Periodic Reporting for period 2 - PRIMAGE (PRedictive In-silico Multiscale Analytics to support cancer personalized diaGnosis and prognosis, Empowered by imaging biomarkers)
Periodo di rendicontazione: 2020-06-01 al 2021-11-30
The digital transformation of healthcare systems has fostered innovative clinical workflows and quality improvements. Several digital diagnosis tools (such as imaging, pathology, genomic analytics, wearable sensors) and patient electronic records are key enabling factors improving routine clinical practice. This change is expected to promote clinical innovation models via real-world data-driven inferences revealing insights implicit in the data. Real-world evidence can help answering existing questions and generating new knowledge in a more reproducible way. The development of predictive models able to predict valid disease-related outcomes by learning from retrospective data, taking into account all types of clinical, pathology, molecular and imaging information, is a hot topic of scientific debate.
The PRIMAGE project focuses on the further development of in silico tools for a more personalized clinical management of cancer by targeting clinical endpoints, with two pilots focused on childhood cancer. Mathematical and computational modelling of biological processes will be used to enhance quantitative understanding of biomedical phenomena, such as cancer progression, potentially incorporating patient-specific data to enrich the scope of therapeutic target identification.
Moreover, since cancer patients are frequently monitored for staging and treatment response follow-up, oncologic imaging represents a suitable field for the discovery and validation of new biomarkers from different imaging modalities.
In this sense, PRIMAGE project aims at developing an open hybrid cloud and high performance computing (HPC) platform with later implementation and validation in non-interventional trials, which will support decision making in the clinical management of two paediatric malignant solid tumours, neuroblastoma (NB) and the diffuse intrinsic pontine glioma (DIPG). The project utilises novel HPC approaches to provide computationally efficient and large scale in silico models resulting in a decision support system (DSS), which will provide improved health outcomes. Since the proposed methodologies for data management, in silico models and visualization tools, will be available to be transferred to other cancer types, it is expected PRIMAGE results have a great impact not only in NB and DIPG but also in other malignant solid tumours.
The PRIMAGE project focuses on the further development of in silico tools for a more personalized clinical management of cancer by targeting clinical endpoints, with two pilots focused on childhood cancer. Mathematical and computational modelling of biological processes will be used to enhance quantitative understanding of biomedical phenomena, such as cancer progression, potentially incorporating patient-specific data to enrich the scope of therapeutic target identification.
Moreover, since cancer patients are frequently monitored for staging and treatment response follow-up, oncologic imaging represents a suitable field for the discovery and validation of new biomarkers from different imaging modalities.
In this sense, PRIMAGE project aims at developing an open hybrid cloud and high performance computing (HPC) platform with later implementation and validation in non-interventional trials, which will support decision making in the clinical management of two paediatric malignant solid tumours, neuroblastoma (NB) and the diffuse intrinsic pontine glioma (DIPG). The project utilises novel HPC approaches to provide computationally efficient and large scale in silico models resulting in a decision support system (DSS), which will provide improved health outcomes. Since the proposed methodologies for data management, in silico models and visualization tools, will be available to be transferred to other cancer types, it is expected PRIMAGE results have a great impact not only in NB and DIPG but also in other malignant solid tumours.
During the first reporting period, the bases of the different components of the PRIMAGE Decision Support System (DSS) were established using a user-centric approach and clinical partners started coordinating efforts with technical partners to allow data extraction from hospital databases. Strategies for data pseudonymization, extraction, curation and transfer to the repository were established. In parallel, technical partners worked on the following aspects:
- Development of the platform’s computing infrastructure: An efficient infrastructure with storage and computing resources was developed, along with the necessary middleware to manage it. A pre-production open cloud infrastructure was deployed.
- Extraction and validation of imaging biomarkers: Reproducible imaging data was generated in a standardized format, designing a common framework to assess the quality and standardization of multi-site and multi-vendor imaging data. The image processing pipeline was optimized to address image motion and noise and intensity inhomogeneity among others. Quantitative imaging descriptors were extracted and AI tools were explored. AI models were developed for the prediction of different clinical outcomes, integrating molecular and genomic biomarkers with imaging and clinical data.
- Multiscale models: In-silico models of solid tumor growth for NB and DIPG were developed and integrated in a multiscale scheme to estimate tumor growth.
- Development of Visual Analytic tools: Visualization methods for high-dimensional data analysis were delivered, and AI methodologies were implemented to generate outcomes to the relevant CEPs for NB and DIPG.
- PRIMAGE cloud-based platform: A functional prototype of the platform was developed and deployed, and tested.
Over the second reporting period, clinical efforts focused on data collection, both from consortium partners and from external collaborator centers for the collection of the validation dataset. This led to a total number of 408 and 54 new cases of NB and DIPG created in the platform, and 12 institutions becoming official external collaborators to the project. Technical partners, in parallel, achieved the following:
- Computing infrastructures: The back-end infrastructure was set-up and validated. The Cloud production infrastructure was developed. Three computational Grants on Prometheus HPC were negotiated.
- Data repositories: The data ingestion processes were deployed in the platform, and an MR series classifier for NB and DIPG was developed.
