Aneurysm Analysis Using Deep Learning.

Area of Science:

• Medical Imaging Analysis
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Current clinical assessments of aneurysms rely on manual diameter measurements, which are time-consuming and lack detailed morphological insights.
• Aneurysm volume measurement offers superior insight into disease progression and severity compared to diameter alone.
• Existing methods lack efficient, automated solutions for aneurysm volume quantification.

Purpose of the Study:

• To develop and validate an automated approach for accurate aneurysm boundary delineation and volume computation.
• To assess the performance of a novel AI-driven method on an unannotated clinical dataset.
• To reduce reliance on manual segmentation and inter-observer variability in aneurysm assessment.

Main Methods:

• Utilized a pre-trained UNet for automatic aorta localization.
• Employed SAM2 for tracking the aorta through complex vascular structures, including aneurysms.
• Applied a Long Short-Term Memory (LSTM) network or expert system to identify aneurysm start and end points.
• Computed aneurysm volumes on a dataset of 60 patients without pixel-wise aorta segmentation.

Main Results:

• Achieved high accuracy in predicting aneurysm start (R²=71%) and end points (R²=76%).
• Demonstrated excellent performance in aneurysm volume prediction with an R² score of 92%.
• Validated the method's efficacy on an unannotated dataset, highlighting its practical applicability.

Conclusions:

• The proposed automated method provides a precise and efficient solution for aneurysm volume measurement.
• This AI-driven technique has the potential to significantly enhance large-scale aneurysm analysis.
• The approach promises to improve clinical decision-making by offering reliable volume data and reducing manual effort.