Supervisors info:
Κωνσταντίνος Λουκάς, Επίκουρος Καθηγητής, Ιατρική, ΕΚΠΑ
Παναγιώτης Παπαγιάννης, Αναπληρωτής Καθηγητής, Ιατρική, ΕΚΠΑ
Παντελής Καραίσκος, Αναπληρωτής Καθηγητής, Ιατρική, ΕΚΠΑ
Summary:
Radiomics is an emerging and rapidly evolving field in quantitative imaging. It exploits the vast amount of information provided by the medical images and extracts a large set of advanced imaging traits (Radiomics Features), which are used in order to describe objectively and quantitatively the various tumor phenotypes. Many papers have been published, which indicate the positive effect of Radiomics on the detection and diagnosis of cancer, selection of the proper treatment (radiation therapy, chemotherapy, etc.), prediction of patient survival and personalized management of each patient. The sample of this study was consisted of two NSCLC patients, who received stereotactic radiation therapy, as a first line treatment. The main objective of our study, was to investigate the variation of 549 CBCT based Radiomics Features (with respect to the pre-therapy values), during the delivery period of radiation therapy.Three different groups of Radiomics Features were computed. Group 1 (Global features) consists of first order statistical features (mean, standard deviation, etc.). Group 2 (GLCM texture features) consists of second order statistical features, extracted from GLCM matrix. Group 3 (GLRLM, GLSZM and NGTDM texture features) consists of higher order statistical features, extracted from GLRLM, GLSZM and NGTDM matrices, respectively. First, Radiomics Features were computed, using the Raw Data (HU’s of tumor ROI) of CBCT examinations. At the next step, a 3D Undecimated Discrete Wavelet Transformation of Raw Data was performed and the whole process of Radiomics Features extraction, was repeated. Finally, Radiomics Features were computed, using the corresponding Saliency Map of each transversal slice of tumor ROI. Fifty (patient A) and ninety (patient B) Radiomics Features, showed constantly positive or negative relative changes (considering as reference, the initial CBCT examination ). Twenty nine (patient A) and thirty six (patient B) Radiomics Features showed statistically significant correlation with time (from the beginning of radiation fractions).We didn’t find any feature significantly correlated with time, in both patients simultaneously. A future study with sufficient sample size is neseccary, in order to confirm or reject our findings.
