Machine Learning Classification of Cerebral Aneurysm Rupture Status with Morphologic Variables and Hemodynamic Parameters

A comprehensive investigation of morphological features responsible for cerebral aneurysm rupture using machine learning

Cerebral aneurysms are a silent yet prevalent condition that affects a significant global population. Their development can be attributed to various factors, presentations, and treatment approaches. The importance of selecting the appropriate treatment becomes evident upon diagnosis, as the severity of the disease guides the course of action. Cerebral aneurysms are particularly vulnerable in the circle of Willis and pose a significant concern due to the potential for rupture, which can lead to irreversible consequences, including fatality. The primary objective of this study is to predict the rupture status of cerebral aneurysms. To achieve this, we leverage a comprehensive dataset that incorporates clinical and morphological data extracted from 3D real geometries of previous patients. The aim of this research is to provide valuable insights that can help make informed decisions during the treatment process and potentially save the lives of future patients. Diagnosing and predicting aneurysm rupture based solely on brain scans is a significant challenge with limited reliability, even for experienced physicians. However, by employing statistical methods and machine learning techniques, we can assist physicians in making more confident predictions regarding rupture likelihood and selecting appropriate treatment strategies. To achieve this, we used 5 classification machine learning algorithms and trained them on a substantial database comprising 708 cerebral aneurysms. The dataset comprised 3 clinical features and 35 morphological parameters, including 8 novel morphological features introduced for the first time in this study. Our models demonstrated exceptional performance in predicting cerebral aneurysm rupture, with accuracy ranging from 0.76 to 0.82 and precision score from 0.79 to 0.83 for the test dataset. As the data are sensitive and the condition is critical, recall is prioritized as the more crucial parameter over accuracy and precision, and our models achieved outstanding recall score ranging from 0.85 to 0.92. Overall, the best model was Support Vector Machin with an accuracy and precision of 0.82, recall of 0.92 for the testing dataset and the area under curve of 0.84. The ellipticity index, size ratio, and shape irregularity are pivotal features in predicting aneurysm rupture, respectively, contributing significantly to our understanding of this complex condition. Among the multitude of parameters under investigation, these are particularly important. In this study, the ideal roundness parameter was introduced as a novel consideration and ranked fifth among all 38 parameters. Neck circumference and outlet numbers from the new parameters were also deemed significant contributors.

Integrating PointNet-Based Model and Machine Learning Algorithms for Classification of Rupture Status of IAs

BACKGROUND

The rupture of intracranial aneurysms (IAs) would result in subarachnoid hemorrhage with high mortality and disability. Predicting the risk of IAs rupture remains a challenge.

METHODS

This paper proposed an effective method for classifying IAs rupture status by integrating a PointNet-based model and machine learning algorithms. First, medical image segmentation and reconstruction algorithms were applied to 3D Digital Subtraction Angiography (DSA) imaging data to construct three-dimensional IAs geometric models. Geometrical parameters of IAs were then acquired using Geomagic, followed by the computation of hemodynamic clouds and hemodynamic parameters using Computational Fluid Dynamics (CFD). A PointNet-based model was developed to extract different dimensional hemodynamic cloud features. Finally, five types of machine learning algorithms were applied on geometrical parameters, hemodynamic parameters, and hemodynamic cloud features to classify and recognize IAs rupture status. The classification performance of different dimensional hemodynamic cloud features was also compared.

RESULTS

The 16-, 32-, 64-, and 1024-dimensional hemodynamic cloud features were extracted with the PointNet-based model, respectively, and the four types of cloud features in combination with the geometrical parameters and hemodynamic parameters were respectively applied to classify the rupture status of IAs. The best classification outcomes were achieved in the case of 16-dimensional hemodynamic cloud features, the accuracy of XGBoost, CatBoost, SVM, LightGBM, and LR algorithms was 0.887, 0.857, 0.854, 0.857, and 0.908, respectively, and the AUCs were 0.917, 0.934, 0.946, 0.920, and 0.944. In contrast, when only utilizing geometrical parameters and hemodynamic parameters, the accuracies were 0.836, 0.816, 0.826, 0.832, and 0.885, respectively, with AUC values of 0.908, 0.922, 0.930, 0.884, and 0.921.

