ABSTRACT
In this chapter, the authors propose a probabilistic framework for classification based on neural networks and apply the framework to the problem of classifying skin lesions. The defined probabilistic framework for classification included optimal decision rules, derivation of error functions, model complexity control, and assessment of generalisation performance. The neural classifier framework was applied to the malignant melanoma classification problem using the extracted dermatoscopic features and results from histological analyses of skin tissue samples. Malignant melanoma is the deadliest form of skin cancer and arises from cancerous growth in pigmented skin lesions. The incidence of malignant melanoma in Denmark has increased five-to sixfold from 1942 to 1982, while the mortality rate has doubled from 1955 to 1982. Due to the rather steep increase in the number of reported malignant melanoma cases, it is becoming increasingly important to develop simple, objective and preferably non-invasive methods that are capable of diagnosing malignant melanoma.
