TRELLIS-Enhanced Surface Features for Comprehensive Intracranial Aneurysm Analysis

2025-09-03

AI Summary

Intracranial aneurysm detection, segmentation, and hemodynamic modeling are severely hindered by the scarcity of high-quality 3D annotated medical data. To address this, we propose a cross-domain feature transfer framework that, for the first time, adapts surface-geometry latent representations from the general-purpose 3D generative model TRELLIS—pretrained on large-scale non-medical data—to the medical domain, replacing hand-crafted geometric descriptors. Our method jointly encodes mesh-surface features extracted by TRELLIS with a graph neural network, enabling end-to-end multi-task optimization for classification, segmentation, and blood flow field prediction on the Intra3D and AnXplore datasets. Experiments demonstrate substantial improvements across all tasks: our approach surpasses current state-of-the-art methods in classification and segmentation metrics, while reducing relative error in hemodynamic simulation by 15%. These results validate the effective transfer of generic 3D priors to data-scarce medical applications.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Intracranial aneurysms pose a significant clinical risk yet are difficult to detect, delineate and model due to limited annotated 3D data. We propose a cross-domain feature-transfer approach that leverages the latent geometric embeddings learned by TRELLIS, a generative model trained on large-scale non-medical 3D datasets, to augment neural networks for aneurysm analysis. By replacing conventional point normals or mesh descriptors with TRELLIS surface features, we systematically enhance three downstream tasks: (i) classifying aneurysms versus healthy vessels in the Intra3D dataset, (ii) segmenting aneurysm and vessel regions on 3D meshes, and (iii) predicting time-evolving blood-flow fields using a graph neural network on the AnXplore dataset. Our experiments show that the inclusion of these features yields strong gains in accuracy, F1-score and segmentation quality over state-of-the-art baselines, and reduces simulation error by 15%. These results illustrate the broader potential of transferring 3D representations from general-purpose generative models to specialized medical tasks.

Problem

• Detecting intracranial aneurysms from limited 3D data
• Segmenting aneurysm and vessel regions on 3D meshes
• Predicting time-evolving blood-flow fields in aneurysms

Innovation

• Cross-domain feature-transfer using TRELLIS embeddings
• Replacing conventional descriptors with generative surface features
• Enhancing aneurysm classification, segmentation, and flow prediction