[Paper Review] Towards Automatic Prediction of Outcome in Treatment of Cerebral Aneurysms

Ashutosh P. Jadhav, Satyananda Kashyap+7 more|arXiv (Cornell University)|Nov 18, 2022

Intracranial Aneurysms: Treatment and Complications6 citations

TL;DR

This study develops a machine learning model that predicts treatment outcomes for wide-neck bifurcation aneurysms using clinical and imaging features derived from 2D and 3D angiographic segmentation. The model achieves 75.31% accuracy and a weighted F1-score of 0.74 in classifying complete versus partial occlusion after endovascular embolization with intrasaccular flow disruptors.

ABSTRACT

Intrasaccular flow disruptors treat cerebral aneurysms by diverting the blood flow from the aneurysm sac. Residual flow into the sac after the intervention is a failure that could be due to the use of an undersized device, or to vascular anatomy and clinical condition of the patient. We report a machine learning model based on over 100 clinical and imaging features that predict the outcome of wide-neck bifurcation aneurysm treatment with an intravascular embolization device. We combine clinical features with a diverse set of common and novel imaging measurements within a random forest model. We also develop neural network segmentation algorithms in 2D and 3D to contour the sac in angiographic images and automatically calculate the imaging features. These deliver 90% overlap with manual contouring in 2D and 83% in 3D. Our predictive model classifies complete vs. partial occlusion outcomes with an accuracy of 75.31%, and weighted F1-score of 0.74.

Motivation & Objective

• To improve prediction of treatment outcomes in cerebral aneurysms treated with intrasaccular flow disruptors.
• To address residual flow into the aneurysm sac post-intervention, a key cause of treatment failure.
• To integrate diverse clinical and imaging features into a unified predictive model for better decision support.
• To automate the extraction of imaging features using deep learning segmentation in 2D and 3D.
• To evaluate model performance in distinguishing complete vs. partial occlusion outcomes after endovascular treatment.

Proposed method

• Collected over 100 clinical and imaging features from patient data, including hemodynamic and anatomical metrics.
• Developed 2D and 3D convolutional neural network (CNN) models for automatic segmentation of the aneurysm sac in rotational angiography.
• Achieved 90% Dice overlap with manual segmentation in 2D and 83% in 3D, validating segmentation accuracy.
• Combined clinical and imaging features into a random forest classifier for outcome prediction.
• Used a weighted F1-score and accuracy as evaluation metrics to assess model performance.
• Trained and validated the model on a dataset of wide-neck bifurcation aneurysms treated with endovascular devices.

Experimental results

Research questions

• RQ1 Can a machine learning model accurately predict complete vs. partial occlusion outcomes after endovascular treatment of wide-neck bifurcation aneurysms?
• RQ2 How well can automated 2D and 3D segmentation of aneurysm sacs in angiographic images match manual contouring?
• RQ3 Which combination of clinical and imaging features yields the highest predictive performance for treatment outcome?
• RQ4 To what extent do novel imaging measurements improve outcome prediction beyond standard clinical factors?
• RQ5 Can deep learning-based segmentation reduce inter-rater variability and labor in feature extraction for aneurysm analysis?

Key findings

• The machine learning model achieved a classification accuracy of 75.31% in predicting complete versus partial occlusion outcomes.
• The model demonstrated a weighted F1-score of 0.74, indicating strong balance between precision and recall across classes.
• The 3D segmentation model achieved 83% Dice overlap with manual segmentation, confirming high anatomical fidelity.
• The 2D segmentation model reached 90% Dice overlap with manual contours, showing strong performance in 2D plane analysis.
• The integration of both clinical and novel imaging features significantly enhanced predictive capability over clinical factors alone.
• The results suggest that automated imaging feature extraction via deep learning can support clinical decision-making in aneurysm treatment planning.