Radiomic and anthropometric analyses in multimodal imaging for exploring predictive and prognostic factors in oncology

Personalized medicine refers to the adaptation of medical treatment to the individual characteristics of each disease and each patient. In oncology, this adaptation depends in particular on the prognosis of the disease in order to adapt a treatment to the severity of the cancer. Medical imaging, represented in particular by the CT scan and the PET scan, allows the extraction and analysis of morphological and functional characteristics of the cancer, called 'radiomics', but also of the patient's characteristics, called 'anthropometry'. The objective of this work was to explore the use of radiomic and anthropometric parameters in multimodal imaging for the exploration of prognostic factors in oncology. For the anthropometric side, we developed and evaluated a software, named 'Anthropometer3D', allowing the measurement of muscle, lean, fat, visceral adipose tissue and subcutaneous adipose tissue masses in a multi-slice and automatic way from PET/CT scans and showed the prognostic value of subcutaneous adipose tissue for stage IV lung cancers treated with immunotherapy and of muscle mass for lung cancers treated with radio-chemotherapy For the radiomic side, we have developed and evaluated algorithms for the automatic segmentation and classification of oncological PET/CT scans and developed a software, named 'Oncometer3D', for the extraction of tumor activity, fragmentation, dispersion and massiveness characteristics from PET scans. We showed that one fragmentation parameter, the volume to total tumor area ratio, was an independent prognostic factor in diffuse large-cell B-cell lymphoma and that several morphological tumor parameters correlated with circulating tumor DNA in B-cell hemopathies. In conclusion, medical imaging participates in the global evaluation of cancer from a macroscopic point of view by allowing a radiomic analysis centered on the tumor but also anthropometrically centered on the patient. Improved prognostication using these two approaches could lead to better therapeutic management of patients. (author)