Periodic Reporting for period 1 - AutoPiX (IMAGING FOR PATIENT BENEFIT IN ARTHRITIS)
Período documentado: 2024-11-01 hasta 2025-10-31
Patients at risk of or diagnosed with arthritis are routinely assessed using medical imaging techniques, such as conventional radiography, to document disease onset and progression. Despite their widespread use and high resolution, these images lack robust analysis and interpretation tools needed to deliver unbiased and precise scoring for diagnosis, monitoring, and prognosis. In addition, highly sensitive imaging modalities such as ultrasound remain largely inaccessible to many patients due to limited expertise and staff shortages. This highlights the urgent need to improve the practicality and accessibility of imaging methods.
The AutoPiX project is an ambitious international, multi-stakeholder public–private partnership jointly led by academic and industry partners. Its goal is to advance the applicability and effective use of imaging technologies by creating powerful new tools for patient care. First, we will develop analysis tools that transform unstructured images into quantitative imaging biomarkers using artificial intelligence (AI) and machine learning (ML). These tools will be clinically validated, substantially increasing the utility of imaging biomarkers for arthritis and bringing them closer to the standard of laboratory biomarkers.
In parallel, we will develop accessible imaging strategies, including remote monitoring and robotic-assisted point-of-care ultrasound examinations. These approaches will help address the shortage of qualified personnel in real-world settings while improving the standardization of ultrasound imaging. To support this, we will further enhance the precision and interpretability of these methods and validate them clinically.
The AutoPiX consortium is built on strong multidisciplinarity and close collaboration among all stakeholders in arthritis care, including rheumatologists, radiologists, patients, researchers, regulators, industry partners, and small and medium-sized enterprises. In the long term, AutoPiX will deliver clinically validated solutions that enable more diagnosis and improved assessment of treatment response and outcome prediction for patients with arthritis.
The AutoPiX project is an ambitious international, multi-stakeholder public–private partnership jointly led by academic and industry partners. Its goal is to advance the applicability and effective use of imaging technologies by creating powerful new tools for patient care. First, we will develop analysis tools that transform unstructured images into quantitative imaging biomarkers using artificial intelligence (AI) and machine learning (ML). These tools will be clinically validated, substantially increasing the utility of imaging biomarkers for arthritis and bringing them closer to the standard of laboratory biomarkers.
In parallel, we will develop accessible imaging strategies, including remote monitoring and robotic-assisted point-of-care ultrasound examinations. These approaches will help address the shortage of qualified personnel in real-world settings while improving the standardization of ultrasound imaging. To support this, we will further enhance the precision and interpretability of these methods and validate them clinically.
The AutoPiX consortium is built on strong multidisciplinarity and close collaboration among all stakeholders in arthritis care, including rheumatologists, radiologists, patients, researchers, regulators, industry partners, and small and medium-sized enterprises. In the long term, AutoPiX will deliver clinically validated solutions that enable more diagnosis and improved assessment of treatment response and outcome prediction for patients with arthritis.
During the first reporting period, AutoPiX laid the scientific and technical groundwork needed to develop and validate AI-based imaging biomarkers for arthritis across different imaging methods and clinical settings.
A key achievement was the establishment of a secure data infrastructure that complies with GDPR and supports large-scale imaging research. This included clear rules for data governance, procedures to protect patient privacy through anonymisation and pseudonymisation, and the creation of a data lake that allows imaging and clinical data from multiple centres to be stored, annotated, and analysed in a structured way. Together, these elements form the technical foundation for training and validating AI models using large, multi-centre datasets.
From a methodological standpoint, AutoPiX developed shared standards and workflows for annotating imaging data. This included an Imaging Charter that defines how images should be acquired, who is qualified to read them, and how quality control and annotations are performed. These common standards ensure consistency and reproducibility across partners and across different imaging modalities.
On the AI development side, existing models within the consortium were systematically identified and reviewed, resulting in an inventory of nine AI and machine-learning models relevant to arthritis imaging. Building on this baseline, automated image-analysis pipelines were further developed, particularly for conventional radiographs. These pipelines combine automated identification of joints with visual explanations of the results, and early internal testing showed strong agreement with expert assessments. This supports their readiness for larger clinical validation studies.
At the same time, the project progressed the technical preparation of clinical validation studies focused on early disease detection, disease monitoring, and assessment of joint damage. Study protocols were prepared, data-collection workflows were defined, and technical systems for image acquisition, storage, and integration were put in place. In addition, remote and point-of-care imaging technologies, such as mobile self-imaging, thermography, and automated ultrasound systems, were technically integrated and aligned with relevant quality and safety standards to enable future prospective studies.
Overall, during the first reporting period, the consortium initiated the development of the key scientific methods, technical platforms, and early validated tools needed to support comprehensive, multi-modal validation of imaging biomarkers in the next phases of the AutoPiX project.
A key achievement was the establishment of a secure data infrastructure that complies with GDPR and supports large-scale imaging research. This included clear rules for data governance, procedures to protect patient privacy through anonymisation and pseudonymisation, and the creation of a data lake that allows imaging and clinical data from multiple centres to be stored, annotated, and analysed in a structured way. Together, these elements form the technical foundation for training and validating AI models using large, multi-centre datasets.
From a methodological standpoint, AutoPiX developed shared standards and workflows for annotating imaging data. This included an Imaging Charter that defines how images should be acquired, who is qualified to read them, and how quality control and annotations are performed. These common standards ensure consistency and reproducibility across partners and across different imaging modalities.
On the AI development side, existing models within the consortium were systematically identified and reviewed, resulting in an inventory of nine AI and machine-learning models relevant to arthritis imaging. Building on this baseline, automated image-analysis pipelines were further developed, particularly for conventional radiographs. These pipelines combine automated identification of joints with visual explanations of the results, and early internal testing showed strong agreement with expert assessments. This supports their readiness for larger clinical validation studies.
At the same time, the project progressed the technical preparation of clinical validation studies focused on early disease detection, disease monitoring, and assessment of joint damage. Study protocols were prepared, data-collection workflows were defined, and technical systems for image acquisition, storage, and integration were put in place. In addition, remote and point-of-care imaging technologies, such as mobile self-imaging, thermography, and automated ultrasound systems, were technically integrated and aligned with relevant quality and safety standards to enable future prospective studies.
Overall, during the first reporting period, the consortium initiated the development of the key scientific methods, technical platforms, and early validated tools needed to support comprehensive, multi-modal validation of imaging biomarkers in the next phases of the AutoPiX project.
The core principle of AutoPiX is to advance imaging biomarkers so they can be used in routine clinical care and in clinical studies. At present, imaging assessments are largely limited to narrative reports. Automated quantification of images through standardized scoring therefore already represents a significant advance beyond the current state of the art. In addition, the development of an imaging platform that is not limited to a single imaging modality is not common practice today. The data-sharing platform used in AutoPiX, based on a data lake architecture, will host diverse data sources and integrate imaging data with clinical information.