Regulation and Validation Challenges in Artificial Intelligence-Empowered Healthcare Applications—The Case of Blood-Retrieved Biomarkers

Abstract

Biomarkers have been proposed as powerful classification features for use in the training of neural network-based and other machine learning and artificial intelligence-based prognostic models in the field of personalized medicine and for targeted interventions in patient management. Biomarkers are measurable indications of a health state, that can be derived from blood sample, tissue or other bodily fluid. An example of a biomarker is the electrocardiogram that records electrical signals from the heart, and thus evaluates heart condition. Biomarkers can lead to actionable insights and for that are important tools for patient management and treatment administration. In this paper, we outline a medical application with a machine learning backbone built with biomarkers retrieved from blood exams that define health states (obesity, metabolic syndrome and systolic pressure), via rational unified process and cross industry standard process for data mining. By adopting novel ways to deploy these industry standards we can identify health sector related requirements and challenges and thus design and propose smart solutions that add value to all stakeholders. New technologies have the potential to create new pathways in medicine by bridging the gap between the laboratory and the patient, however strong medical validation of processes is required to ensure usability and patient’s safety. We recognise regulation and validation as key challenges and important factors for the improvement of the development of health care applications. Towards this we shall define when a software application is considered as a medical device. Since the regulator is identified as an important stakeholder, strategies are suggested for the proper handling of this stakeholder through out the production cycle.