ABSTRACT
The prominence of machine learning at leading radiological conferences testifies to how radiologists have embraced the transformative technology. Despite some bold announcements on the existential threat that artificial intelligence (AI) poses for radiology as a profession, many prominent radiologists and scientists are optimistic regarding its potential long term benefits. Investments and research in applied AI have skyrocketed, thanks to the successes of deep learning in computer vision, natural language processing and many other fields. However, healthcare applications require extensive clinical validation, and their true clinical utility on health outcomes may not be truly assessed until they are available on the market. Many AI-based applications are at the research and development stage, or have been cleared by regulatory agencies. The AI development model is largely data-driven, and as such the quality and quantity of data available for training are of paramount importance. Difficulties in accessing adequate datasets for training and testing are key barriers to be overcome in the near future.
