Counting Moles Automatically From Back Images

Abstract

Abstract-Density of moles is a strong predictor of malignant melanoma, therefore, enumeration of moles is often an integral part of many studies that look at malignant melanoma. An automatic method of segmenting and counting moles would help standardize studies, compared with manual counting. We have developed an unsupervised algorithm for segmenting and counting moles from two-dimensional color images of the back torso region, as part of a study to evaluate the effectiveness of sunscreen. The method consists of a new variant of mean shift filtering that forms clusters in the image and removes noise, a region growing procedure to select candidates, and a rule-based classifier to identify moles. When this algorithm was compared to an assessment by an expert dermatologist, the algorithm showed a sensitivity rate of 91% and diagnostic accuracy of 90% on the test set, for moles larger than 1.5 mm in diameter.

Skin lesion tracking using structured graphical models

“…The most closely related PSL detection method for our application (i.e. back images) is the one by Lee et al (2005), in which they, first, applied thresholding to the output of an image enhanced by an adaptive Gaussian kernel. Then, they detected potential PSL candidates in the thresholded binary image based on geometric feature (e.g.…”

“…Unary Binary Ternary Optimization approach Learning Lee et al (2005), Pierrard and Vetter (2007), Sang et al (2007) × (13) and two newly appearing PSLs are highlighted within the red box (for interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).…”

“…Note that since PSLs may appear or disappear, |V| is not necessarily equal to |V |. Further, in our study, PSLs must be larger than 100 pixels 6 mm 2 to be clinically relevant (Lee et al, 2005). Therefore, to prepare ground truth for the PSLs, we select large enough PSLs and annotate one point per PSL (please refer to Section 3 for more details on PSL ground truth preparation).…”

Face Mole Detection, Classification and Application

“…Lee et al [2] proposed a detection method of moles on human back torso with recognition rate 90%. Mean shift [3] is used to locate mole candidates and noises are removed by morphological operations.…”

A graph-based approach to skin mole matching incorporating template-normalized coordinates

“…The second class detects one or typically more moles in an image capturing a larger field of view, in which each mole occupies only a few pixels, such as those appearing in the full back images of interest in this current work (Figure 1(a)). In [10], for example, back image moles are extracted using a variant of the mean shift algorithm, whereas in [16], facial skin irregularities (nevi) are localized using a multi-scale template matching and texture analysis procedure. The focus of our current work is on mole matching and not on the skin or mole segmentation steps.…”