AutoPiX Revolutionizes Arthritis Care with AI-Powered Imaging

The Challenge

Before partnering with us, the AutoPiX consortium faced significant hurdles in establishing a cohesive data infrastructure. Patients with arthritis are constantly assessed by imaging techniques, but these images often lack the analysis and interpretation tools needed for unbiased diagnosis, monitoring, and prognosis. Existing quantitative scoring methods were time-consuming and not suitable for routine settings, leading to a vast amount of unutilized quantitative information. The project needed to overcome challenges related to data standardization across multiple sources, ensuring GDPR and IP compliance, and creating a scalable environment for developing and validating AI models. It was crucial to integrate heterogeneous imaging datasets and algorithms from diverse partners—including existing images from routine clinical care, cohorts, registries, and clinical trials, as well as newly generated images—into a unique and sustainable "data lake".

The Solution

The AutoPiX consortium recognized our established expertise in providing cloud infrastructure and platform solutions for healthcare collaboration. They selected us as the lead for Work Package 2 (WP2), responsible for "Data and AI infrastructure, legal issues, and data protection". Our platform was chosen for its proven regulatory compliance (MDR, GAMP5, 21 CFR Part 11, ISO27001 certified) and its design based on privacy-by-design and GDPR principles.

We provided a highly secure, FAIR (Findable, Accessible, Interoperable, Reusable), and GDPR-compliant interface for clinical site connections, data storage, curation, classification, and processing through our cloud platform. Key features that enabled AutoPiX's success include:

• CM Connect: An edge server installed at each hospital partner to de-identify and pseudonymize imaging data on-premise before encrypted transfer to our central repository, ensuring data remains traceable for data controllers while being de-identified from other repository members' perspectives.
• Subject-Centric Data Model: Unlike traditional data/modality-centric repositories, our platform adopts a patient/subject-centric approach, where multi-modal, multi-versioned data is structured by subject, making it instantly available for analysis and AI/ML algorithm creation.
• Secure Processing Environment (SPE): Our platform acts as a secure processing environment compliant with EU guidelines and regulations, including the European Health Data Space (EHDS) framework, providing robust data security and access control.
• Annotation, Segmentation, and eCRF Tools: For retrospective data classification, our Research Platform offers tools that allow unlimited users to collaborate on data classification at scale with a clinical-grade User Interface (UI) accessible via a web browser.
• Audit Trail and E-Signatures: The platform maintains an audit trail and supports e-signatures for task sign-offs, providing an auditable research environment approved for sponsored clinical trials.
• Central Compute Capabilities: We added cloud compute capabilities to the repository, allowing compute tasks to be run on the AutoPiX repository without transferring data outside the virtual private cloud (VPC).

Our robust legal and technical framework, coupled with our experience from other European Horizon/IHI projects (such as EUCanImage, PancAIM, PancAID, GUIDE.mrd, AI-Pod, and NetZeroAICT), solidified the consortium's decision.

The Results

Leveraging the Collective Minds Platform, the AutoPiX consortium achieved significant milestones and impacts:

• Enhanced Data Accessibility and Validation: The centralized data lake enabled rapid cross-validation of AI models across multiple cohorts, providing robust evidence of their reliability against population, design, and "batch effects" variations.
• Streamlined Data Management: We coordinated and delivered the initial Data Management Plan (DMP), with ongoing progress in connecting clinical partners and establishing legal frameworks for data transfer.
• Accelerated Research and Innovation: The secure, interoperable data infrastructure allows AI models to be refined and tested more efficiently, leading to faster development of imaging biomarkers. This is expected to reduce drug development costs and shorten clinical study durations.
• Improved Patient Outcomes: The project's advancements in AI-supported image analysis are anticipated to lead to more precise, accessible, and effective diagnoses, shortened treatment paths, and improved treatment response assessments for arthritis patients.
• Increased Competitiveness: Our solution directly contributes to increasing the competitiveness of European health industries by providing a GDPR-compliant infrastructure that advances the utility of imaging data for precision medicine and novel therapeutics. This also empowers SMEs in the consortium to integrate digital tools into clinical care and expand their market presence.