Intracranial Aneurysm Classification and Segmentation via Tri-Axial ROI and Multi-Task Learning

2026-06-25

AI Summary

Current intracranial aneurysm detection methods are limited to binary classification and lack fine-grained anatomical localization and multi-class segmentation, hindering their utility in clinical risk assessment and treatment planning. This work proposes a two-stage multitask framework: first, 2D triplanar regions of interest (ROIs) efficiently localize lesions; then, a 3D nnU-Net performs multi-label classification across 13 anatomical locations alongside joint aneurysm–vessel segmentation, supporting four imaging modalities. To address the severe volume imbalance between aneurysms and vessels, a novel dual-decoder architecture is introduced. Cross-attention pooling and modality-specific auxiliary heads enhance feature learning from heterogeneous imaging data, while a two-fold ensemble strategy improves robustness. The method achieved second place in the RSNA 2025 Challenge, and its code, model weights, and 3D Slicer plugin have been publicly released.

Abstract

Intracranial aneurysms are often asymptomatic until rupture, which carries high mortality. Rupture risk assessment and treatment planning depend on both aneurysm morphology and anatomical location, yet existing automated methods remain limited to binary detection without fine-grained anatomical classification or multi-class segmentation. We present a multi-task framework that simultaneously performs multi-label classification, multi-class aneurysm segmentation, and multi-class vessel segmentation across 13 anatomical locations and four imaging modalities (CTA, MRA, T2, T1-post). Our two-stage approach combines a fast 2D tri-axial Region of Interest (ROI) extraction method with a 3D multi-task nnU-Net backbone. A dual-decoder design mitigates the extreme volume imbalance between aneurysm and vessel classes, while cross-attention pooling and modality-specific auxiliary heads improve feature learning across heterogeneous inputs. Our two-fold ensemble achieved 2nd place in the RSNA 2025 Intracranial Aneurysm Detection challenge. Code, model weights, and a 3D Slicer plugin are publicly available.

Problem

• Intracranial Aneurysm
• Anatomical Classification
• Multi-class Segmentation
• Rupture Risk Assessment
• Medical Image Analysis

Innovation

• multi-task learning
• tri-axial ROI
• dual-decoder architecture
• cross-attention pooling
• heterogeneous medical imaging