Streamlining neuroradiology workflow with AI for improved cerebrovascular structure monitoring

Subhashis Banerjee ¹, Fredrik Nysjö ², Dimitrios Toumpanakis ³

⁰Department of Information Technology, Uppsala University, Uppsala, Sweden. subhashis.banerjee@it.uu.se.

Scientific Reports +

|

April 22, 2024

View abstract on PubMed

Summary

This study presents a new AI tool for precise segmentation of brain blood vessels from medical images. The advanced model improves accuracy in monitoring cerebrovascular structures, aiding radiologists in diagnosis and treatment planning.

Area Of Science

Medical Imaging
Artificial Intelligence
Neuroscience

Background

Accurate visualization of intracranial blood vessels is crucial for neurosurgical planning and patient follow-up.
Automated segmentation of cerebrovascular anatomy from Time-Of-Flight Magnetic Resonance Angiography (TOF-MRA) offers enhanced precision for radiologists.
Current methods may lack robustness across diverse imaging datasets and patient populations.

Purpose Of The Study

To introduce a domain-generalized artificial intelligence (AI) solution for volumetric monitoring of cerebrovascular structures.
To develop a robust and accurate method for segmenting the cerebrovascular tree from multi-center TOF-MRA datasets.
To create an AI-assisted tool to improve the efficiency of radiological workflows.

Main Methods

Utilized a multi-task deep convolutional neural network (CNN) with a topology-aware loss function for voxel-wise segmentation.
Employed Decorrelation Loss for domain regularization of the encoder network.
Incorporated auxiliary tasks for enhanced regularization and learning of higher-level representations.

Main Results

The proposed AI model achieved superior performance compared to six state-of-the-art 3D vessel segmentation methods.
The model demonstrated effectiveness across multi-center TOF-MRA datasets, including those with vascular pathologies.
Qualitative performance measures indicated the best scores for the developed method.

Conclusions

The AI solution provides accurate and domain-generalized segmentation of cerebrovascular structures.
The developed AI-assisted Graphical User Interface (GUI) streamlines radiological workflows and saves time.
This technology has the potential to significantly improve preoperative planning and postoperative follow-up for intracranial vessel examination.