SIBGRAPI: Experiments & Results

This page details the experiments of the paper "Statistical Learning Approach for Robust Melanoma Screening". Here we show intermediate results and summarize the pipeline.

Experimental setup

The experiments of that paper are based on a classical 5-fold cross validation schema. The significance of the parameters and their interactions are analysed by a fractional factorial ANOVA, using the AUC as evaluation measure. The ANOVA analysis was done with R language.

Dataset

The dataset is a third-party development and was created by the Department of Medical Informatics, RWTH Aachen University. It is composed of 747 dermoscopic images with resolution of 512x512 pixels, of which 187 images are melanomas and 560 images are benign skin lesions.

Final results

This is the full results of our experiments. Each row contains the values used on each parameter of the pipeline. We also show the AUC for every configuration.