CONCLUSION

In this paper, classification models for IAs rupture status were constructed by integrating a PointNet-based model and machine learning algorithms. Experiments demonstrated that hemodynamic cloud features had a certain contribution weight to the classification of IAs rupture status. When 16-dimensional hemodynamic cloud features were added to the morphological and hemodynamic features, the models achieved the highest classification accuracies and AUCs. Our models and algorithms would provide valuable insights for the clinical diagnosis and treatment of IAs.

Prediction of small intracranial aneurysm rupture status based on combined Clinical-Radiomics model

Background

Accumulating small unruptured intracranial aneurysms are detected due to the improved quality and higher frequency of cranial imaging, but treatment remains controversial. While surgery or endovascular treatment is effective for small aneurysms with a high risk of rupture, such interventions are unnecessary for aneurysms with a low risk of rupture. Consequently, it is imperative to accurately identify small aneurysms with a low risk of rupture. The purpose of this study was to develop a clinically practical model to predict small aneurysm ruptures based on a radiomics signature and clinical risk factors.

Methods

A total of 293 patients having an aneurysm with a diameter of less than 5 mm, including 199 patients (67.9 %) with a ruptured aneurysm and 94 patients (32.1 %) without a ruptured aneurysm, were included in this study. Digital subtraction angiography or surgical treatment was required in all cases. Data on the clinical risk factors and the features on computed tomography angiography images associated with the aneurysm rupture status were collected simultaneously. We developed a clinical-radiomics model to predict aneurysm rupture status using multivariate logistic regression analysis. The combined clinical-radiomics model was constructed by nomogram analysis. The diagnostic performance, clinical utility, and model calibration were evaluated by operating characteristic curve analysis, decision curve analysis, and calibration analysis.

Results

A combined clinical-radiomics model (Area Under Curve [AUC], 0.85; 95 % confidence interval [CI], 0.757-0.947) showed effective performance in the operating characteristic curve analysis. In the validation cohort, the performance of the combined model was better than that of the radiomics model (AUC, 0.75; 95 % CI, 0.645-0.865; Delong's test p-value = 0.01) and the clinical model (AUC, 0.74; 95 % CI, 0.625-0.851; Delong's test p-value <0.01) alone. The results of the decision curve, nomogram, and calibration analyses demonstrated the clinical utility and good fitness of the combined model.

Conclusion

Our study demonstrated the effectiveness of a clinical-radiomics model for predicting rupture status in small aneurysms.

Advances in research and application of artificial intelligence and radiomic predictive models based on intracranial aneurysm images

Intracranial aneurysm is a high-risk disease, with imaging playing a crucial role in their diagnosis and treatment. The rapid advancement of artificial intelligence in imaging technology holds promise for the development of AI-based radiomics predictive models. These models could potentially enable the automatic detection and diagnosis of intracranial aneurysms, assess their status, and predict outcomes, thereby assisting in the creation of personalized treatment plans. In addition, these techniques could improve diagnostic efficiency for physicians and patient prognoses. This article aims to review the progress of artificial intelligence radiomics in the study of intracranial aneurysms, addressing the challenges faced and future prospects, in hopes of introducing new ideas for the precise diagnosis and treatment of intracranial aneurysms.

A Systematic Review and Meta-Analysis of 3-Dimensional Morphometric Parameters for Cerebral Aneurysms

BACKGROUND

Imaging modalities with increased spatial resolution have allowed for more precise quantification of cerebral aneurysm shape in 3-dimensional (3D) space. We conducted a systematic review and meta-analysis to assess the correlation of individual 3D morphometric measures with cerebral aneurysm rupture status.

METHODS

Two independent reviewers performed a PRISMA (preferred reporting items of systematic reviews and meta-analysis)-guided literature search to identify articles reporting the association between 3D morphometric measures of intracranial aneurysms and rupture status.

