All Times EDT
There is an increasing number of AI-related devices including algorithm and software developed for aid in the disease diagnosis and management. It is very challenge to validate the AI-related devices with complex models in helping the users including physicians and patients to identify, monitor, and manage the disease beyond the traditional metrics such as the sensitivity and specificity for diagnostic devices. This presentation will discuss the issues and examples in the study design and statistical analysis for these AI related devices including special analytical studies for repeatability and reproducibility, validating risk probability/risk scores, studies for disease with low prevalence, validation using single arm study with paired data or two arm study, and validation using clinical outcome studies.