Work Package 2

Work Package 2 (WP2) is about creating a solid foundation for sharing data securely while following legal, ethical, and privacy standards (like GDPR). It involves setting up clear rules for intellectual property and ensuring that all data from academic and industry partners can be shared properly.

These are WP2’s tasks:

• Legal Framework: This task ensures that all data-sharing activities comply with legal rules like GDPR and that all partners sign agreements on sharing and processing data.
• Data Infrastructure: A central data "lake" will be set up where all collected data (such as medical images and patient information) will be stored. This system will allow for easy management and access, ensuring the data can be used to develop AI models.
• Data Curation: Data will be carefully organised, cleaned, and annotated to prepare it for AI research, ensuring it’s usable for tasks like training AI models.
• Data Warehousing: A secure cloud-based environment will be created where AI models can be run and analysed using the data stored in the system.
• AI Model Inventory: At the beginning of the project, an inventory of available AI models will be created to understand their strengths and weaknesses so the team can decide how best to use them.

All these tasks aim to ensure that data are appropriately managed, safely shared, and ready to help develop AI tools that can analyse medical images and related data.

Work Package 3

Work Package 3 (WP3) is focused on using remote patient monitoring (RPM) technology to help track and manage conditions like rheumatoid arthritis (RA) and psoriatic arthritis (PsA) while following high ethical standards and protecting patient data.

These are WP3’s tasks:

• Ethics and Risk/Benefit Assessment: This task ensures that all RPM technologies follow legal and ethical standards, including patient privacy (GDPR) and medical device regulations while evaluating their risks and benefits for patients.
• Creating a Monitoring Dashboard: A cloud-based dashboard will be developed, allowing healthcare providers to remotely monitor images and data sent by patients, making it easier to track their condition.
• Testing RPM for At-Risk Patients: A study will be conducted on patients at risk for RA to test how RPM technology can detect early signs of arthritis. Patients will use RPM to self-monitor and send data, which doctors will check during in-clinic visits.
• Expanding RPM for Established RA/PsA Patients: The RPM technology will be expanded to monitor patients with RA or PsA, especially those in remission. The goal is to detect flare-ups early using remote monitoring, with further in-clinic checks for validation.
• Point-of-Care Ultrasound: Robotic ultrasound machines will be used to conduct regular scans of joints in patients with RA or PsA to detect signs of worsening disease. These images will be analysed with AI and reviewed by ultrasound experts to improve diagnosis and help refine the technology for future use.

These tasks aim to improve patient monitoring and make early detection of disease changes easier through remote tools, ultimately improving care and treatment outcomes for patients with RA and PsA.

Work Package 4

Work Package 4 (WP4) focuses on developing AI-based systems to automatically assess structural and inflammatory changes in patients with RA, PsA, and osteoarthritis (OA) joints. These systems will use images from different types of scans (radiography, MRI, and ultrasound) to improve disease diagnosis and track changes over time.

These are WP4’s tasks:

• Image Annotation: The goal is to comprehensively annotate images (from CR, MRI, and US) to create a detailed database (data lake). These images will be labelled with specific features (e.g., bone erosion, structural damage), forming the foundation for training AI models.
• AI Scoring Systems: Several AI models will be developed and validated for automating the scoring of images in RA, PsA, and OA across different imaging methods:
  1. Conventional Radiography (CR): Scoring systems will be developed to assess structural damage.
  2. Magnetic Resonance Imaging (MRI): AI models will be designed to score joint changes.
  3. Ultrasound (US): AI will be used to assess both inflammatory changes and structural damage.
• Longitudinal Validation: The AI models will be validated to detect short-term changes in joint conditions and predict long-term outcomes in RA and PsA.
• Differential Diagnosis: An AI model will be trained to distinguish between RA, PsA, and OA based on the images, using data from clinical cohorts, imaging from routine clinical practice, and studies conducted in the AutoPiX project.

The overall goal of WP4 is to improve the diagnosis and tracking of RA, PsA, and OA through AI-assisted analysis of radiographic, MRI, and ultrasound images. This will involve annotating thousands of images to create robust datasets, developing automated scoring systems, and validating these systems for both short-term and long-term use. Ultimately, this will lead to better diagnostic tools and more personalised treatment strategies.

Work Package 5

Work Package 5 (WP5) is similar to WP4 but focuses on developing AI systems for automated analysis of images from conventional radiographs (CR) and MRI scans of the sacroiliac joints (SIJ) and spine in axial arthritis (including ankylosing spondylitis (axSpA) and PsA). The ultimate goal is to improve diagnosis, track disease progression, and differentiate between types of axial arthritis and other spine-related diseases.

