it is not feasible to implement the training process and the feature
extraction on current smartphones due to its computational demands.
Furthermore, the supervised learning algorithm requires
a large number of training image samples and experienced clinical
input, which is difficult and costly.