🤖 AI Summary
Existing brain tumor classification methods rely on predefined categories and thus fail to identify novel tumor types absent from training data; while Generalized Category Discovery (GCD) enables joint partitioning of known and unknown classes, it neglects the inherent hierarchical taxonomic structure of brain tumors. Method: We propose HGCD-BT, the first GCD framework incorporating hierarchical class structure, which integrates contrastive learning with semi-supervised hierarchical clustering and introduces a novel hierarchical clustering loss. This loss leverages labeled-data priors while modeling hierarchical semantic relationships among unlabeled samples. Contribution/Results: Evaluated on the OpenSRH dataset, HGCD-BT achieves a 28% absolute improvement in classification accuracy over state-of-the-art GCD methods. Moreover, it demonstrates strong generalization capability for slice-level classification of multimodal digital pathology images.
📝 Abstract
Accurate brain tumor classification is critical for intra-operative decision making in neuro-oncological surgery. However, existing approaches are restricted to a fixed set of predefined classes and are therefore unable to capture patterns of tumor types not available during training. Unsupervised learning can extract general-purpose features, but it lacks the ability to incorporate prior knowledge from labelled data, and semi-supervised methods often assume that all potential classes are represented in the labelled data. Generalized Category Discovery (GCD) aims to bridge this gap by categorizing both known and unknown classes within unlabelled data. To reflect the hierarchical structure of brain tumor taxonomies, in this work, we introduce Hierarchical Generalized Category Discovery for Brain Tumor Classification (HGCD-BT), a novel approach that integrates hierarchical clustering with contrastive learning. Our method extends contrastive learning based GCD by incorporating a novel semi-supervised hierarchical clustering loss. We evaluate HGCD-BT on OpenSRH, a dataset of stimulated Raman histology brain tumor images, achieving a +28% improvement in accuracy over state-of-the-art GCD methods for patch-level classification, particularly in identifying previously unseen tumor categories. Furthermore, we demonstrate the generalizability of HGCD-BT on slide-level classification of hematoxylin and eosin stained whole-slide images from the Digital Brain Tumor Atlas, confirming its utility across imaging modalities.