Fast density-based lesion detection in dermoscopy images
Introduction
Melanoma is the fifth most common malignancy in the United States [1]. Malignant melanoma, the most deadly form of skin cancer, is one of the most rapidly increasing cancers in the world. 8441 deaths out of 68,720 incidences are estimated numbers in the United States during 2009 [2]. Early diagnosis is particularly important for melanoma since it can be cured with a simple excision operation in early stages of the disease.
Dermoscopy, which is one of the non-invasive skin imaging techniques, has become a principal tool in the diagnosis of melanoma and other pigmented skin lesions. It involves optical magnification of the region-of-interest, which makes subsurface structures more visible than conventional macroscopic images [3]. This in turn improves screening characteristics and provides greater differentiation between difficult lesions such as pigmented Spitz nevi and small, clinically equivocal lesions [4]. However, it has also been demonstrated that dermoscopy may actually lower the diagnostic accuracy in the hands of inexperienced dermatologists [5]. Therefore, novel computerized image understanding tools are needed to minimize the diagnostic errors. These errors are generally caused by the complexity of the incidents and the subjectivity of visual interpretations [6], [7].
For many reasons, delineation of region-of-interest is the first and key step in the computerized analysis of skin lesion images. First of all, the border structure provides essential information for an accurate diagnosis. For instance, asymmetry, border irregularity, and abrupt border cutoff are some of the critical features calculated based on the lesion border. Furthermore, the extraction of other critical clinical indicators such as atypical pigment networks, globules, and blue-white areas depend on the border detection [8].
In the literature, many algorithms were proposed to detect the borders in dermoscopy images. Those include the principal component transform (PCT)/median cut algorithm [9], adaptive thresholding, the first image plane of the PCT [10], thresholding in the blue image plane [11], k-means clustering [12], split-and-merge [9], [13], a segmentation technique based on a Markov random field (MRF) image model [14], and a non-linear diffusion technique [12].
Schmid [15] proposed an algorithm based on color clustering. First, a two-dimensional histogram is calculated from the first two principal components of the CIE L*u*v* color space. The histogram is then smoothed and initial cluster centers are obtained from the peaks using a perceptron classifier. At the final step, the lesion image is segmented by using a modified version of the fuzzy c-means clustering algorithm. Gao et al. [12] created two methods: one based on stabilized inverse diffusion equations, a form of non-linear diffusion and another one based on Markov random fields in which the model parameters are estimated using the mean field theory.
Regarding boundary of clusters, Lee and Estivill-Castro [16] introduced a new algorithm of polygonization based on boundary of resulting point clusters. Recently Nosovskiy et al. [17] used another theoretical approach to find boundary of a cluster in order to infer accurate boundary between close neighboring clusters. These two works principally study boundaries of finalized data groups (clusters), which is not the case in our study.
In this study, we proposed a novel data mining approach to the problem of border identification: a boundary-driven density-based algorithm. It is based on a well-known density-based approach, called DBSCAN (A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise) [18]. Since DBSCAN is very effective in discovering clusters of arbitrary shapes, it efficiently finds different structures in a given dataset. Therefore, it was successfully used for synthetic datasets as well as earth science and protein datasets. Although DBSCAN can handle multidimensional datasets, in this study our focus is on dataset of 2D dermoscopy images.
Clustering is an unsupervised learning process to find related items in groups, called clusters. For dermoscopy images studied in this communication, clustering corresponds to separating background from skin lesion. To do that; first, we cluster pixels of a thresholded image. Our algorithm takes a binary (pre-segmented) image, and delineates only significantly important regions. The significantly important regions are determined by using our clustering algorithm. The outcome of this framework is the desired skin lesion and its boundary in a dermoscopy image.
The rest of the paper is organized as follows. In Section 2, we first describe the segmentation technique that we employ as a preprocessing step (pre-segmentation). Next, we introduce a novel density-based clustering algorithm for 2D datasets for further segmentation. Section 3 details image dataset, experimental results, discusses findings. Section 4 concludes this study emphasizing future works.
Section snippets
Methodology
The original dermoscopy dataset is in PNG format and each sample includes three color channels, RGB. To make use of our improved density-based algorithm, we first represent each image's RGB color channels in single channel called luminance. This is achieved by calculating the lightness component of the HSL color space [19] as follows:
Fig. 1 demonstrates an original (RGB) image and corresponding grayscale image was obtained using the above formula.
