Classifying Histopathologic Glioblastoma Sub-regions with EfficientNet

Sanyukta Adap, Ujjwal Baid, Spyridon Bakas · 2025

Read Classifying Histopathologic Glioblastoma Sub-regions with EfficientNet by Sanyukta Adap, Ujjwal Baid, Spyridon Bakas (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-08896.txt.

Description

Sanyukta Adap, Ujjwal Baid, and Spyridon Bakas train neural networks to label the sub-regions inside glioblastoma, the most common and aggressive brain tumor, from stained tissue slides. Reading these slides by hand is the first step of diagnosis, and automated region maps could help researchers understand the tumor's morphology at scale.

A four-step pipeline classifies six histological regions in the BraTS-Path 2024 challenge data. Several EfficientNet variants were tried, and the B1 and B4 sizes performed best, reaching an F1 score of 0.98 under cross-validation on the training set. The same approach scored near 0.55 on held-out validation data and about 0.52 on the hidden test set. That drop is the paper's honest warning: models that shine in training can falter on new hospitals' slides.

Catalog record

CREATORSanyukta Adap, Ujjwal Baid, Spyridon Bakas
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-08896

Browse the full library · About the Tower of Babel