Factor analysis-based approach for early uptake automatic quantification of breast cancer by ~(18)F-FDG PET images sequence

摘要

Factor Analysis of Medical Image Sequences (FAMIS) is recognized as one pioneer successfully used approach for analyzing especially dynamic images' sequence for estimating kinetics and associated compartments having a physiological meaning. Some studies tried to extend the exploring of this approach to analyze Positron Emission Tomography (PET) image modality for dynamic sequences. PET images with ~(18)F-fluorodesoxyglucose (~(18)F-FDG) is the gold standard for in vivo, evaluation of tumor glucose metabolism and is widely used in clinical oncology. In this paper, a novel approach is proposed to obtain an automated quantification method for early accumulation of ~(18)F-FDG tracer in order to explore breast cancer, by applying FAMIS tool on dynamic first pass ~(18)F-FDG PET dynamic sequences. This approach starts by an automated identification of a tumor Region of Interest (ROI) from PET dynamic images' sequence. Then, a FAMIS approach is applied to separate two compartments: one compartment is associated to the vascular and a second one is associated to the purely tumor compartment. The latter allows the evaluation of the temporal evolution of the glucose tracer metabolism and therefore for pursuing cancer characterization. A new empiric parameter K_FPQ (First Pass Quantification), computed from the evolution of the ~(18)F-FDG radiotracer accumulation using the first 11 min PET early images, is proposed. This parameter is found to be correlated to standardized uptake value maximal index (SUV_max) metabolism tumor. The proposed framework is tested using image sequences' database for 25 different pathology cases, which is considered as largely sufficient by the clinical team. Among clinicians' experience, using a large dataset permits the possibility to obtain accurate information and precise early diagnosis. Pearson correlation coefficient is computed to evaluate as well as to analyze the relationship between the proposed empiric parameter K_FPQ and glucose tracer metabolism SUV_max for the overall pathology cases. K_FPQ is successfully evaluated by the dynamic first-pass ~(18)F-FDG PET image sequences for exploring early breast cancer diagnosis. Quantitative evaluations, as discussed and validated by clinicians, confirmed the efficiency of the modeling and the usefulness of the new empiric parameter K_FPQ to predict tumor glucose metabolism for early uptake. This can be considered as a significant indication for quantification as well as evaluation of early relapse and disease progression during the therapy.