Abstract
An automated melanocytic lesion image-analysis algorithm is described that aims to reproduce the decision-making of a dermatologist. The utility of the algorithm lies in its ability to identify lesions requiring excision from lesions not requiring excision. Using only wavelet coefficients as features, and testing three different machine learning algorithms, a cohort of 250 images of pigmented lesions is classified based on expert dermatologists’ recommendations of either excision (165 images) or no excision (85 images). It is shown that the best algorithm utilises the Shannon4 wavelet coupled to the support vector machine, where the latter is used as the classifier. In this case the algorithm, utilising only 22 othogonal features, achieves a 10-fold cross validation sensitivity and specificity of 0.96 and 0.87, resulting in a diagnostic-odds ratio of 261. The advantages of this method over diagnostic algorithms–which make a melanoma/no melanoma decision–are twofold: first, by reproducing the decision-making of a dermatologist, the average number of lesions excised per melanoma among practioners in general can be reduced without compromising the detection of melanoma; and second, the intractable problem of clinically differentiating between many atypical dysplastic naevi and melanoma is avoided. Since many atypical naevi that require excision on clinical grounds will not be melanoma, the algorithm–in contrast to diagnostic algorithms–can aim for perfect specificities without clinical concerns, thus lowering the excision rate of non-melanoma. Finally, the algorithm has been implemented as a smart phone application to investigate its utility in clinical practice and to streamline the assimilation of hitherto unseen tested images into the training set.
Highlights
The incidence of melanoma has increased substantially in the United States, Europe and Australia over the last 30 years [1]
Stabilisation of melanoma rates in Australia is largely thought to be a consequence of the public awareness campaigns that began in the early 1980s
Decision support and melanoma pigmented lesions excised are not melanomas [2]. This phenomenon can be captured by a measure known as the ‘Number Needed to Treat’ (NNT), a term loosely defined as the number of benign lesions excised per melanoma [3]
Summary
The incidence of melanoma has increased substantially in the United States, Europe and Australia over the last 30 years [1]. There will exist considerable overlap between any set of morphologic criteria that identifies melanoma and identifies lesions requiring excision (after all, Decision support and melanoma the purpose of clinical evaluation of melanocytic lesions is to identify potential melanoma), the feature set cannot be the same with respect to the classification problem because the respective algorithms are classifying different classes of objects. While these observations do not prove that a feature set derived using well-defined morphologic parameters could not be an efficient classifier with respect to a decision-making algorithm, it is apparent that feature selection would involve considerable subjectivity, and require multiple rounds of training and testing Motivated by these considerations, an alternative approach will be implemented: the feature selection process desribed here will be limited to obtaining and analysing the statistical properties of wavelet coefficients derived from dermoscopic image data of melanocytic lesions. The following sections describe the approach to this problem, the results, and discuss the utility of the algorithm–which has been developed as a smart-phone application for research purposes–in the clinical setting
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
