Semantic Segmentation of Diabetic Foot Ulcer Images: Dealing with Small Dataset in DL Approaches

Niri, Rania; Douzi, Hassan; Lucas, Yves; Treuillet, Sylvie

doi:10.1007/978-3-030-51935-3_17

Cited by 31 publications

(9 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The distance and angles calculated in this step are used to automatically select the furthest frontal frame, i.e., the reference thermal frame. The automated 2D wound segmentation process was performed based on a deep learning model developed in a previous study by the same research group [24]. The reference surface thermography provides an overview of the temperature pattern of the wound and its surroundings.…”

Section: Reference Surface Thermographymentioning

confidence: 99%

See 1 more Smart Citation

Multimodal and Multiview Wound Monitoring with Mobile Devices

et al. 2021

View full text Add to dashboard Cite

Along with geometric and color indicators, thermography is another valuable source of information for wound monitoring. The interaction of geometry with thermography can provide predictive indicators of wound evolution; however, existing processes are focused on the use of high-cost devices with a static configuration, which restricts the scanning of large surfaces. In this study, we propose the use of commercial devices, such as mobile devices and portable thermography, to integrate information from different wavelengths onto the surface of a 3D model. A handheld acquisition is proposed in which color images are used to create a 3D model by using Structure from Motion (SfM), and thermography is incorporated into the 3D surface through a pose estimation refinement based on optimizing the temperature correlation between multiple views. Thermal and color 3D models were successfully created for six patients with multiple views from a low-cost commercial device. The results show the successful application of the proposed methodology where thermal mapping on 3D models is not limited in the scanning area and can provide consistent information between multiple thermal camera views. Further work will focus on studying the quantitative metrics obtained by the multi-view 3D models created with the proposed methodology.

show abstract

Section: Reference Surface Thermographymentioning

confidence: 99%

“…The wound segmentation in the 3D model was created from segmentations of the 2D color images. The 2D segmentation was performed using a deep learning model previously developed by the same research group [24]. All 2D segmentations were reprojected to the 3D point cloud in order to create the wound segmentation in 3D.…”

Section: D Wound and Periwound Areamentioning

confidence: 99%

Multimodal and Multiview Wound Monitoring with Mobile Devices

et al. 2021

View full text Add to dashboard Cite

show abstract

“…To perform wound delineation, we based our segmentation on our previous work [36] while proposing more robust background elimination using skin correction. We opted for the state-of-the-art semantic segmentation network U-net for medical images [37].…”

Section: A Robust Wound Segmentationmentioning

confidence: 99%

“…Both U-net networks for wound and skin segmentation tasks were trained for 200 epochs and were implemented in Keras with TensorFlow backend using the Adam optimization algorithm [39] with a learning rate of 0.001. The performance of the proposed segmentation procedure helped to improve the accuracy to reach a Jaccard index of 98.48% and a Dice score of 99.26% instead of 94.96% and 97.25% in the previous version [36] using the same testing set. The proposed 2D segmentation method overcomes perfectly complex background elimination and uncontrolled lighting conditions, but it still can be greatly affected by high angle and distance variation due to camera position and orientation in the image sequence.…”

Section: A Robust Wound Segmentationmentioning

confidence: 99%

Multi-View Data Augmentation to Improve Wound Segmentation on 3D Surface Model by Deep Learning

Niri¹,

Gutierrez

Douzi

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Wound area segmentation really progressed with the emergence of deep learning, due to its robustness in uncontrolled lighting and no need to design hand-crafted features but two limits have still to be overcome : firstly, its performance relies on the size and quality of the training dataset in the medical field, where data annotation is costly and time-consuming ; secondly the accuracy of the segmentation depends highly on the camera distance and angle and moreover perspective effects prevent measuring real surfaces in single views. To address concurrently these two issues, we propose to apply multi-view modeling : an image sequence is acquired around the wound site and enables wound 3D reconstruction. Then, a segmentation step is run to extract roughly the wound from the background in each view and to select the best view with an original strategy. This view provides the most accurate segmentation and the real wound bed area even on non planar wounds. Finally, this segmentation is backprojected in each view to generate a complete set of well annotated real images to reinforce the learning step of the neural network.In our experiments, we compare several strategies to select the best view in the image sequence. The proposed method, tested on a dataset of 270 images, outperforms standard deep learning approach based on a single view, as recorded with DICE index and IoU score which rise respectively from 36.53% to 86.3% and 29.48% to 77.09% for the wound class to achieve an overall DICE and IoU score of 93.04% and 86.61% including background class. These results attest to the robustness of our method and its improved accuracy in the wound segmentation task.

show abstract

“…Similar to other medical image segmentation tasks such as skin lesion segmentation [13] or nuclei segmentation in histological images [14], deep learning and convolutional neural networks (CNN)-based approaches have shown to outperform other approaches for foot ulcer segmentation [15]. Some well-known CNN-based architectures such as fully convolutional neural network (FCN), UNet, and mask-RCNN were utilised to perform ulcer segmentation in former studies [15,16,17,18]. In this work, inspired by our former studies for other medical image segmentation tasks [13,19,20], we proposed and developed a model based on two well-established encoderdecoder-based CNNs, namely UNet and LinkNet, to segment wounds in clinical foot images.…”

Section: Introductionmentioning

confidence: 99%

Automatic Foot Ulcer Segmentation Using an Ensemble of Convolutional Neural Networks

Mahbod¹,

Schaefer²,

Ecker³

et al. 2021

Preprint

View full text Add to dashboard Cite

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.

show abstract

Semantic Segmentation of Diabetic Foot Ulcer Images: Dealing with Small Dataset in DL Approaches

Cited by 31 publications

References 16 publications

Multimodal and Multiview Wound Monitoring with Mobile Devices

Multimodal and Multiview Wound Monitoring with Mobile Devices

Multi-View Data Augmentation to Improve Wound Segmentation on 3D Surface Model by Deep Learning

Automatic Foot Ulcer Segmentation Using an Ensemble of Convolutional Neural Networks

Contact Info

Product

Resources

About