Abstract 9
Category: Basic Science
At the end of the session, participants will be able to:
- Introduce a platform that enables users to carry out morphometric analysis on digital whole slide images
- Demonstrate how morphometric features can be integrated with molecular information to improve stratification of tumours
- Understand how to use morphometric analysis to answer important questions in histolog
COI Disclosure:
None to disclose.
Authors:
Dimitrios G. Oreopoulos1,2, Ameesha Paliwal2, Parsa Babaei-Zadeh2, Alberto Leon2, Kevin Faust2, Phedias Diamandis2,3,4
1 Temerty Centre for Artificial Intelligence Research and Education in Medicine, University of Toronto, Toronto, ON, Canada
2 Department of Pathology, Princess Margaret Cancer Centre, 101 College Street, Toronto, ON, M5G 1L7, Canada.
3 Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada
6 Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada
Presenter
Dimitrios Oreopoulos recently completed his second year in the Bachelor of Health Sciences program at Western University, where he is currently taking courses that focus primarily on the social determinants of health and multidisciplary approaches to research. Dimitrios gained the opportunity to apply these skills in the lab of Dr. Phedias Diamandis for the first time last summer and assisted the lab in developing an Artificial Intelligence workflow and accompanying platform for PHenotyping And Regional Analysis Of Histology (PHARAOH). This year, Dimitrios’ work with Dr. Diamandis has focused on finding novel applications for this tool, which inspired his most recent work integrating molecular and morphometric features with the aim of improving prognostication of diffuse gliomas. Dimitrios has developed a keen interest in neuropathology under the mentorship of Dr. Diamandis and is excited to remain involved in bringing innovation to the forefront of this field.
In his spare time, Dimitrios is passionate about giving back to his community and is a student leader in multiple clubs atWestern University. Dimitrios was recently named co-director of Learning it Together, a volunteer-driven health promotion club that aims to improve early literacy, numeracy, and healthy living skills through providing mentorship and role modeling for underprivileged elementary school children in the London community. Dimitrios also recently began volunteering at a Southwest Ontario Aboriginal Health Access Centre community clinic in London, where he helps coordinate the organization’s “Farm Box” food delivery program with its clients.
Target Audience:
Pathologists, Residents
CanMEDS:
Communicator, Collaborator, Scholar
Integration of molecular and morphometric features improves prognostication of diffuse gliomas
Abstract
BACKGROUND: Diffuse gliomas are the most common forms of adult brain cancer, but exhibit substantial heterogeneity in outcomes. While much of the clinical variability has recently been objectively resolved by molecular profiling and subclassification of gliomas into isocitrate dehydrogenase (IDH)-wildtype (Glioblastoma) and IDH-mutant tumors, the later with and without 1p19q chromosomal codeletions (Oligodendroglioma, Astrocytoma respectively), these subgroups still exhibit patient-level differences in progression that challenge precision management and care.
HYPOTHESIS: We hypothesized that integrating Artificial Intelligence-derived tissue features with modern molecular subgrouping may resolve further variation of outcome within established glioma subtypes.
SCIENTIFIC APPROACH: To explore this hypothesis, we retrieved the clinical cohort of gliomas from The Cancer Genome Atlas and used a publicly available tool we previously developed, PHARAOH, (bioRxiv 2024), to automatically extract and analyze morphometric features of all available cases (TCGA-LGG, n = 722; TCGA-GBM, n= 670). We systematically analyzed 160 different nuclear features and correlated the results with accompanying clinical pathological survival data in 3-5 relevant glioma subcohorts.
RESULTS: This systematic analysis revealed morphometric features that track variations in nuclear staining and cellularity as strong predictors of aggressive disease and poorer outcomes in patients with grade 3 oligodendrogliomas; a notoriously clinical heterogenous glioma subgroup. Importantly, this feature extraction resource is publicly available, allowing these results to be generalized to other cohorts.
CONCLUSIONS: Systematic characterization and validation of morphometric features in additional cohorts and clinical practice may support the development of more advanced multimodal biomarkers that can better predict outcome in this aggressive and heterogeneous disease.