You are here

Article Text

Article menu
Download PDFPDF
Hemorrhagic stroke
Original research
Evaluation of predictive models of aneurysm focal growth and bleb development using machine learning techniques
  1. Sara Hadad1,
  2. Fernando Mut1,
  3. Martin Slawski2,
  4. Anne M Robertson3,4,
  5. Juan R Cebral1,5
  1. 1 Department of Bioengineering, George Mason University, Fairfax, Virginia, USA
  2. 2 Statistics Department, George Mason University, Fairfax, Virginia, USA
  3. 3 Departmnet of Mechanical enginering and Material Science, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
  4. 4 Department of Bioengineering, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
  5. 5 Department of Mechanical Engineering, George Mason University, Fairfax, Virginia, USA
  1. Correspondence to Sara Hadad, Department of Bioengineering, George Mason University, Fairfax, Virginia, USA; Shadad{at}gmu.edu

Abstract

Background The presence of blebs increases the rupture risk of intracranial aneurysms (IAs).

Objective To evaluate whether cross-sectional bleb formation models can identify aneurysms with focalized enlargement in longitudinal series.

Methods Hemodynamic, geometric, and anatomical variables derived from computational fluid dynamics models of 2265 IAs from a cross-sectional dataset were used to train machine learning (ML) models for bleb development. ML algorithms, including logistic regression, random forest, bagging method, support vector machine, and K-nearest neighbors, were validated using an independent cross-sectional dataset of 266 IAs. The models' ability to identify aneurysms with focalized enlargement was evaluated using a separate longitudinal dataset of 174 IAs. Model performance was quantified by the area under the receiving operating characteristic curve (AUC), the sensitivity and specificity, positive predictive value, negative predictive value, F1 score, balanced accuracy, and misclassification error.

Results The final model, with three hemodynamic and four geometrical variables, along with aneurysm location and morphology, identified strong inflow jets, non-uniform wall shear stress with high peaks, larger sizes, and elongated shapes as indicators of a higher risk of focal growth over time. The logistic regression model demonstrated the best performance on the longitudinal series, achieving an AUC of 0.9, sensitivity of 85%, specificity of 75%, balanced accuracy of 80%, and a misclassification error of 21%.

Conclusions Models trained with cross-sectional data can identify aneurysms prone to future focalized growth with good accuracy. These models could potentially be used as early indicators of future risk in clinical practice.

  • Aneurysm
  • Blood Flow
  • Brain

Data availability statement

Data are available upon reasonable request. The data that support the findings of this study are available from the corresponding author, upon reasonable request.

https://doi.org/10.1136/jnis-2023-020241

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Data availability statement

    Data are available upon reasonable request. The data that support the findings of this study are available from the corresponding author, upon reasonable request.

    View Full Text

    Footnotes

    • Contributors SH and JRC designed the study. SH and FM contributed to the development of the methodology. FM and JRC designed the software tools. SH and JRC identified focal growing aneurysms in the dataset. SH curated the data. SH and MS performed the data analysis. SH, FM, MS, and JRC contributed to interpretation of the results. AMR and JRC acquired funding, supervised students, and coordinated the project. SH and JRC drafted the manuscript. All authors contributed to the manuscript edition and approved the final manuscript. JRC acted as the guarantor.

    • Funding This work was supported by the NIH grants 2R01NS097457, and R01NS121286.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.