Machine Learning for Rupture Risk Prediction of Intracranial Aneurysms: Challenging the PHASES Score in Geographically Constrained Areas

Abstract

Intracranial aneurysms represent a potentially life-threatening condition and occur in 3–5% of the population. They are increasingly diagnosed due to the broad application of cranial magnetic resonance imaging and computed tomography in the context of headaches, vertigo, and other unspecific symptoms. For each affected individual, it is utterly important to estimate the rupture risk of the respective aneurysm. However, clinically applied decision tools, such as the PHASES score, remain insufficient. Therefore, a machine learning approach assessing the rupture risk of intracranial aneurysms is proposed in our study. For training and evaluation of the algorithm, data from a single neurovascular center was used, comprising 446 aneurysms (221 ruptured, 225 unruptured). The machine learning model was then compared with the PHASES score and proved superior in accuracy (0.7825), F1-score (0.7975), sensitivity (0.8643), specificity (0.7022), positive predictive value (0.7403), negative predictive value (0.8404), and area under the curve (0.8639). The frequency distributions of the predicted rupture probabilities and the PHASES score were analyzed. A symmetry can be observed between the rupture probabilities, with a symmetry axis at 0.5. A feature importance analysis reveals that the body mass index, consumption of anticoagulants, and harboring vessel are regarded as the most important features when assessing the rupture risk. On the other hand, the size of the aneurysm, which is weighted most in the PHASES score, is regarded as less important. Based on our findings we discuss the potential role of the model for clinical practice in geographically confined aneurysm patients.

Endovascular treatment of multiple intracranial aneurysms in patients with subarachnoid hemorrhage: one or multiple sessions?

“…Tang et al ( 18 ) developed a predictive model by studying intracranial mirror aneurysms and concluded that neck width, bleb formation, and size ratio were independent risk factors for aneurysm rupture, which yielded good results in a validation cohort. Walther et al ( 19 ) demonstrated that the machine learning approach is better than the PHASES score for rupture prediction of UIAs, particularly for patients in geographically constrained areas.…”

Neuroscience Scaffolded by Informatics: A Raging Interdisciplinary Field

“…In this issue, a brand-new approach by a research group from the University of Leipzig uses artificial intelligence for aneurysm rupture risk assessment. A large (N = 446), balanced database from the authors' medical center is engaged in a retrospective study using a gradient-boosting machine from the scikit-learn library and the R-package for statistical computing [4]. Challenging results came from this research, calling for optimistic yet cautious interpretations.…”

“…Aneurysm size and other comorbidities such as diabetes and hypertension, classically considered triggers for aneurysm rupture, were less critical. The machine learning approach [4] improved the aneurysm rupture prediction performance [16]. However, the receiver operating curve of classical scoring ("PHASES score") in the databases of Bijlenga et al, and Walther et al, is not similar, which might harbor a confounding factor.…”

“…Video-tracking and automated detection of mitosis in vitro [1], the macromolecular modeling of kinesin motion [2], and the assumptionfree functional profiling of brain's macrophages [3] have as a common denominator a deep learning model. Essential clinical neuroscience questions such as the estimated risk of brain aneurysm rupture [4] and the surgical outcome of facial nerve transplantation [5] require rapid and complex converging data handling, enabled by deep learning methods that perform beyond the limits of linear algebraic modeling.…”