RESULTS

A total of 15,122 articles were identified. After screening, 39 studies were included. We identified 17 3D morphometric measures, with 11 eligible for the meta-analysis. The meta-analysis showed a significant association with rupture status for the following measures: nonsphericity index (standardized mean difference [SMD], 0.66; 95% confidence interval [CI], 0.53-0.79; P < 0.0001; I2 = 55.2%), undulation index (SMD, 0.55; 95% CI, 0.26-0.85; P = 0.0017; I2 = 68.1%), ellipticity index (SMD, 0.53; 95% CI, 0.29-0.77; P = 0.0005; I2 = 70.8%), volume (SMD, 0.18; 95% CI, 0.02-0.35; P = 0.0320; I2 = 82.3%), volume/ostium ratio (SMD, 0.43; 95% CI, 0.16-0.71; P = 0.0075; I2 = 90.4%), elongation (SMD, -0.94; 95% CI, -1.12 to -0.76; P = 0.0005; I2 = 0%), flatness (SMD, -0.87; 95% CI, -1.04 to -0.71; P = 0.0005; I2 = 0%), and sphericity (SMD, -0.62; 95% CI, -1.06 to -0.17; P = 0.0215; I2 = 67.9%). A significant risk of publication bias was estimated for the ellipticity index (P = 0.0360) and volume (P = 0.0030).

CONCLUSIONS

Based on the results of a meta-analysis containing 39 studies, the nonsphericity index, undulation index, elongation, flatness, and sphericity demonstrated the most consistent correlation with rupture status.

Machine learning based outcome prediction of microsurgically treated unruptured intracranial aneurysms

Machine learning (ML) has revolutionized data processing in recent years. This study presents the results of the first prediction models based on a long-term monocentric data registry of patients with microsurgically treated unruptured intracranial aneurysms (UIAs) using a temporal train-test split. Temporal train-test splits allow to simulate prospective validation, and therefore provide more accurate estimations of a model's predictive quality when applied to future patients. ML models for the prediction of the Glasgow outcome scale, modified Rankin Scale (mRS), and new transient or permanent neurological deficits (output variables) were created from all UIA patients that underwent microsurgery at the Kepler University Hospital Linz (Austria) between 2002 and 2020 (n = 466), based on 18 patient- and 10 aneurysm-specific preoperative parameters (input variables). Train-test splitting was performed with a temporal split for outcome prediction in microsurgical therapy of UIA. Moreover, an external validation was conducted on an independent external data set (n = 256) of the Department of Neurosurgery, University Medical Centre Hamburg-Eppendorf. In total, 722 aneurysms were included in this study. A postoperative mRS > 2 was best predicted by a quadratic discriminant analysis (QDA) estimator in the internal test set, with an area under the receiver operating characteristic curve (ROC-AUC) of 0.87 ± 0.03 and a sensitivity and specificity of 0.83 ± 0.08 and 0.71 ± 0.07, respectively. A Multilayer Perceptron predicted the post- to preoperative mRS difference > 1 with a ROC-AUC of 0.70 ± 0.02 and a sensitivity and specificity of 0.74 ± 0.07 and 0.50 ± 0.04, respectively. The QDA was the best model for predicting a permanent new neurological deficit with a ROC-AUC of 0.71 ± 0.04 and a sensitivity and specificity of 0.65 ± 0.24 and 0.60 ± 0.12, respectively. Furthermore, these models performed significantly better than the classic logistic regression models (p < 0.0001). The present results showed good performance in predicting functional and clinical outcomes after microsurgical therapy of UIAs in the internal data set, especially for the main outcome parameters, mRS and permanent neurological deficit. The external validation showed poor discrimination with ROC-AUC values of 0.61, 0.53 and 0.58 respectively for predicting a postoperative mRS > 2, a pre- and postoperative difference in mRS > 1 point and a GOS < 5. Therefore, generalizability of the models could not be demonstrated in the external validation. A SHapley Additive exPlanations (SHAP) analysis revealed that this is due to the most important features being distributed quite differently in the internal and external data sets. The implementation of newly available data and the merging of larger databases to form more broad-based predictive models is imperative in the future.

Topological Data Analysis and Use of Mapper for Cerebral Aneurysm Rupture Status Discrimination Based on 3-Dimensional Shape Analysis

BACKGROUND AND OBJECTIVES

Topological data analysis (TDA), which identifies patterns in data through simplified topological signatures, has yet to be applied to aneurysm research. We investigate TDA Mapper graphs (Mapper) for aneurysm rupture discrimination.

