CAPABLE D5.7: Refined Framework and Models of All Prototypes Based on Accumulated Data

The main goals of the statistical and prediction models in WP5 include clinical aspects, research aspects, and technical aspects, as described below:

1. Clinical goal – Enable more informed treatment selection and patient management by using automatic data-driven analysis models derived from CAPABLE clinical and sensors data. These models can support users of the Decision Support System (DSS) as well as of the Virtual Coach System that is part of the AI framework in CAPABLE architecture. 
2. Research goal – Advance the state-of-the-art methods and tools in predictive models and their interpretation for clinical practice. Specifically, we concentrate on multimodal models including imaging data that can benefit CAPABLE when imaging data will be accumulated in the future.
3. Technical goal - Use and contribute to open-source frameworks and tools to enrich the biomedical research community and foster collaboration. This supports making technical assets developed within CAPABLE, sustainable beyond the project pilot. 

The document is organized according to these three goals in the following way: 

• Section 3 provides a brief overview of the AI framework and its components, healthcare professionals' (HCPs) needs for statistical-based decision support, and the implementation of these needs by corresponding components in the AI framework. 
• Section 4 demonstrates the statistical analysis models performed on data collected from sensors (watches) provided to the CAPABLE pilot participants. It shows the correlation between the sensor time series and the side effects that the patient encounters. The demonstration is presented as a video and includes all the steps in data analysis. Voice-over and a walk-through in this document explain all the steps of the demonstration. This section supports the clinical goal of the statistical and prediction models in WP5. 
• Section 5 provides state-of-the-art multimodal models including clinical and imaging data to predict the disease progression of kidney disease. This complements the prediction models created from the clinical data as described in previous deliverables and the published paper. It shows models that can be used in CAPABLE if additional multimodal imaging data is provided. This section supports the research goal of the statistical and prediction models in WP5. 
• Section 6 describes our contribution to open source, specifically to the BiomedSciAI GitHub organization. It shows the creation of sustainable assets and our increasing contribution to the biomedical community. This section supports the technical goal of the statistical and prediction models in WP5.