Periodic Reporting for period 1 - AISN (Integrating AI in Stroke Neurorehabilitation)
Période du rapport: 2022-12-01 au 2024-05-31
The AISN project aims to revolutionize post-stroke rehabilitation by integrating AI into healthcare, addressing a critical global health issue. Annually, 13.5 million people suffer strokes, with 40% requiring rehabilitation and 80 million chronic stroke survivors worldwide. AISN's comprehensive AI health platform integrates validated systems for data acquisition, clinical interpretation, whole-brain simulation, and intervention delivery. This platform will be validated, transforming care pathways and informing new treatment guidelines. AISN emphasizes ethical AI deployment by developing robust legal and ethical guidelines, ensuring fair and trustworthy implementation, and validating acceptance and transparency. The project focuses on evidence-based interventions, transparency, prognostics, personalized treatments, and information access for clinicians, patients, and caregivers. Stroke's economic burden is significant, with annual costs of 60 billion euros in Europe and 46 billion dollars in the United States. AISN's AI-enhanced healthcare aligns with Value-Based Healthcare (VBHC) principles, linking reimbursement to clinical outcomes, enhancing results, and reducing costs. The project tackles barriers to AI adoption, such as staff shortages, skill gaps, and interoperability issues, offering a TRL6 integrated platform for AI-enhanced VBHC. To achieve these outcomes, AISN has 4 key objectives:
1. AISN Platform with integration of EBRAINS' Knowledge Graph, CHARITE’s Virtual Research Environment, and EODYNE's Rehabilitation Gaming System with advanced AI tools like SADDLE's Bayesian Inference Engine.
2. AI-based Decision-Support Module for Clinicians to tailor treatment recommendations for stroke patients rehabilitating at home.
3. Clinical validation and guidelines for long-term stroke care, aligning with European objectives for personalized rehabilitation and ethical AI deployment.
4. Inclusion of clinicians, patients, and caregivers through the AISN education platform, enhancing AI adoption and creating tailored guides for clinicians, patients, and caregivers.
1. AISN Platform with integration of EBRAINS' Knowledge Graph, CHARITE’s Virtual Research Environment, and EODYNE's Rehabilitation Gaming System with advanced AI tools like SADDLE's Bayesian Inference Engine.
2. AI-based Decision-Support Module for Clinicians to tailor treatment recommendations for stroke patients rehabilitating at home.
3. Clinical validation and guidelines for long-term stroke care, aligning with European objectives for personalized rehabilitation and ethical AI deployment.
4. Inclusion of clinicians, patients, and caregivers through the AISN education platform, enhancing AI adoption and creating tailored guides for clinicians, patients, and caregivers.
AISN has focused on scientific advancements and developing signal processing toolboxes and computational modeling. The team implemented a whole-brain Wilson-Cowan model to study brain dynamics post-lesion, validated with EEG and fMRI data from stroke patients. They analyzed EEG data, quantified metastability over time, and correlated it with recovery using clinical scales, identifying metastability and synchrony as potential biomarkers. Preliminary analysis of RGS@home (RGS: Rehabilitation Gaming System) intervention data evaluated progress, revealing covariate effects on clinical scores. The team developed pipelines to extract kinematic variables correlated with clinical scales and deployed a Docker-based infrastructure with a Vector DB and FastAPI for a recommender engine, planning to validate it with real clinical data. They advanced RGS solutions by improving RGSapp with new features like video tutorials, a more interactive user interface, and enhanced messaging systems with machine learning-driven personalized coaching. Efforts to improve cybersecurity included upgrading servers and optimizing data transfer methods. The integration of the algorithm for diagnosis and prognosis on the server has been successful. Improvements in RGSweb protocol development, RGSwear's user interface, and data transfer reliability were achieved. Accessibility enhancements included adding Romanian language support and creating user manuals. The Medical Information Management System (MIMS) platform now offers personalized patient experiences and preliminary diagnostic outputs. AISN developed a robust AI pipeline for patient prognosis, using statistical models for interpretable AI algorithms. The prognosis module, integrated into the AISN system, supports real-time prediction and data storage, enhancing clinical trials' prognostic predictions and providing insights for clinicians. The project developed standards for structuring computational modeling data, applied existing standards to multimodal brain imaging data, and organized and annotated stroke datasets accordingly. These datasets are openly discoverable via EBRAINS KG, with guidelines for GDPR-compliant data sharing.
The AISN project has identified metastability and synchrony as potential biomarkers for stroke recovery, enhancing the understanding of brain dynamics post-lesion. The detailed examination of RGS@home intervention data provides insights into key factors affecting recovery and suggests directions for future improvements. Pipelines for extracting kinematic variables allow for clinical scale-free assessments of recovery. The deployment of a Docker-based infrastructure for a recommender engine is a substantial technological advancement, with the potential to significantly impact personalized healthcare solutions. The advancements in RGS solutions, such as personalized and interactive elements in RGSapp, combined with enhanced cybersecurity and reliable data handling, ensure robust and secure user experiences. The successful integration of the algorithm marks a step forward in personalized healthcare, with potential for improved diagnostic and prognostic capabilities. Continued improvements in RGSweb and RGSwear enhance the usability and accuracy of the solutions, while the personalized features of the MIMS platform address patient needs directly. AISN has achieved advancements in patient prognosis and clinical decision support. By integrating the prognosis module into the AISN system, real-time prediction and data storage enhance clinical decision-making efficiency and accuracy. Next steps are to conduct more research and data collection from diverse sources, improve the prognosis model's accuracy, and add new features to the decision support module. This involves continuous platform maintenance, additional data gathering, and model refinement. By extending features, optimizing AI modules for scalability, and ensuring efficient and secure operation, we advance clinical decision support. Charité’s standardized stroke patient datasets, processing workflows, GDPR ready infrastructure, and guidelines for data sharing benefit the European and international research community, creating an impact beyond the project.