METHODS

Two hundred sixteen bifurcation aneurysms (90 ruptured) from 3-dimensional rotational angiography were segmented from vasculature and evaluated for 12 size/shape and 18 enhanced radiomics features. Using Mapper, uniformly dense aneurysm models were represented as graph structures and described by graph shape metrics. Mapper dissimilarity scores (MDS) were computed between pairs of aneurysms based on shape metrics. Lower MDS described similar shapes, whereas high MDS represented shapes that do not share common characteristics. Ruptured/unruptured average MDS scores (how "far" an aneurysm is shape-wise to ruptured/unruptured data sets, respectively) were evaluated for each aneurysm. Rupture status discrimination univariate and multivariate statistics were reported for all features.

RESULTS

The average MDS for pairs of ruptured aneurysms were significantly larger compared with unruptured pairs (0.055 ± 0.027 vs 0.039 ± 0.015, P < .0001). Low MDS suggest that, in contrast to ruptured aneurysms, unruptured aneurysms have similar shape characteristics. An MDS threshold value of 0.0417 (area under the curve [AUC] = 0.73, 80% specificity, 60% sensitivity) was identified for rupture status classification. Under this predictive model, MDS scores <0.0417 would identify unruptured status. MDS statistical performance in discriminating rupture status was similar to that of nonsphericity and radiomics Flatness (AUC = 0.73), outperforming other features. Ruptured aneurysms were more elongated ( P < .0001), flatter ( P < .0001), and showed higher nonsphericity ( P < .0001) compared with unruptured. Including MDS in multivariate analysis resulted in AUC = 0.82, outperforming multivariate analysis on size/shape (AUC = 0.76) and enhanced radiomics (AUC = 0.78) alone.

CONCLUSION

A novel application of Mapper TDA was proposed for aneurysm evaluation, with promising results for rupture status classification. Multivariate analysis incorporating Mapper resulted in high accuracy, which is particularly important given that bifurcation aneurysms are challenging to classify morphologically. This proof-of-concept study warrants future investigation into optimizing Mapper functionality for aneurysm research.

Hemodynamic differences of posterior communicating artery aneurysms between adult and fetal types of posterior cerebral artery

BACKGROUND

The recanalization of posterior communicating artery (PCoA) aneurysms after endovascular treatment has been analyzed by various factors. However, the differences between adult and fetal types of posterior cerebral artery (PCA) have not been fully investigated. The main aim of this study was to investigate hemodynamic differences of PCoA aneurysms between adult and fetal types using computational fluid dynamics (CFD).

METHODS

Fifty-five PCoA aneurysms were evaluated by 3D CT angiography and divided into unruptured aneurysms with adult-type or fetal-type PCAs (19 cases, UA group; 9 cases, UF group) and ruptured aneurysms with adult-type or fetal-type PCAs (17 cases, RA group; 10 cases, RF group). These native aneurysms were analyzed by CFD regarding morphological and hemodynamic characteristics. To evaluate simulated endovascular treatment of aneurysms, CFD was performed using porous media modeling.

RESULTS

Morphologically, the RA group had significantly smaller parent artery diameter (2.91 mm vs. 3.49 mm, p=0.005) and higher size ratio (2.54 vs. 1.78, p=0.023) than the RF group. CFD revealed that the UA group had significantly lower oscillatory shear index (OSI) (0.0032 vs. 0.0078, p=0.004) than the UF group and that the RA group had lower WSS (3.09 vs. 11.10, p=0.001) and higher OSI (0.014 vs. 0.006, p=0.031) than the RF group, while the RF group presented significantly higher intra-aneurysmal flow velocity (0.19 m/s vs. 0.061 m/s, p=0.002) than the RA group. Porous media modeling of simulated treatment revealed higher residual flow volume in the fetal-type groups.

CONCLUSIONS

These results suggested that PCoA aneurysms with fetal-type PCAs had different morphological features and hemodynamic characteristics compared with those with adult-type PCAs, leading to high risks of recanalization.

Predicting the formation of mixed pattern hemorrhages in ruptured middle cerebral artery aneurysms based on a decision tree model: A multicenter study

OBJECTIVE

Mixed-pattern hemorrhages (MPH) commonly occur in ruptured middle cerebral artery (MCA) aneurysms and are associated with poor clinical outcomes. This study aimed to predict the formation of MPH in a multicenter database of MCA aneurysms using a decision tree model.

