Abstract
419
Objectives Despite significant efforts in developing therapies, median survival from glioblastoma is about 14 months, and thus new therapies are needed. We present the results of a patient-specific bio-mathematical model that predicts glioma growth and is coupled to the appearance of PET hypoxia marker (FMISO) images.
Methods Our bio-mathematical model characterizes gliomas into distinct cellular phenotypes including normoxic, hypoxic and necrotic cells that interact with themselves and the microenvironment (vasculature). Applying this model to specific glioma patients, we simulated spatial maps of tumor hypoxia to allow direct comparison with patient FMISO-PET images. We applied a pharmacokinetic model to the patient-specific hypoxic maps to capture the FMISO tracer activity. The resultant FMISO tracer activity maps were processed through a simulator for the PET image acquisition and reconstruction. This analysis was performed on 3 newly diagnosed glioblastoma test cases.
Results Visual appearances were qualitatively closely coupled between the patient and simulated FMISO-PET images. Quantitative spatial comparisons indicate that the distribution of pixel intensities were not statistically different between the simulated and actual images (97.5% certainty, Kolmogorov-Smirnov test).
Conclusions Patient-specific mathematical modeling tools can provide a quantitative dynamic understanding of the biological connection between anatomical changes seen on MRI and biochemical activity seen on PET of gliomas in vivo and may be utilized to investigate the degree to which PET imaging can detect clinically relevant phenotypic and histologic characteristics. Furthermore this predictive ability may improve glioma treatments by providing quantitative information on the spread of microscopic disease in individual patients.
Research Support James S. McDonnell Foundation, the National Institutes of Health (R01NS060752, U54CA143970), the Academic Pathology Fund at the University of Washingto