on
A Two-Phase Learning Approach for the Segmentation of Dermatological Wounds
Abstract
Tissue segmentation in photographs of lower limb chronic ulcers is a non-intrusive approach that supports dermatological analyses. This paper presents 2PLA, a method that combines supervised and unsupervised learning strategies for enhancing the segmentation of dermatological wounds. Given an ulcer photo captured according to a fixed protocol, 2PLA first phase performs a pixelwise classification of points of interest, whereas pre-processing filters are employed for the smoothing of image noise. The cleaned image is further sent to the 2PLA divide-and-conquer second phase. It builds upon SLIC superpixel construction algorithm for dividing the lower limb into regions of interest with well-defined borders, and clusters the superpixels by taking advantage of the similarity-based DBSCAN algorithm. et up the phases of our method by using a real annotated set of dermatological wounds, and empirical evaluations on representative samples up to 100,000 points showed a compact Multi-Layer Perceptron with Levenberg-Marquardt training algorithm (Cohen-Kappa = .971, Sensitivity = .98, and Specificity = .98) outperformed other classifiers as 2PLA first phase. Additionally, experimental trials on DBSCAN with five distance functions (L 1 , L 2 , L ∞ , Canberra, and BrayCurtis) indicated L 1 function provided fewer groups in comparison to the competitors, and the number of clusters was an exponential decay to the similarity ratio. Accordingly, we used the elbow criterion for finding the L 1 -based DBSCAN threshold as 2PLA second phase parameterization. We evaluated the fine-tuned setting of our method over a labeled set of ulcer images, and wounded tissues were segmented within a .05 Mean Absolute Error ratio. These results illustrate the impact of learning parameters on 2PLA as well as the method efficacy for wound segmentation.
Published in: 2019 IEEE 32nd International Symposium on Computer-Based Medical Systems
Date of Conference: 5-7 June 2019
Date Added to IEEE Xplore: 05 August 2019
INSPEC Accession Number: 18886298
DOI: 10.1109/CBMS.2019.00076
Publisher: IEEE
Conference Location: Cordoba, Spain, Spain
Github
2PLA: A Two-Phase Learning Approach
WARNING: Before we get started, please notice 2PLA is NOT a clinical software. It is built for reproducibility and demonstration purposes ONLY.
This repository contains the 2PLA (Two-Phase Learning Approach), a method that combines supervised and unsupervised learning strategies for enhancing the segmentation of dermatological wounds.
Given an ulcer photo captured according to a fixed protocol, 2PLA first phase performs a pixelwise classification of points of interest, whereas pre-processing filters are employed for the smoothing of image noise.
The cleaned image is then further sent to the 2PLA divide-and-conquer second phase. It builds upon SLIC superpixel construction algorithm for dividing the lower limp into regions of interest with well-defined borders, and clusters the superpixels by taking advantage of the similarity-based DBSCAN algorithm.
1. Minimum Requirements
- Matlab R2018b.
- Screen resolution: 1920×1080 (or higher).
- 350 MB free disk space.
- 4 GB RAM.
- Intel i5 4th generation core processor.
2. Folder Description
The content of this repository is organized as follows:
- dataset: pixels extracted from various sample values of images of ULCER_SET dataset [1] and employed for the tuning and evaluation of our approach.
- image: 6 images processed by the first phase of the method and the same images processed by the second phase of the method.
- src: implementation of 2PLA, where the main function implements Algorithm 1 of the submitted paper.
3. Third-party software and data
-
This demonstration includes third-party software from Matlab. It uses both ANN and DBSCAN native packages. SLIC superpixel implementation was used as is provided in the Peter Kovesi’s original code, which is similar to the Matlab Toolbox.
-
Images employed in the experiments are derived from ULCER_SET dataset [1]. To obtain the original dataset, please contact the authors and request their permission.
3. Additional Information and Legal Note
The code is provided ‘as is’ and any expressed or implied warranties, including but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed. In no event shall the authors of this software or its contributors be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including but not limited to, procurement of substitutive goods or services, loss of use, data, or profits; or business interruption) however caused and on any theory of liability wheter in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this demonstration, even if advised of the possibility of such damage.
Again, 2PLA is NOT a clinical software.
References
[1] S. M. Pereira, M. A. C. Frade, R. M. Rangayyan, and P. M. Azevedo-Marques. Classification of color images of dermatological ulcers. Journal of Biomedicine and Health, 17(1):136–142, 2013.