METHODS

We retrospectively reviewed patients with ruptured MCA aneurysms between January 2009 and June 2020. The MPH was defined as subarachnoid hemorrhages with intracranial hematomas and/or intraventricular hemorrhages and/or subdural hematomas. Univariate and multivariate logistic regression analyses were used to explore the prediction factors of the formation of MPH. Based on these prediction factors, a decision tree model was developed to predict the formation of MPH. Additional independent datasets were used for external validation.

RESULTS

We enrolled 436 patients with ruptured MCA aneurysms detected by computed tomography angiography; 285 patients had MPH (65.4%). A multivariate logistic regression analysis showed that age, aneurysm size, multiple aneurysms, and the presence of a daughter dome were the independent prediction factors of the formation of MPH. The areas under receiver operating characteristic curves of the decision tree model in the training, internal, and external validation cohorts were 0.951, 0.927, and 0.901, respectively.

CONCLUSION

Age, aneurysm size, the presence of a daughter dome, and multiple aneurysms were the independent prediction factors of the formation of MPH. The decision tree model is a useful visual triage tool to predict the formation of MPH that could facilitate the management of unruptured aneurysms in routine clinical work.

A novel clinical-radscore nomogram for predicting ruptured intracranial aneurysm

Objectives

Our study aims to find the more practical and powerful method to predict intracranial aneurysm (IA) rupture through verification of predictive power of different models.

Methods

Clinical and imaging data of 576 patients with IAs including 192 ruptured IAs and matched 384 unruptured IAs was retrospectively analyzed. Radiomics features derived from computed tomography angiography (CTA) images were selected by t-test and Elastic-Net regression. A radiomics score (radscore) was developed based on the optimal radiomics features. Inflammatory markers were selected by multivariate regression. And then 4 models including the radscore, inflammatory, clinical and clinical-radscore models (C-R model) were built. The receiver operating characteristic curve (ROC) was performed to evaluate the performance of each model, PHASES and ELAPSS. The nomogram visualizing the C-R model was constructed to predict the risk of IA rupture.

Results

Five inflammatory features, 2 radiological characteristics and 7 radiomics features were significantly associated with IA rupture. The areas under ROCs of the radscore, inflammatory, clinical and C-R models were 0.814, 0.935, 0.970 and 0.975 in the training cohort and 0.805, 0.927, 0.952 and 0.962 in the validation cohort, respectively.

Conclusion

The inflammatory model performs particularly well in predicting the risk of IA rupture, and its predictive power is further improved by combining with radiological and radiomics features and the C-R model performs the best. The C-R nomogram is a more stable and effective tool than PHASES and ELAPSS for individually predicting the risk of rupture for patients with IA.

USING CONVOLUTIONAL NEURAL NETWORK-BASED SEGMENTATION FOR IMAGE-BASED COMPUTATIONAL FLUID DYNAMICS SIMULATIONS OF BRAIN ANEURYSMS: INITIAL EXPERIENCE IN AUTOMATED MODEL CREATION

"Image-based" computational fluid dynamics (CFD) simulations provide insights into each patient's hemodynamic environment. However, current standard procedures for creating CFD models start with manual segmentation and are time-consuming, hindering the clinical translation of image-based CFD simulations. This feasibility study adopts deep-learning-based image segmentation (hereafter referred to as Artificial Intelligence (AI) segmentation) to replace manual segmentation to accelerate CFD model creation. Two published convolutional neural network-based AI methods (MIScnn and DeepMedic) were selected to perform CFD model extraction from three-dimensional (3D) rotational angiography data containing intracranial aneurysms. In this study, aneurysm morphological and hemodynamic results using models generated by AI segmentation methods were compared with those obtained by two human users for the same data. Interclass coefficients (ICC), Bland-Altman plots, and Pearson's correlation coefficients (PCC) were combined to assess how well AI-generated CFD models were performed. We found that almost perfect agreement was obtained between the human and AI results for all eleven morphological and five out of eight hemodynamic parameters, while a moderate agreement was obtained from the remaining three hemodynamic parameters. Given this level of agreement, using AI segmentation to create CFD models is feasible, given more developments.

Toward Grading Subarachnoid Hemorrhage Risk Prediction: A Machine Learning-Based Aneurysm Rupture Score

OBJECTIVE

Existing approaches neither provide an accurate prediction of subarachnoid hemorrhage (SAH) nor offer a quantitative comparison among a group of its risk factors. To evaluate the population, hypertension, age, size, earlier subarachnoid hemorrhage, and location (PHASES) and unruptured intracranial aneurysm treatment score (UIATS) scores and develop an Artificial Intelligence-based 5-year and lifetime aneurysmal rupture criticality prediction (ARCP) score for a set of risk factors.

