De noodzaak aan data voor artificiële intelligentie in de geneeskunde

Abstract

The need for data for artificial intelligence in medicine In recent decades, there has been a digital revolution in medicine, with an increasing integration of innovative technologies across different disciplines in the medical world. Artificial intelligence (AI), in particular, has the potential to have a groundbreaking impact on the healthcare of the future. However, the core of this promising technology heavily relies on data. Relevant literature was systematically and structurally searched through the databases of PubMed and Embase. Interviews were conducted with experts based on the insights and considerations from the literature. These interviews formed the foundation of this paper. Finally, the interviews were supported by relevant websites and literature found through Google Scholar. To develop a generalizable algorithm, the used data should not only have a high quality, but must also be numerous and diverse. However, there is not necessarily a need for more data, but rather for accessibility of the data. In clinical practice, a standardized format to store data is lacking. Furthermore, the data are scattered across different centres, with data-sharing heavily protected by the GDPR. There is a need for uniform and linkable data that can be collected from multiple healthcare institutions in a structured and protected manner using a centralized data platform. This data should have a high quality and must be sufficient in number to develop a robust and representative algorithm. The entire process must comply with the strict obligations imposed by the GDPR, ensuring the protection of the patients’ privacy.