Research Projects
Image-Localized Biopsy Collection and Analysis
I lead the collection of image-localized biopsies in the operating room for our brain tumor biopsy protocol, having attended over 100 surgeries to date and collected over 500 samples. This successful project has so far consented over 220 patients with over 1000 biopsies collected so far. This project aims to connect heterogeneous biopsies with their local presentation on standard and advanced imaging, to ultimately make maps of key tumor features such as cell density and genetic phenotype. This work has given me insight into clinical workflows and how to seamlessly connect research with clinical care. See our recently published paper on this workflow here.
I also work further down this pipeline, with data analysis of whole exome sequencing and bulk RNA-Seq. We're currently using a deconvolution method alongside a single cell reference to determine the cellular population makeup of our bulk samples. We can then test the abundance of these populations against survival metrics and between imaging regions.
Pretreatment Cysts in Glioblastoma
I led a project using retrospective patient data alongside pretreatment imaging to determine the survival impact of cysts in patients with glioblastoma (e.g. subpanel A). This dataset provided evidence for a prolonged overall survival for patients with cystic glioblastoma, that didn't show significant impact of the introduction of the current standard of care. This work may provide a link between older papers that saw a survival impact of cystic glioblastoma against more recent work that didn't, after the standard of care was introduced. This survival benefit also only appeared significant in male patients, hinting at possible sex differences.
MRI Abnormality Shape at Pretreatment
I have led another project based on lacunarity and fractal dimension - quantitative values describing heterogeneity and connectedness of shapes - to show that they can have prognostic impact when applied to pretreatment standard-of-care MRI.
This work was inspired by Liu et al. who showed lacunarity and fractal dimension of necrotic regions had prognostic impact. I extended on this methodology to include the total abnormality present on T1Gd MRI and T2/FLAIR MRI on a different dataset, and include a p-value adjustment method for finding optimal cutoffs that distingush survival. I found that large lacunarity values and smaller fractal dimension values in T2 regions were associated with shorter survival, an effect that was independent of both the abnormality size and age of the patient.
The figure shows some significant Kaplan-Meier curves, and two example patients with their lacunarity and fractal dimension values.
Modeling Impact of Ischemia on Tumor Recurrence
Inspired by a paper on distal glioblastoma recurrence following perioperative ischemia (i.e. tumors recurred far away or more spread out following a cutoff in blood supply), we wanted to test if the Proliferation Invasion Hypoxia Necrosis Angiogenesis model could recapitulate this behavior.
By modeling two different scenarios of perioperative ischemia, we saw that distal recurrence was more likely to occur for tumors with high cellular migration relative to proliferation rates. The mechanism of action is that tumors with these simulated characteristics were more likely to grow through the ischemic region at MRI-undetectable densities.
Future work in this space may change the type of model used, to provide the tumor cells with a memory of nutrient-poor environments they have been in.
