38] FUSeg dataset [28] Diabetic foot ulcer 1210 Yes [29] Sårwebben [39] Mixed -No [40] Chronic wound database [41] Chronic wound 188 Yes [42] AHZ dataset [28] Diabetic foot ulcer 1109 Yes [28] AHZ&UWM dataset [34] Mixed 538 Yes [34] publicly available [44]. Second, the primary job of medical professionals is not data collection, and the acquisition of a batch of images may be done by multiple personnel, which can lead to inconsistent standards of the collected images.…”

2D/3D Wound Segmentation and Measurement Based on a Robot-Driven Reconstruction System

“…Researchers have employed a variety of approaches to perform 2D wound segmentation, including using K-means clustering [ 6 , 7 ], deep neural networks [ 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ], support vector machines [ 16 , 17 ], k-nearest neighbors [ 4 ], and simple feedforward networks [ 18 ]. Other approaches include using superpixel region-growing algorithms, color histograms, or combined geometric and visual information of the wound surface to segment wounds.…”

“…A segmentation technique made up of the U-Net and LinkNet deep neural networks was proposed by Mahbod et al in [ 13 ]. These deep neural networks are basically encoder–decoder convolutional networks.…”

Wound Detection by Simple Feedforward Neural Network

“…Mahbod et al [17] proposed a segmentation method that consists of two encoderdecoder convolutional neural networks, U-Net and LinkNet. Both networks were pretrained on the Medetec database [18] on 152 images and afterwards, the model was trained on the chronic wound dataset shared by the MICCAI 2021 Foot Ulcer Segmentation Challenge containing 1010 annotated images.…”