Country / Region
EMEA
Tags
Artificial intelligence, Clinical Practice, Research
Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous cancer, with 30 to 40% of patients experiencing early relapse or being refractory to standard treatment. This project, conducted within the artificial intelligence laboratory (LAB-AI), aims to structure the care pathways of DLBCL patients to train AI algorithms for the early detection of primary refractory disease. Semantic quality, interoperability, and structured clinical data are essential, with SNOMED CT playing a central role. This standardized terminology enables precise representation of medical concepts, supporting integration into relational databases and FHIR resources, and ensuring reliable use by AI. SNOMED CT also fosters transparency, traceability, and explainability (e.g., with LIME), aligning the project with TRIPOD+AI guidelines. Moreover, it supports large-scale interoperability, which is key to advancing precision medicine through standardized, reliable, and explainable data.
Description
Diffuse large B-cell lymphoma (DLBCL) is the most common type of non-Hodgkin lymphoma in adults and is characterized by significant clinical and biological heterogeneity. Despite a standardized first-line treatment, 30 to 40% of patients experience early relapse or are refractory to treatment, resulting in a poor prognosis. The development of predictive tools based on artificial intelligence (AI) to rapidly identify these cases is a major challenge in personalized medicine, especially given the effectiveness of second-line immunotherapy.
This project, conducted as part of an internship within the artificial intelligence laboratory (LAB-AI), aims to structure the care pathways of patients with DLBCL in order to train algorithms for the early detection of primary refractory disease to chemotherapy. To achieve this, structuring, interoperability, and semantic quality of clinical data are essential components.
Scope
SNOMED CT was chosen because it provides precise and unambiguous coding of clinical concepts. It ensures semantic interoperability across different data sources like EHRs, labs, and imaging systems. Its compatibility with FHIR makes data integration easier and more standardized. SNOMED CT also supports traceability and explainability, which are essential for ethical AI development. Finally, it is internationally recognized and widely adopted, making it a sustainable choice for large-scale and long-term projects.
How SNOMED CT will be used
In this project, SNOMED CT is used as a standardized clinical terminology to structure and harmonize data from various sources, such as electronic health records, registries, laboratories, and imaging systems. It enables the precise and unambiguous representation of key medical concepts, including patient characteristics, features of DLBCL, adverse events, treatment responses, and follow-up information. By encoding each data point with a unique SNOMED CT concept, the project ensures semantic interoperability and facilitates the integration of data into FHIR resources and relational databases. This standardized approach not only supports the reliable use of data by AI algorithms but also enhances transparency and traceability, which are essential for explainability using tools like LIME. Furthermore, the use of SNOMED CT aligns the project with international recommendations such as TRIPOD+AI, promoting ethical, reproducible, and robust predictive modeling. Finally, it enables large-scale data sharing and interoperability, paving the way for broader applications in precision medicine.
Why SNOMED CT will be used
Contact
