Time filter

Source Type

Wang H.,Missouri University of Science and Technology | Chen X.,Apple Inc | Moss R.H.,Missouri University of Science and Technology | Stanley R.J.,Missouri University of Science and Technology | And 8 more authors.
Skin Research and Technology | Year: 2010

Background/purpose: Automatic lesion segmentation is an important part of computer-based image analysis of pigmented skin lesions. In this research, a watershed algorithm is developed and investigated for adequacy of skin lesion segmentation in dermoscopy images. Methods: Hair, black border and vignette removal methods are introduced as preprocessing steps. The flooding variant of the watershed segmentation algorithm was implemented with novel features adapted to this domain. An outer bounding box, determined by a difference function derived from horizontal and vertical projection functions, is added to estimate the lesion area, and the lesion area error is reduced by a linear estimation function. As a post-processing step, a second-order B-Spline smoothing method is introduced to smooth the watershed border. Results: Using the average of three sets of dermatologist-drawn borders as the ground truth, an overall error of 15.98% was obtained using the watershed technique. Conclusion: The implementation of the flooding variant of the watershed algorithm presented here allows satisfactory automatic segmentation of pigmented skin lesions. © 2010 John Wiley & Sons A/S.

Moloney F.J.,University of Sydney | Moloney F.J.,Royal Prince Alfred Hospital | Moloney F.J.,Sydney Melanoma Diagnostic Center | Moloney F.J.,Materials Misericordiae University Hospital | And 24 more authors.
JAMA Dermatology | Year: 2014

IMPORTANCE: The clinical phenotype and certain predisposing genetic mutations that confer increased melanoma risk are established; however, no consensus exists regarding optimal screening for such individuals. Early identification remains the most important intervention in reducing melanoma mortality. OBJECTIVE: To evaluate the impact of full-body examinations every 6 months supported by dermoscopy and total-body photography (TBP) on all patients and sequential digital dermoscopy imaging (SDDI), when indicated, on detecting primary melanoma in an extreme-risk population. DESIGN, SETTING, AND PARTICIPANTS: Prospective observational study from February 2006 to February 2011, with patients recruited from Sydney Melanoma Diagnostic Centre and Melanoma Institute Australia who had a history of invasive melanoma and dysplastic nevus syndrome, history of invasive melanoma and at least 3 first-degree or second-degree relatives with prior melanoma, history of at least 2 primary invasive melanomas, or a CDKN2A or CDK4 gene mutation. EXPOSURES: Six-month full-body examination compared with TBP. For equivocal lesions, SDDI short term (approximately 3 months) or long term (≥6 months), following established criteria, was performed. Atypical lesions were excised. MAIN OUTCOMES AND MEASURES: New primary melanoma numbers, characteristics, and cumulative incidence in each patient subgroup; effect of diagnostic aids on new melanoma identification. RESULTS: In 311 patients with a median (interquartile range [IQR]) follow-up of 3.5 (2.4-4.2) years, 75 primary melanomas were detected, 14 at baseline visit. Median (IQR) Breslow thickness of postbaseline incident melanomas was in situ (in situ to 0.60 mm). Thirty-eight percent were detected using TBP and 39% with SDDI. Five melanomas were greater than 1 mm Breslow thickness, 3 of which were histologically desmoplastic; the other 2 had nodular components. The benign to malignant excision ratio was 1.6:1 for all lesions excised and 4.4:1 for melanocytic lesions. Cumulative risk of developing a novel primary melanoma was 12.7% by year 2, with new primary melanoma incidence during the final 3 years of follow-up half of that observed during the first 2 years (incidence density ratio, 0.43 [95% CI, 0.25-0.74]; P = .002). CONCLUSIONS AND RELEVANCE: Monitoring patients at extreme risk with TBP and SDDI assisted with early diagnosis of primary melanoma. Hypervigilance for difficult-to-detect thick melanoma subtypes is crucial. Copyright 2014 American Medical Association. All rights reserved.

Wang H.,Missouri University of Science and Technology | Moss R.H.,Missouri University of Science and Technology | Chen X.,Apple Inc | Stanley R.J.,Missouri University of Science and Technology | And 8 more authors.
Computerized Medical Imaging and Graphics | Year: 2011

In previous research, a watershed-based algorithm was shown to be useful for automatic lesion segmentation in dermoscopy images, and was tested on a set of 100 benign and malignant melanoma images with the average of three sets of dermatologist-drawn borders used as the ground truth, resulting in an overall error of 15.98%. In this study, to reduce the border detection errors, a neural network classifier was utilized to improve the first-pass watershed segmentation; a novel "edge object value (EOV) threshold" method was used to remove large light blobs near the lesion boundary; and a noise removal procedure was applied to reduce the peninsula-shaped false-positive areas. As a result, an overall error of 11.09% was achieved. © 2010 Elsevier Ltd.

Loading Sydney Melanoma Diagnostic Center collaborators
Loading Sydney Melanoma Diagnostic Center collaborators