Its tasks are:

• Image Annotation: Comprehensive annotation of CR and MRI images of the SIJ and spine will be conducted to allow supervised AI learning models. These annotations will focus on assessing both inflammatory and post-inflammatory changes related to axial arthritis using established scoring methods like mSASS (CR) for axSpA and SvdH (CR) for PsA, along with SPARCC (MRI) and PSAMRIS (MRI).
• Automated Scoring for CR: Develop and validate AI-based automated scoring systems for CR images of the SIJ and spine in axial arthritis, detecting structural damage.
• Longitudinal Validation for CR Scoring: Validate the AI models for CR images to assess their ability to detect short-term changes and predict long-term disease outcomes.
• Automated Scoring for MRI: Develop and validate AI-based systems for automated scoring of MRI images of the SIJ and spine, detecting disease-related lesions like inflammation, erosion, and sclerosis.
• Longitudinal Validation for MRI Scoring: Validate the AI models for MRI images to track short-term changes and predict long-term disease outcomes.
• Differential Diagnosis: Develop and validate AI models that use CR and MRI data to differentiate between various forms of axial arthritis (axSpA, PsA) and other spine-related conditions (DISH and degenerative spine disease).

Work Package 6

Work Package 6 (WP6) focuses on improving precision medicine for different types of arthritis. It uses advanced imaging techniques to better predict how patients will respond to treatments. The goal is to replace invasive procedures like biopsies with non-invasive imaging methods to guide treatment decisions.
WP6 will conduct a series of studies with cutting-edge imaging techniques to make arthritis treatment more precise, reducing the need for invasive procedures while improving patient care:

The PIX-CELL study will use AI to analyse blood cells from RA patients to predict their response to specific treatments. This could help personalise therapy and improve outcomes.

The PIX-MATCH study will test whether special imaging techniques (immunoscintigraphy) using labelled medications can predict how well patients with PsA or axSpA will respond to new treatments.

The PIX-TISSUE 1 study will compare new PET/CT imaging techniques with actual joint tissue samples from RA patients to see if imaging can accurately classify different types of RA.

The PIX-TISSUE 2 study will use PET-MRI imaging to analyse inflammation and bone changes in axSpA patients. This study aims to replace biopsies with imaging as a more patient-friendly diagnostic tool.

Work Package 7

Work Package 7 (WP7) focuses on developing an advanced AI model to analyse arthritis-related medical images (X-rays, ultrasounds, and MRIs). The goal is to improve diagnosis, predict treatment responses, and create a user-friendly platform for healthcare professionals, making advanced imaging analysis more accessible to clinicians and researchers.

What will be done in WP7:

• Build an AI model for arthritis imaging. A Visual Foundation Model will be trained using a large dataset of arthritis images. This AI model will help detect and classify arthritis more accurately.
• Enhance Prediction Accuracy. The AI model will be tested to ensure it improves existing diagnostic and predictive tools from other research areas within AutoPiX.
• Identify New Arthritis Patterns with AI. AI-powered clustering techniques will be used to discover new arthritis subtypes, potentially leading to better-tailored treatments.
• Developing a Web-Based Research Tool. A web platform will be created to allow healthcare professionals to analyse medical images, generate automated reports, and translate findings into simple language for patients.

Work Package 8

Work Package 8 (WP8) focuses on maximizing the impact of AutoPiX by ensuring effective communication, engagement with key stakeholders, sustainable exploitation of results, and regulatory compliance. By integrating patient perspectives, engaging regulatory bodies, and fostering long-term collaborations, the project will create lasting value in arthritis research and AI-driven medical imaging.

Its key objectives are:

• To develop a robust communication plan to engage researchers, clinicians, industry, and patients.
• Foster long-term partnerships through co-design and shared accountability.
• Ensure effective exploitation of project results and long-term sustainability.

Its core activities will be:

• Dissemination, Communication, and Engagement, including a dedicated board, open-access scientific publications, conference presentations, multimedia content (newsletters, videos, lay summaries), or a position paper on AI in arthritis imaging.
• Project Branding and Digital Presence (project website, social media strategy, consistent visual identity to streamline communication).
• Ongoing Communication and Networking through regular meetings, synergies with related EU projects, or public awareness events.
• Exploitation and Sustainability of Results, establishing an intellectual property strategy to manage research outputs, an Innovation Board to guide the commercialisation and sustainability of AutoPiX results, and a long-term sustainability plan.
• Regulatory Guidance and Compliance, working closely with regulatory bodies (EMA, FDA) to align AI and biomarker validation with European Medical Device Regulations (MDR/IVDR).
• Education and Training, producing training materials (guides, videos, courses) in collaboration with EULAR to help clinicians and researchers use the AutoPiX platform.
• Strengthening Patient-Researcher Collaboration with a Patient Advisory Panel that will actively contribute to research, co-author scientific publications, and help develop educational materials, and a structured Patient Engagement Plan to ensure meaningful collaboration between researchers and patient communities.