METHODS

We design various location-specific and ensemble learning models to develop lifetime rupture risk, employ the longitudinal data to develop a linear regression-based model to predict an aneurysm's growth score, and use the Apriori algorithm to identify risk factors strongly associated with SAH. We develop ARCP by integrating output of Apriori algorithm and ML models and compare with PHASES and UIATS scores along with the scores of a multidisciplinary team of neurosurgeons.

RESULTS

The PHASES and UIATS scores show sensitivities of 22% and 35% and specificities of 76% and 79%, respectively. Location-specific models show precision and recall of 93% and 90% for the middle cerebral artery, 83% and 80% for the anterior communicating artery, and 80% and 80% for the supraclinoid internal carotid artery, respectively. The ensemble method shows both precision and recall of 80%. The validation of the models shows that ARCP performs better than our control group of neurosurgeons. Data-driven knowledge produces comparisons among 61 risk factor combinations, 11 ranked minor, 8 moderate, and 41 severe, and 1 of which is a critical factor.

CONCLUSIONS

The PHASES and UIATS are weak predictors, and the ARCP score can identify, and grade, risk factors associated with SAH.

Can we explain machine learning-based prediction for rupture status assessments of intracranial aneurysms?

Although applying machine learning (ML) algorithms to rupture status assessment of intracranial aneurysms (IA) has yielded promising results, the opaqueness of some ML methods has limited their clinical translation. We presented the first explainability comparison of six commonly used ML algorithms: multivariate logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost), multi-layer perceptron neural network (MLPNN), and Bayesian additive regression trees (BART). A total of 112 IAs with known rupture status were selected for this study. The ML-based classification used two anatomical features, nine hemodynamic parameters, and thirteen morphologic variables. We utilized permutation feature importance, local interpretable model-agnostic explanations (LIME), and SHapley Additive exPlanations (SHAP) algorithms to explain and analyze 6 Ml algorithms. All models performed comparably: LR area under the curve (AUC) was 0.71; SVM AUC was 0.76; RF AUC was 0.73; XGBoost AUC was 0.78; MLPNN AUC was 0.73; BART AUC was 0.73. Our interpretability analysis demonstrated consistent results across all the methods; i.e., the utility of the top 12 features was broadly consistent. Furthermore, contributions of 9 important features (aneurysm area, aneurysm location, aneurysm type, wall shear stress maximum during systole, ostium area, the size ratio between aneurysm width, (parent) vessel diameter, one standard deviation among time-averaged low shear area, and one standard deviation of temporally averaged low shear area less than 0.4 Pa) were nearly the same. This research suggested that ML classifiers can provide explainable predictions consistent with general domain knowledge concerning IA rupture. With the improved understanding of ML algorithms, clinicians' trust in ML algorithms will be enhanced, accelerating their clinical translation.

Association of local solid mechanical, hemodynamic and morphological characteristics with ruptured intracranial aneurysm

The rupture of intracranial aneurysms (IAs) is a complicated phenomenon of which the mechanism is not fully understood. The purpose of this study is to associate local solid mechanical, hemodynamic, and morphological characteristics with rupture regions through statistical means, in an attempt to identify the parameters that are indicative of rupture propensity for IAs. Twenty patient-specific ruptured IA models were reconstructed from digital subtraction angiography (DSA), and applied in the analysis of wall tension, wall shear stress (WSS) and curvature. The precise rupture locations were marked out through intraoperative videos. Pearson correlation analysis was employed to investigate the correlations of these three parameters with patient characteristics and global geometric features. Univariate and multivariate logistic regression analysis were further performed on wall tension, WSS and curvature with regards to rupture and nonrupture regions. Receiver operating characteristic (ROC) analysis defining area under the curve (AUC) was performed on these three parameters. The univariate model of wall tension (AUC, 0.9750), WSS (AUC, 0.9300), curvature (0.8150) and their combined multivariate model (AUC, 0.9875) all present high AUC values. The wall tension, WSS and curvature are acceptable parameters relating to rupture regions. The rupture odd is more sensitive to the wall tension and WSS than curvature. Each logistic model is capable in discriminating ruptures from nonrupture regions, while the multivariate model is the most efficient.

