About the project
RACOON CORE-PE is a clinical, multicentre, retrospective study. The RACOON CORE-PE project aims to improve the prognosis and treatment decisions for patients with acute pulmonary embolism. This is to be achieved through AI-supported models that combine comprehensive image analyses with laboratory values and clinical information.
► more
The aim of RACOON CORE-PE is to develop and validate an AI-driven assistance module to improve prognosis and treatment decisions in patients with acute pulmonary embolism. This is to be achieved through advanced image analysis and the integration of imaging biomarkers with laboratory and clinical data. The development of this module could subsequently be used for other medical applications and is therefore highly relevant.
The following specific objectives of the CORE-PE project result from this general objective:
- Development of standardised procedures for assessing the prognosis of patients with acute pulmonary embolism
- Development of a large data set of CT studies in acute pulmonary embolism, including clinical, laboratory and outcome data
- Creation of an AI-capable structured database
- Development of AI algorithms for the fully automated detection of prognostic quantitative imaging biomarkers from CT images in acute pulmonary embolism.
- Development of AI-driven risk prediction models and identification of new image-based prognostic biomarkers in acute pulmonary embolism.
- Conducting a thorough validation of all developed AI models
The RACOON CORE-PE study involves close collaboration between 11 university sites (radiology and cardiology), site IT specialists and established technology partners. The aim is to develop an AI-based procedure for standardised risk stratification in pulmonary embolism patients. The following eight work packages are planned to achieve this goal
- Coordination of the overall project, provision of the necessary infrastructure and regulations
- Definition and standardisation of input and outcome variables and definition of the patient cohort
- Data selection and import of all imaging and non-imaging data into the RACOON infrastructure
- Manual segmentation and annotation of CT datasets to create a reference dataset
- Development of AI models for automatic image segmentation and annotation
- Development of comprehensive AI-driven models for risk stratification in acute LAE, incorporating imaging biomarkers as well as clinical and laboratory parameters
- Quality control and validation of the AI models
- Dissemination and utilisation of the results as well as the introduction of algorithms in the form of software tools for further research and clinical use within NUM