Periodic Reporting for period 1 - ARTEMIS (AcceleRating the Translation of virtual twins towards a pErsonalised Management of fatty lIver patients)
Período documentado: 2024-01-01 hasta 2025-06-30
The ARTEMIs project is motivated by the urgent need to address the global epidemic of Metabolic dysfunction-associated steatotic liver disease (MASLD). This condition, marked by fat accumulation in the liver, is the leading cause of chronic liver disease in Europe, with a prevalence of over 25%. Its prevalence is rising due to increased rates of associated metabolic disorders like obesity, type 2 diabetes, and high blood pressure, which are caused by unhealthy lifestyles. While disease progression is slow in most patients, an estimated 10% develop advanced fibrosis, risking complications like cirrhosis. Cirrhosis is a non-curable disease that significantly contributes to morbidity and mortality in Europe and is a major risk factor for hepatocellular carcinoma (HCC). MASLD is also linked to an increased risk of cardiovascular diseases, which are the primary cause of death in MASLD patients. Effective management is currently hindered by a limited understanding of MAsLD progression and the complex interactions of the "liver-heart axis".
The project's main aim is to tackle these challenges by proposing a new method to control the MAsLD epidemic using advanced "virtual twin" liver-heart models. Building on previous successful work with liver computational models, the primary goal is to integrate existing multilevel computational models and expand them into a multi-organ approach that includes the heart and circulatory systems. This will be achieved using both mechanistic and machine-learning techniques. The models will be validated in clinical practice through Proof of Concept demonstrators to provide meaningful clinical insights to healthcare professionals.
To achieve this, ARTEMIs must address several key challenges. The models must be clinically relevant and integrated into understandable decision support tools to deliver actionable, personalized insights to clinicians. This requires virtual twins capable of integrating models across multiple scales (molecular, cellular, tissue, organ) and systems to simulate disease progression and predict comorbidities dynamically. Access to a large, high-quality set of multimodal data from diverse patient populations is essential, while adhering to data protection and sovereignty regulations. It is also critical to maintain clinicians' trust in the models' accuracy. Finally, to ensure acceptance and market readiness, the project involves clinical Key Opinion Leaders (KOLs), patients, and industrialists to address clinical relevance, regulatory aspects, and interoperability.
The project's main aim is to tackle these challenges by proposing a new method to control the MAsLD epidemic using advanced "virtual twin" liver-heart models. Building on previous successful work with liver computational models, the primary goal is to integrate existing multilevel computational models and expand them into a multi-organ approach that includes the heart and circulatory systems. This will be achieved using both mechanistic and machine-learning techniques. The models will be validated in clinical practice through Proof of Concept demonstrators to provide meaningful clinical insights to healthcare professionals.
To achieve this, ARTEMIs must address several key challenges. The models must be clinically relevant and integrated into understandable decision support tools to deliver actionable, personalized insights to clinicians. This requires virtual twins capable of integrating models across multiple scales (molecular, cellular, tissue, organ) and systems to simulate disease progression and predict comorbidities dynamically. Access to a large, high-quality set of multimodal data from diverse patient populations is essential, while adhering to data protection and sovereignty regulations. It is also critical to maintain clinicians' trust in the models' accuracy. Finally, to ensure acceptance and market readiness, the project involves clinical Key Opinion Leaders (KOLs), patients, and industrialists to address clinical relevance, regulatory aspects, and interoperability.
The retrospective study protocol was finalised. A complete list of clinical variables was defined for each use case, and an inventory was created of all available data at clinical sites. T
A solution for clinical data collection and harmonisation is fully implemented for all use cases. The architecture of the ARTEMIS platform has been fully defined and described, and on-site infrastructure setup has been partially completed.
Customised tools were developed for segmenting and quantifying large-scale structures and for identifying precancerous and cancerous regions. A pipeline was also developed to isolate liver vessels for image analysis. Imaging biomarkers, quantitative features, and radiomics features were selected for the ARTEMIs platform. Additionally, domain adaptation and continual deep learning approaches were evaluated for transferring image analysis algorithms across sites and managing scanner variability.
The automated sample preparation pipeline for liquid and organ biopsies was completed, and the proteomic characterisation of 62 liver tissue samples was finalised. A foundational, unsupervised, multi-modal deep federated clustering pipeline was successfully established as a crucial baseline, and omics data exploration was carried out to identify relevant subpopulations and biomarkers for MASH disease.
