Development of an End-to-End Web Application for Visualization, Evaluation, and Post-processing of Result Data from Neural Network Predictions for the Melanoma Use Case

Image acquisition technology advances in various fields, yet current image analysis tools limit the effective application of image analysis due to their cumbersome operation process and the requirement of professional knowledge and skills. In this paper, we develop a semi-automated web application for image segmentation and classification tasks with the support of neural networks (relieving the above-mentioned current research dilemma) using melanoma detection as a use case. The web application enables scientists to participate and improve the decision-making process of the neural network through the concept of “human-in-the-loop”, while saving expensive labor costs due to its automation in image annotation and classification. In addition, our web application achieves high usability in the general user community by testing seven aspects of usability: the first impression, distinguishability and clarity of the tools, intuitive characteristics, learnability, feedback and reaction, implementation of expected functionality, and the fulfillment of usability.