Quantifying post-treatment vascular remodeling in brain aneurysms using WEKA-based machine learning: a pilot study.

Area of Science:

• Neuroimaging
• Machine Learning
• Vascular Surgery

Background:

• Middle cerebral artery aneurysms require effective treatment monitoring.
• Assessing post-treatment hemodynamic remodeling is crucial for patient outcomes.
• Current methods for evaluating vascular changes can be labor-intensive.

Purpose of the Study:

• To assess the feasibility of a WEKA-based machine learning pipeline for detecting hemodynamic remodeling after treatment for middle cerebral artery aneurysms.
• To compare pre- and postoperative cerebral angiographic images using machine learning.
• To evaluate the potential for automated imaging biomarkers in neurovascular care.

Main Methods:

• Retrospective analysis of 60 patients with middle cerebral artery aneurysms treated with microsurgical clipping or endovascular intervention.
• A WEKA-based Random Forest classifier trained on digital subtraction angiography (DSA) image pairs.
• Custom Python-based post-processing for image denoising and refinement.
• Assessment of vascular surface area changes via pixel count comparison.

Main Results:

• 75% of analyzable image pairs showed increased vascular pixel counts postoperatively.
• Significant increases in vascular pixel counts were observed in the endovascular group (p=0.034 for segmented, p=0.017 for refined pixels).
• No significant differences were found in the neurosurgical group; between-group comparisons did not reach significance.

Conclusions:

• The WEKA pipeline successfully quantified vascular remodeling but requires further refinement and external validation.
• Machine learning-guided segmentation can detect treatment-induced vascular changes, particularly after endovascular therapy.
• This approach holds promise for developing automated imaging biomarkers to aid clinical decision-making in neurovascular treatment.