For modelling and simulation, three established intracellular dynamic models relevant for ARTEMIS were selected. A blood flow model was extended to consider the elasticity of blood vessels, which is impaired by disease progression. A first tissue model for disease progression from fatty liver to fibrosis was also developed. Model components for clinical use cases in flow and transport were established, including the first-time segmentation and modelling of natural shunts in the liver.
Finally, user needs and expectations for the user experience with the Clinical Decision Support System (CDSS) were identified. The first version of the UI design and the CDSS's technical and legal specifications were delivered. A preliminary methodology for prioritising Key Exploitable Results (KERs) was established, and an initial analysis for the sustainability strategy was carried out.
A solution for clinical data collection and harmonisation is fully implemented for all use cases. The architecture of the ARTEMIS platform has been fully defined and described, and on-site infrastructure setup has been partially completed.
Customised tools were developed for segmenting and quantifying large-scale structures and for identifying precancerous and cancerous regions. A pipeline was also developed to isolate liver vessels for image analysis. Imaging biomarkers, quantitative features, and radiomics features were selected for the ARTEMIs platform. Additionally, domain adaptation and continual deep learning approaches were evaluated for transferring image analysis algorithms across sites and managing scanner variability.
The automated sample preparation pipeline for liquid and organ biopsies was completed, and the proteomic characterisation of 62 liver tissue samples was finalised. A foundational, unsupervised, multi-modal deep federated clustering pipeline was successfully established as a crucial baseline, and omics data exploration was carried out to identify relevant subpopulations and biomarkers for MASH disease.
For modelling and simulation, three established intracellular dynamic models relevant for ARTEMIS were selected. A blood flow model was extended to consider the elasticity of blood vessels, which is impaired by disease progression. A first tissue model for disease progression from fatty liver to fibrosis was also developed. Model components for clinical use cases in flow and transport were established, including the first-time segmentation and modelling of natural shunts in the liver.
Finally, user needs and expectations for the user experience with the Clinical Decision Support System (CDSS) were identified. The first version of the UI design and the CDSS's technical and legal specifications were delivered. A preliminary methodology for prioritising Key Exploitable Results (KERs) was established, and an initial analysis for the sustainability strategy was carried out.
It is foreseen that ARTEMIs will have the following Key Exploitable Results:
-Processing pipelines for the extraction of quantitative metrics as inputs to the models
-Multi-modal clustering of patients and pattern identification
-Multiorgan Multiscale Multilevel model of liver-heart axis for MAFLD -Virtual Twin models
-Advanced visualisation tool for multimodal data
-ARTEMIs CDSS
-Federated data exploitation infrastructure for some 7500 patients, along the MAFLD spectrum
-Methodology for model performance, trustworthiness, utility and usability assessment
That will have the following impacts:
▪ Scientific impact: Advancement in translational research in the field of Virtual Twins for patient management, in the field of liver diseases and their cardiac complications.
▪ Social impact: Assisting to the achievement of better health outcomes in MAFLD patients at different stages, through personalised management, and contributing to leveraging acceptance for computer-based aid systems.
▪ Economic impact: Contributing to more cost-effective patient management approaches. Accelerating market uptake of Virtual Twins applied to patient care.
-Processing pipelines for the extraction of quantitative metrics as inputs to the models
-Multi-modal clustering of patients and pattern identification
-Multiorgan Multiscale Multilevel model of liver-heart axis for MAFLD -Virtual Twin models
-Advanced visualisation tool for multimodal data
-ARTEMIs CDSS
-Federated data exploitation infrastructure for some 7500 patients, along the MAFLD spectrum
-Methodology for model performance, trustworthiness, utility and usability assessment
That will have the following impacts:
▪ Scientific impact: Advancement in translational research in the field of Virtual Twins for patient management, in the field of liver diseases and their cardiac complications.
▪ Social impact: Assisting to the achievement of better health outcomes in MAFLD patients at different stages, through personalised management, and contributing to leveraging acceptance for computer-based aid systems.
▪ Economic impact: Contributing to more cost-effective patient management approaches. Accelerating market uptake of Virtual Twins applied to patient care.