Key points are not available for this paper at this time.
Artificial intelligence (AI) has significant potential to transform healthcare and improve patient care. However, successful development and integration of AI models requires careful consideration of study designs and sample size calculations for development and validation of models, publishing standards, prototype development for translation and collaboration with stakeholders. As the field is relatively new and rapidly evolving there is a lack of guidance and agreement on best practices for most of these steps. We engaged stakeholders in the form of clinicians, researchers from academia and industry, and data scientists to discuss various aspects of the translational pipeline and identified the challenges researchers in the field face and potential solutions to them. In this viewpoint, we present the summary of our discussions as a brief guide on the process of developing AI-based applications for healthcare using medical images. We organized the entire process into six major themes (i.e., The gaps AI can fill in healthcare, Development of AI models for healthcare: practical and important things to consider, Good practices for validation of AI models for healthcare: study designs and sample size calculation, Points to consider when publishing AI models, Translation towards products, Challenges and potential solutions from a technical perspective) and presented important points as a rule of thumb. We conclude that successful integration of AI in healthcare requires a collaborative approach, rigorous validation, adherence to best practices as described and cited, and consideration of technical aspects.
Desiraju et al. (Fri,) studied this question.