88 Views

MRI-based Radiomics for Prognosis of Pediatric High-Grade Glioma

Friday, May 3, 2024

Presenting Author(s)

MT

Mariam Tolba, BS

Medical Student
University of Massachusetts Chan Medical School

Co-Author(s)

AC

Andrew Campion, MD

Neuroradiology Fellow
Department of Radiology, Stanford Hospitals and Clinics, Stanford, CA, USA, United States
SC

Samuel H. Cheshier, MD, PhD

Pediatric Neurosurgeon
Department of Neurosurgery, University of Utah School of Medicine, Salt Lake City, UT, USA
Salt Lake City, Utah, United States
AD

Allison Duh, BS

Co-Author
Department of Neurosurgery, Stanford Hospital and Clinics, Stanford, CA, USA, United States
RL

Robert Lober, MD, PhD

Pediatric Neurosurgeon
Department of Neurosurgery, Dayton Children's Hospital, Dayton, OH, USA
Dayton, Ohio, United States
Laura M. Prolo, MD, PhD

Assistant Professor
Division of Pediatric Neurosurgery, Lucile Packard Children’s Hospital at Stanford
Stanford, California, United States
KY

Kristen Yeom, MD

Attending Physician
Department of Radiology, Phoenix Children’s Hospital, United States
Michael Zhang, MD

Adjunct Clinical Instructor
Department of Neurosurgery, Stanford Hospital and Clinics, Stanford, CA, USA
Stanford, California, United States

Introduction: Pediatric high-grade glioma (pHGG) represents one of the leading causes of cancer-related deaths in children with underlying molecular genetics and driver mutations that are distinct from adult HGG. Image-based biomarkers that assist pHGG risk-stratification could potentiate treatment strategies. Towards this, we investigated radiomics approaches for pHGG prognosis.

Methods: We identified 77 children (mean age: 140 months; 34 females) with non-midline-origin, hemispheric pHGG tumors across 5 pediatric institutions. We extracted 1800 image-biomarker standardization initiative (IBSI)-based radiomics tumor features from treatment-naïve, axial T1- and T2-weighted brain MRI. We performed k-fold cross validation and Cox regression to identify optimal predictive features of overall survival. We calculated the risk scores for each patient using a linear combination of the selected features weighted by their coefficients determined by the Cox regression model. Patients were stratified based on the median risk score into high and low risk groups. We used Python (version 3.10.0) for all model developments.

Results: Optimal features predictive of overall survival included three T2- and two T1-based intensity and texture-based features on the wavelet-filtered images. Our best performing multivariable Cox model demonstrated a concordance of 0.75 (95% CI: 0.69-0.82) using radiomics features compared to 0.61 (95% CI: 0.54-0.67) using clinical features (age and sex). Using the risk scores derived from our radiomics model, the high and low risk groups had distinct survival curves on Kaplan-Meier graphs for the test pHGG cohort (log-rank P < 0.05).

Conclusion : In this multi-center study, we identify optimal radiomics features in the creation of a prognostic pHGG model. Computational MRI approaches may offer new approaches for quantitatively evaluating tumor phenotype and serve a potential future role in therapy planning and clinical trials eligibility.