| Abstract |
Background: Melanoma is a malignant tumor derived from melanocytes. Depth of invasion is the single most important prognostic factor and ,in order to ensure good prognosis, melanoma must be recognized and excised at an early stage. Certain types of melanocytic nevi are recognized as precursors of melanoma and one should distinguish between them before recommending an excisional biopsy. For early melanoma diagnosis, experienced dermatologists have an accuracy of 64-80% using clinical diagnostic criteria, usually the ABCDΕ rule, while automated melanoma diagnosis systems are still considered to be experimental and serve as adjuncts to the naked-eye expert prediction. In an attempt to aid in early melanoma diagnosis, an image processing program was developed on the aim to discriminate melanoma from melanocytic nevi, establishing a mathematical model to come up with a melanoma probability. Color texture parameters, Fractal Dimension and Lacunarity of Melanoma and other Melanocytic nevi were estimated on the same basis. Methods: Digital images of one hundred and thirty two melanocytic skin lesions (23 melanomas and 109 melanocytic nevi) were studied in features of geometry, color and color texture. A total of forty-three variables were studied for all lesions e.g. geometry, color texture, sharpness of border and color variables. Univariate logistic regression analysis followed by 2Log Likelihood test and Spearmans rank correlation coefficient were used to eliminate inappropriate variables, since presence of multicollinearity among variables could cause severe problems in any stepwise variable selection method. Initially, 2Log Likelihood and non-parametric Spearmans rho picked 5 variables to be included in a multivariate model of prediction. The five-variable model was then reduced to three variables and the performance of each model was tested. The "jackknife" method was performed in order to validate the model with the three variables and its accuracy was weighed versus the five variable model by ROC curve plotting. It was concluded that the reduced model did not compromise discriminatory power. Graphic three-dimensional pseudoelevation images of the lesions and surrounding skin were produced to identify irregularities in color texture within the lesions. Estimation of Lacunarity and Fractal Dimension followed in order to produce a numerical estimate of the coarseness of color texture. Estimation of Lacunarity and Fractal Dimension followed in order to produce a numerical estimate of the coarseness of color texture. Results: Not all 43 variables contributed much to the model, therefore they were progressively eliminated and the model was finally reduced to three covariates of significance. A predictive equation was calculated, incorporating parameters of geometry, color and color texture as independent covariates for the prediction of melanoma. The proposed model provides melanoma probability with a 60.9% sensitivity and 95.4% specificity of prediction, an overall accuracy of 89.4% (probability level 0.5) and 8% false negative results. Of the fractal parameters studied, Lacunarity proved to be significant in discriminating melanoma from other benign melanocytic lesions. Graphic 3D pseudoelevation of the studied lesions revealed delicate aspects of irregular pigmentation, imperceptible by naked eye. Conclusions: Through a digital image processing system and the development of a mathematical model of prediction, discrimination between melanomas and melanocytic nevi seems feasible with a high rate of accuracy using multivariate logistic regression analysis. It is not suggested that this approach should replace excision and histological examination of a suspect lesion. The proposed model is a clinical diagnostic tool to aid in early melanoma diagnosis. Expensive and sophisticated equipment is not required and it can be easily implemented in a reasonably priced portable programmable computer, in order to predict previously undiagnosed skin melanoma before histopathology results confirm diagnosis. Furthermore, clinicians readily perceive the resulting geographical images. Irregularity in the anaglyph, which might veil malignancy, is effortlessly identified through these images, and therefore an early excision of a suspect lesion is indicated. Furthermore it could be applied to monitor a suspect lesion over a period of time. Currently such lesions are monitored clinically or by standard photographic means, hoping to detect any change that could be suggestive of malignant transformation. By application of the proposed system in clinical monitoring, regions of abnormal coloration (veiling malignancy) could be identified at an early stage.
|
Μορφομετρική μελέτη μελαγχρωματικών βλαβών προμελανωματικών δερματικών βλαβών και μελανωμάτων στην Κρήτη
