Up to one third of diffuse large B-cell lymphoma (DLBCL) patients experience relapse or fail to achieve complete remission during first-line treatment. To improve outcome, early identification of patients at risk of treatment failure is of paramount importance.

Identification of patients with poor prognosis might be further improved by radiomics. Radiography uses ionizing radiation (i.e. X-rays) and non-ionizing radiation (i.e. ultrasound, MRI) to view the internal form of an object. By using artificial intelligence, radiomics technology extracts a large number of features from radiographic images. In these large datasets, machine learning is employed to search for distinct cancer characteristics which a human observer may fail to recognize. Radiomics analysis of medical scans provides quantifiable features of tumor characteristics such as intensity, shape, volume, texture, and intra-and inter-lesion heterogeneity. The aim of this project is to identify and validate radiomics features, both at patient- and cancer level, that predict treatment response and to compare these results to currently used prognostic markers.

For scientific information about this project, please contact Ronald Boellaard:

Radiomics pipeline – automated delineation of lesions on FDG PET/CT followed with feature extraction providing information on shape, size, tracer uptake distribution and texture and, finally, use of these features as input for prediction with a model

Radiomics pipeline – Left: Metabolic active cancer cells in the patient are captured by positron emission tomography (PET). Middle: Automated processing and analysis of radiographic medical images by radiomics and machine learning. Right:  Modeling and prediction of disease outcome.

Researchers involved

PhD students/Research assistent: J. Eertink, S. Wiegers