Using artificial intelligence and MRI to address limitations in glioblastoma

Division Of Basic Sciences - Nci

Investigators

Linked publications & trials

Abstract

We hypothesize that MRI images of patients with glioma when subjected to change over time analysis (at diagnosis, prior to and post radiation therapy) can identify features predictive of treatment failure helping guide patient management in the clinic. Successful algorithms can be validated using large scale publicly available data sets fostering the secondary advancement of currently lacking ground truth data sets for glioma progression. Our methodology will include MRI (Magnetic Resonance Imaging) of the brain using standard of care sequences, generally carried out at most centers will initially be employed to ensure transferrable findings. However, we would like to identify and test additional features including ADC and CBV maps that may enhance the predictive ability of algorithms to identify true progression vs. pseudoprogression acknowledging that these may not be consistently carried out outside of centers of excellence secondary to limitations in resources and expertise. The patient cohort initially employed consists of glioblastoma patients treated on IRB approved ROB protocols at NCI NIH for whom biospecimens for correlative omic analysis are available.