The Future of AI-Driven Technologies in Medical Laboratories

Abstract

In this systematic review, we conclude that many AI models in health care result from volumetric exploration and almost exclusively focus on black box models whose underlying mechanisms or reasons for making specific predictions are not understood by the user. However, most of the current methods are not transparent and difficult to understand due to their inherent black-box nature. Our analysis also shows that medical decision relevant AI models do not yet consider the factors of explainability, interpretability, reproducibility, robustness, and trustworthiness of deployment in clinical routine. All the above-described properties are essential to ensure the quality and safety of clinical decisions based on AI.

Ähnliche Arbeiten