Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Machine learning for performance

An integrated machine learning framework for cerebral aneurysm rupture status classification using clinical, morphological, and hemodynamic features

Zakeri M; Atef A; Aziznia M; Jafari A

Originally published in PLoS One · 2026-10-30

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

AbstractAccurate assessment of cerebral aneurysm rupture status remains a critical challenge in neurovascular research due to the complex interplay between geometric, hemodynamic, and clinical factors. In this study, we developed and evaluated a machine-learning framework for retrospective rupture-status classification in 708 patient-specific aneurysms by integrating 55 predictors: three clinical variables, 35 morphological descriptors, and 17 hemodynamic parameters derived from computational fluid dynamics (CFD) simulations. Blood flow was modeled as incompressible laminar flow with a non-Newtonian Carreau–Yasuda viscosity model to capture shear-thinning behavior in cerebral arteries. We compared 13 supervised classifiers, 10 unsupervised clustering approaches, and two neural-network models. Model training and hyperparameter…

Results

Unsupervised clustering was used as an exploratory check of whether rupture status emerges as an intrinsic structure in the feature space. Fig 4 summarizes the results, showing that best-mapped accuracy ranged from 0.56 to 0.64 across the evaluated algorithms, with DBSCAN and Birch achieving the highest agreement (0.643), followed by Spectral (0.629) and Agglomerative clustering (0.616). Table 3 summarizes the cross-validated performance of the supervised classifiers included in the comparative analysis. Overall discrimination, as measured by mean ROC–AUC, was highest for SVM (0.817 ± 0.041), followed closely by Gaussian Process (0.814 ± 0.041) and BaggingSVM (0.813 ± 0.041). BaggingSVM achieved the highest mean precision (0.747 ± 0.051) and the highest specificity (0.907 ± 0.027), but with a comparatively lower recall (0.487 ±…

Discussion

This study provides a comprehensive evaluation of machine learning approaches for classifying rupture status in cerebral aneurysms using an integrated set of clinical, morphological, and hemodynamic features. By comparing a wide range of supervised and unsupervised algorithms and incorporating physiologically informed hemodynamic simulations, the results provide insights into both the classification performance of different modeling strategies and the underlying factors associated with aneurysm rupture. The unsupervised clustering results (Fig 4) suggest that rupture status does not emerge as a strongly separable structure in the feature space when label information is absent. Best-mapped clustering accuracy ranged from approximately 0.56 to 0.64 across the evaluated methods, indicating only modest alignment between cluster…

Editorial reading guide

  • What data created and tested the model, and was evaluation independent?
  • What is the cost of a false positive, false negative or out-of-distribution decision?