Abstract
Glioblastoma is a highly aggressive primary brain tumor whose diagnosis and grading rely on the identification of heterogeneous histologic sub-regions in digitized tissue sections. The BraTS-Path 2026 challenge frames this as a fine-grained, ten-class classification of H&E-stained image patches. We present a submission built on two complementary pathology foundation models, H-optimus-1 and Virchow2, used as frozen feature extractors. For each backbone we train a lightweight multilayer-perceptron classifier on the extracted embeddings, using focal loss and class-balanced sampling to counter the strong class imbalance of the dataset. We further apply a per-class probability calibration to the H-optimus-1 head and combine its output with that of the Virchow2 head by weighted soft-voting. On the official challenge validation set, the calibrated H-optimus-1 head and the Virchow2 head reach F1 scores of 0.555 and 0.538 respectively, and the weighted ensemble improves over both, attaining an F1 of 0.570, an accuracy of 0.878, an AUC of 0.922, a Matthews correlation coefficient of 0.830, a sensitivity of 0.578, and a specificity of 0.982. The ensemble improves over or matches each individual backbone on every reported metric. Our source code is available at: https://github.com/threyareddy.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BraTS_Path_006.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=XSOIw7RNXA
BibTex
@InProceedings{RedThr_Ensemble_MICCAISAT2026,
author = { Reddy, Threya AND Menon, Arya Pradeep AND Rahman, Farzana AND Makris, Dimitrios},
title = { { Ensemble of Pathology Foundation Models for Fine-Grained Histologic Classification of Glioma } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17252},
month = {pending},
page = {pending}
}