Skin Lesion Boundary Detection with Neural Networks on iOS Devices

Abstract

Automated skin lesion boundary detection has become a common issue in Health Care. On the one hand, a broad variety of image processing algorithms already exist and they are power consuming on mobile devices. On the other hand, the use of machine learning algorithms is on the rise and new frameworks have been developed to use these techniques with improved on-device-performance. Since iOS 11.0, Apple is providing a Core Machine Learning Interface to use machine learning models. Moreover, conversion tools allow integration of 3rd party models into iOS applications. In this paper, we present an overview of available frameworks for iOS devices as well as their limitations and evaluate in practice the performance and maturity level of Neural Network frameworks for skin lesion boundary detection using only freely available pictures.