- Imaging biomarkers: The relationship of the MYCN gene with image-based data was elucidated. A pre-processing and registration pipeline was built and validated for both cancers, and manual and automatic segmentation methods were developed.
- Multiscale models: Four new in-silico models were developed. In parallel 2 in vitro experiments with NB and DIPG cells were performed. Lastly, USFD’s cell-based NB model was refined, calibrated and integrated into the orchestrated model.
- Visual analytic tools: The interactive visual analytics framework was finalized and integrated in the production environment.
- PRIMAGE could-platform: New functionalities were incorporated to the platform, adjusting to current project needs. An advanced visualization environment was integrated in the infrastructure.
- Development of the platform’s computing infrastructure: An efficient infrastructure with storage and computing resources was developed, along with the necessary middleware to manage it. A pre-production open cloud infrastructure was deployed.
- Extraction and validation of imaging biomarkers: Reproducible imaging data was generated in a standardized format, designing a common framework to assess the quality and standardization of multi-site and multi-vendor imaging data. The image processing pipeline was optimized to address image motion and noise and intensity inhomogeneity among others. Quantitative imaging descriptors were extracted and AI tools were explored. AI models were developed for the prediction of different clinical outcomes, integrating molecular and genomic biomarkers with imaging and clinical data.
- Multiscale models: In-silico models of solid tumor growth for NB and DIPG were developed and integrated in a multiscale scheme to estimate tumor growth.
- Development of Visual Analytic tools: Visualization methods for high-dimensional data analysis were delivered, and AI methodologies were implemented to generate outcomes to the relevant CEPs for NB and DIPG.
- PRIMAGE cloud-based platform: A functional prototype of the platform was developed and deployed, and tested.
Over the second reporting period, clinical efforts focused on data collection, both from consortium partners and from external collaborator centers for the collection of the validation dataset. This led to a total number of 408 and 54 new cases of NB and DIPG created in the platform, and 12 institutions becoming official external collaborators to the project. Technical partners, in parallel, achieved the following:
- Computing infrastructures: The back-end infrastructure was set-up and validated. The Cloud production infrastructure was developed. Three computational Grants on Prometheus HPC were negotiated.
- Data repositories: The data ingestion processes were deployed in the platform, and an MR series classifier for NB and DIPG was developed.
- Imaging biomarkers: The relationship of the MYCN gene with image-based data was elucidated. A pre-processing and registration pipeline was built and validated for both cancers, and manual and automatic segmentation methods were developed.
- Multiscale models: Four new in-silico models were developed. In parallel 2 in vitro experiments with NB and DIPG cells were performed. Lastly, USFD’s cell-based NB model was refined, calibrated and integrated into the orchestrated model.
- Visual analytic tools: The interactive visual analytics framework was finalized and integrated in the production environment.
- PRIMAGE could-platform: New functionalities were incorporated to the platform, adjusting to current project needs. An advanced visualization environment was integrated in the infrastructure.
PRIMAGE is facing different technological challenges during its development, which may have a great impact in the translation of Big & Multi-disciplinary Data into predictors for personalised decision making, the improvement of data quality, the interoperability and standards, within others. Some of these challenges are the development of:
- Open cloud infrastructures for healthcare, ensuring privacy management, transparency, liability and long-term sustainability.
- A Hybrid Cloud Model combining restricted sites within healthcare networks processing the sensitive data and external clouds that only deal with anonymised data.
- Reproducible imaging data in a standardized format to provide large sample size statistics and generalizable trends.
- A multi-step image preprocessing pipeline for the image preparation addressing the challenges of medical image analysis.
- Automatic segmentation algorithms based on encoder-decoder CNNs and the addition of multi-sequence information.
- An innovative Fit-Cluster-Fit methodology for the selection of more homogeneous and representative clusters as habitats.
- A multiscale hybrid approach that combines both stochastic and continuum strategies in order to simulate NB and DIPG tumour growth, vascularization, residual stresses and extracellular matrix stiffening.
- A DSS offering visual analytics and predictive tools to assist management of NB and DIPG paediatric cancers based on the use of novel imaging biomarkers, tumour growth models, advanced visualisation of predictions with weighted confidence scores and responses to a set of CEPs.
- Open cloud infrastructures for healthcare, ensuring privacy management, transparency, liability and long-term sustainability.
- A Hybrid Cloud Model combining restricted sites within healthcare networks processing the sensitive data and external clouds that only deal with anonymised data.
- Reproducible imaging data in a standardized format to provide large sample size statistics and generalizable trends.
- A multi-step image preprocessing pipeline for the image preparation addressing the challenges of medical image analysis.
- Automatic segmentation algorithms based on encoder-decoder CNNs and the addition of multi-sequence information.
- An innovative Fit-Cluster-Fit methodology for the selection of more homogeneous and representative clusters as habitats.
- A multiscale hybrid approach that combines both stochastic and continuum strategies in order to simulate NB and DIPG tumour growth, vascularization, residual stresses and extracellular matrix stiffening.
- A DSS offering visual analytics and predictive tools to assist management of NB and DIPG paediatric cancers based on the use of novel imaging biomarkers, tumour growth models, advanced visualisation of predictions with weighted confidence scores and responses to a set of CEPs.