2026 Proffered Presentations
S051: ESTABLISHING THE UTILITY OF DEEP LEARNING FOR H&E-BASED MENINGIOMA MOLECULAR CLASSIFICATION AND OUTCOME PREDICTION
Alex Landry1; Farshad Nassiri1; Leeor Yefet1; H Lalchungnunga2; Eldad Shulman2; Justin Wang1; Yosef Ellenbogen1; Chloe Gui1; Andrew Ajisebutu1; Vikas Patil1; Jeff Liu1; Qingxia Wei1; Olivia Singh1; Julio Sosa1; Sheila Mansouri1; Shihao Ma1; Bo Wang1; Aditya Raghunathan3; Andrew Gao1; Eytan Ruppin2; Kenneth Aldape2; Gelareh Zadeh3; 1University of Toronto; 2National Cancer Institute; 3Mayo Clinic
Background: The introduction of genomic profiling as a tool for molecular classification and outcome prediction has revolutionized the care of patients with brain tumors. Artificial intelligence (AI) provides advanced avenues to convert complex genomic information into routinely available patient-level information. In this study, we leverage deep learning to demonstrate that H&E can robustly characterize the molecular subtypes, risk groups, and salient chromosomal copy number alterations of the most common brain tumor, meningioma.
Methods: We created an invaluable cohort of 605 meningioma cases with paired DNA methylation and matched digitized H&E images, with additional clinical validation generated using 67 WHO grade 2 meningiomas after gross total resection. We trained and validated five dedicated deep learning models to predict molecular classification of meningiomas (MGs), relevant chromosomal arm aneuploidies (1p loss, 1q gain, 22q loss), and DNA methylation-based 5-year progression free survival risk group (high vs low) using H&E alone. AUROCs and balanced accuracy were used to assess classifier performance and differences between risk-group-specific PFS were compared using the log-rank test.
Results: Our deep learning classifier achieved balanced accuracies of 87-97% for predicting MGs of meningiomas. Similar performance was achieved in the prediction of chromosomal aneuploidies. Our dedicated outcome prediction model was remarkably prognostic even when adjusting for WHO grade, extent of resection, and age (HR 3.49, 95% CI 1.54-7.91, p = 0.003), demonstrating the clear and immediate translational value of deep learning in this context. Beyond this direct translational utility, the generated AI models also provide novel insights into group-specific molecular heterogeneity which have not been detected using bulk genomic approaches to date.
Conclusions: This work is the first to demonstrate the ability to apply deep learning models that can, beyond diagnosis, determine molecular subtypes and predict outcomes in a single brain tumour entity (meningioma) using H&E alone, which to date is only possible using resource-intensive genomic profiling. It further demonstrates the broad clinical utility of applying AI modelling to readily available H&E to democratize access to genomic information globally.