Research

Research Highlights

RESEARCH THEME AND MOTIVATION

The overarching theme of our research is to identify radiomic (extracting computerized sub-visual features from radiologic imaging that are visually not discernible to an expert reader), radio-genomic (identifying radiologic features associated with molecular phenotypes), and radio-pathomic (radiologic features associated with pathologic phenotypes) features that can provide insight into the underlying tumor biology across different disease conditions as reflected on radiologic imaging. Current projects involve quantification of morphological and functional neuroimaging attributes via radiomics, radio-genomics, and radio-pathomics associations, to address questions such as: What to treat? Which treatment? Did the treatment work?

BrIC logo with images representing multimodal data fusion and outcome prediction, image-based morphometry and prognosis, neuro-informatics and computer aided diagnosis, and treatment evaluation and response.

DIAGNOSTIC MODELING: Diagnosis of true-progression from treatment effects in brain tumors

Radiation necrosis vs. tumor recurrence using radiomics on routine MRI

One of the most challenging problems in management of cancer patients (especially in brain tumors) is to distinguish pseudo-progression or radiation necrosis (benign conditions caused due to radiation), affecting 20-40% of all brain tumor patients, from recurrent brain tumors (true progression). Currently, invasive biopsy offers the only reliable diagnosis, since these treatment effects and true progression have a similar visual appearance on routine MRI. We have developed a computerized decision support software toolkit, NeuroRadVisionTM, employing novel radiomic tools that in conjunction with multi-parametric-MRI enable discrimination of true progression from benign treatment effects. NeuroRadVisionTM has so far yielded 89% accuracy, while currently the diagnostic accuracy via visual inspection by radiologists (standard-of-care) on MRI is only 50-65% at best.

PROGNOSTIC MODELING: Identifying prognostic markers for survival prediction on pre-treatment MRI

Survival prediction in Glioblastoma patients using radiomics analysis

Despite the variability in clinical responses, the majority of cancer patients are presently treated in a uniform, standardized way, following a ‘one-fits-all’ therapeutic approach, regardless of the individual characteristics of each tumor that most likely affect patient prognosis. The identification of accurate prognostic factors on pre-treatment imaging can serve as clinical indicators for tailoring personalized therapy for individual cancer patients. With our clinical collaborators, we have been exploring the role of radiomic features in capturing subtle, morphologic attributes relating to the extent of heterogeneity of tumor/edema in brain tumor patients on pre-treatment MRI. Our preliminary analysis suggests that radiomic features have the potential to provide prognostic information regarding patients with short-term (<6-months) from prolonged survival (>2-years) and can assist in making personalized therapy decisions based on patient’s survival characteristics on pre-treatment MRI.

TREATMENT EVALUATION: Evaluating early response to treatment of neurological disorders

TREATMENT EVALUATION: Evaluating early response to treatment of neurological disordersCurrently post-treatment changes for evaluating patient’s response to treatment are monitored qualitatively via comparing volumetric changes of contrast enhancement on T1w MRI protocol acquired for follow-up (24-hours, 1-month, 3-months, 6-months post-treatment), with reference to pre-treatment T1w MRI (known as Recist criterion). Volumetric analysis provides a single global measurement of morphologic volume differences; and may not capture the local treatment effects (such as swelling, tissue necrosis) introduced due to treatment on and around the lesion. We have been working on developing a radiomic framework to evaluate early treatment related changes that may be better reflected via a voxel-by-voxel analysis of changes in MRI markers monitored over time. These per-voxel changes over time may provide cues about treatment-failure and serve as surrogate markers for evaluating patient’s response to treatment.