The next step is to
Experiments
The proposed method is tested on a set of 100 dermoscopy images obtained from the EDRA Interactive Atlas of Dermoscopy [22]. They are 24-bit RGB color images with dimensions ranging from 577 × 397 to 1921 × 1285 pixels. The benign lesions include nevocellular nevi and dysplastic nevi. The distance function used is Euclidean distance between pixels p and q, and given aswhere p.x and p.y denote position of pixel p at xth column and yth row with respect to top-left corner
Conclusion
In this study, we introduced a novel framework for automatic detection of skin lesions. First, intermeans algorithm, non-parametric segmentation method, for pre-segmentation of dermoscopy images is introduced. Second, in order to further improve segmentation results for dermoscopy images an improved version of DBSCAN algorithm – a fast density-based algorithm (BD-DBSCAN) – is introduced. Results showed that the number of region queries produced by BD-DBSCAN is 1/7th of DBSCAN.
The assessments
Acknowledgment
The authors thank Dr. Emre Celebi from Department of Computer Science, Louisiana State University, Shreveport for providing the dataset.
References (26)
- et al.
Techniques for a structural analysis of dermatoscopic imagery
Computerized Medical Imaging and Graphics
(1998) - et al.
Lesion border detection in dermoscopy images
Computerized Medical Imaging and Graphics
(2009) - et al.
Segmentation of images of skin lesions using color and texture information of surface pigmentation
Computerized Medical Imaging and Graphics
(1992) - et al.
Automatic clustering and boundary detection algorithm based on adaptive influence function
Pattern Recognition
(2008) - et al.
A Generalized lightness, hue, and saturation color model
CVGIP: Graphical Models and Image Processing
(1993) Another efficient algorithm for convex hulls in two dimensions
Information Processing Letters
(1979)- American Cancer Society. Cancer Facts & Figures, Available at...
- et al.
Cancer statistics
A Cancer Journal for Clinicians
(2008) - et al.
Dermoscopy: a tutorial
(2002) - et al.
Statistical evaluation of epiluminescence dermoscopy criteria for melanocytic pigmented lesions
Journal of the American Academy of Dermatology
(1993)
Microscopy: a useful tool for the diagnosis of pigmented skin lesions for formally trained dermatologists
Archives Of Dermatology
Unsupervised border detection in dermoscopy images
Skin Research and Technology
Unsupervised color image segmentation with application to skin tumor borders
IEEE Engineering in Medicine and Biology
Cited by (51)
Application of the novel harmony search optimization algorithm for DBSCAN clustering
2021, Expert Systems with ApplicationsCitation Excerpt :As an unsupervised classification technique, the objective of clustering is to sort a sample of objects into different groups without any prior knowledge, which makes objects in the same cluster are more similar to each other than objects in different clusters (Boonchoo, Ao, Liu, Zhao, Zhuang & He, 2019). Clustering has been used in many applications, like document management (Hatamlou, Abdullah & Nezamabadi-Pour, 2012; Santra & Christy, 2012), pattern recognition (Elshourbagy, Hemayed & Fayek, 2016; Rodriguez, Alex, Laio & Alessandro, 2014; Tian, Li, Zeng, Evans & Zhang, 2019), marketing (Ben Salem et al., 2018; Chen, 2015), and health-care (Mete, Kockara & Aydin, 2011). Generally, the clustering methods can be broadly divided into five types: partitional clustering, density-based clustering, hierarchical clustering, grid-based clustering, and model-based clustering (Chen & Xinquan, 2015; Lu, Zhu & Wu, 2018).
A survey on automated melanoma detection
2018, Engineering Applications of Artificial IntelligenceCitation Excerpt :Various methods have been proposed for lesion segmentation, which is principally an image segmentation task. Typical examples are based on thresholding (Garnavi et al., 2011; Celebi et al., 2013), clustering (Schmid, 1999; Mete et al., 2011; Melli et al., 2006), fuzzy logic (Baral et al., 2014), supervised learning (Wu et al., 2013), and graph theory (Yuan et al., 2009). In addition, some of these methods can be combined to achieve higher segmentation accuracy (Celebi et al., 2009a).
CNN-based dermoscopic analysis of vascular skin lesions in the prognosis of skin lesion sarcoma based on ensemble learning
2023, Journal of Intelligent and Fuzzy SystemsSegmentation of the Melanoma Lesion and its Border
2022, International Journal of Applied Mathematics and Computer ScienceEnhanced Diagnostic Methods for Identifying Anomalies in Imaging of Skin Lesions
2022, International Journal of Electrical and Electronics ResearchComputer Aided Skin Disease (CASD) Classification Using Machine Learning Techniques for iOS Platform
2022, Intelligent Systems Reference Library