Related Experiment Video
Updated: Aug 6, 2026
09:41
A Pipeline for 3D Multimodality Image Integration and Computer-assisted Planning in Epilepsy Surgery
Published on: May 20, 2016
Multimodal Deep Learning and Knowledge-Enhanced Intelligent Decision Support System for Pipeline Embolization Device
Zhihong Wen1, Shengli Guo2, Yulin Peng1
1College of Information, Mechanical & Electrical Engineering, Shanghai Normal University, Shanghai, China.
CNS Neuroscience & Therapeutics
|July 22, 2026
Summary
This study introduces NeurAneuNet, an AI system that automates Pipeline Embolization Device (PED) selection for intracranial aneurysms. It significantly reduces planning time and cognitive workload for clinicians.
Area of Science:
- Medical Artificial Intelligence
- Neurosurgery
- Medical Imaging Analysis
Background:
- Intracranial aneurysms pose significant risks, necessitating precise treatment planning.
- Current treatment planning for Pipeline Embolization Devices (PEDs) relies heavily on operator expertise, leading to variability.
- Automating device selection and landing zone identification is crucial for consistent and efficient aneurysm treatment.
Purpose of the Study:
- To develop NeurAneuNet, an intelligent decision support system for automating Pipeline Embolization Device (PED) size selection and landing zone identification.
- To reduce reliance on operator experience and enhance consistency in intracranial aneurysm treatment planning.
- To leverage multimodal deep learning for accurate and efficient neurovascular intervention planning.
Main Methods:
- NeurAneuNet integrates multimodal deep learning, employing a dual-path attention U-Net++ for 3D rotational angiography (3DRA) image segmentation.
- Knowledge-augmented features are extracted and fused across five modalities using tensor decomposition.
- A high-order Kolmogorov-Arnold Network (KAN) predicts optimal PED sizing and landing zones.
Main Results:
- NeurAneuNet achieved high accuracy in aneurysm segmentation (Dice 0.874 ± 0.03), PED size classification (91.8%), and diameter prediction (0.24 ± 0.10 mm).
- The system's recommendations agreed with expert consensus in 95.2% of cases in an independent clinical cohort.
- Planning time was reduced by 44.8%, and cognitive workload (NASA-TLX) significantly decreased.
Conclusions:
- NeurAneuNet demonstrates clinically relevant accuracy and efficiency in automating PED treatment planning for intracranial aneurysms.
- The AI system provides robust, intelligent support, enhancing neurosurgical interventions.
- This work lays the foundation for a generalizable AI decision support framework in complex medical interventions.