Reliability of using generic flow conditions to quantify aneurysmal haemodynamics: A comparison against simulations incorporating boundary conditions measured in vivo

BACKGROUND AND OBJECTIVES

Initiation, growth, and rupture of intracranial aneurysms are believed to be closely related to their local haemodynamic environment. While haemodynamics can be characterised by use of computational fluid dynamics (CFD), its reliability depends heavily upon accurate assumption of the boundary conditions. Herein, we compared the simulated aneurysmal haemodynamics obtained by use of generic boundary conditions against those obtained under flow conditions measured in vivo.

METHODS

We prospectively recruited 19 patients with intracranial aneurysms requiring 3-dimensional rotational angiography, during which blood pressure at the internal carotid artery was probed by catheter and flowrate measured by a dedicated software tool. Using these flow conditions measured in vivo, we quantified the aneurysmal haemodynamics for each patient by CFD, and then compared the results with those derived from a generic condition reported in the literature, in terms of the time-averaged wall shear stress (TAWSS), oscillatory shear index (OSI), relative residence time (RRT), and percentage of the intra-aneurysmal flow (PIAF). In addition, the effects on aneurysmal haemodynamics of different outflow strategies (splitting method vs. Murray's Law) and simulation schemes (transient vs. steady-state) relative to each flow condition were also assessed.

RESULTS

Differences in the simulated TAWSS (-6.08 ± 10.64 Pa, p = 0.001), OSI (0.06 ± 0.13, p = 0.001), and PIAF (-0.05 ± 0.20, p = 0.012) between the patient-specific and generic boundary conditions were found to be statistically significant, in contrast to that in the RRT (49 ± 307 Pa^-1, p = 0.062). Outflow strategies did not yield statistically significant differences in any of the investigated parameters (all p > 0.05); rather, the resulting parameters were found to be in good correlations (all r > 0.71, p < 0.001). Difference between the aneurysmal TAWSS and the WSS derived from cycle-averaged flowrate condition was found to be minor (0.66 ± 1.36 Pa, p = 0.000), so was that between PIAFs obtained respectively from the transient and steady-state simulations (0.02 ± 0.05, p = 0.000).

CONCLUSIONS

Incorporating into simulation the patient-specific boundary conditions is critical for CFD to characterise aneurysmal haemodynamics, while outflow strategies may not introduce significant uncertainties. Steady-state simulation incorporating the cycle-averaged flow condition may produce unbiased WSS and PIAF compared to the transient analysis.

Machine Learning Algorithms for Rupture Risk Assessment of Intracranial Aneurysms: A Diagnostic Meta-Analysis

OBJECTIVE

Several machine learning algorithms have been increasingly applied to predict the rupture risk of intracranial aneurysms. We performed the present diagnostic meta-analysis to comprehensively evaluate the diagnostic value of machine learning algorithms for assessing the rupture risk of intracranial aneurysms.

METHODS

We systematically searched 3 electronic databases, including Medline (via PubMed), the Cochrane Register of Controlled Trials (via Ovid), and Embase (via Elsevier), to retrieve eligible studies from the databases' inception through March 2021. The latest update was performed in June 2021. StataMP, version 14, was used to estimate all pooled diagnostic values.

RESULTS

A total of 4 studies involving 6 reports were considered to meet the inclusion criteria. Our diagnostic meta-analysis generated the following pooled diagnostic values: sensitivity, 0.84 (95% confidence interval [CI], 0.75-0.90); specificity, 0.78 (95% CI, 0.68-0.85); positive likelihood ratio, 3.8 (95% CI, 2.4-5.9); negative likelihood ratio, 0.21 (95% CI, 0.12-0.35), diagnostic odd ratio, 18 (95% CI, 7-46), and area under the summary receiver operating characteristic curve, 0.88 (95% CI, 0.85-0.90).

CONCLUSIONS

Our findings have demonstrated that the diagnostic performance of machine learning algorithms for the rupture risk assessment of AIs is excellent. Considering that the negative effects resulted from the limited number of eligible studies, we suggest developing more well-designed studies with larger sample sizes to validate our findings.

Predicting the rupture status of small middle cerebral artery aneurysms using random forest modeling

Objective

Small intracranial aneurysms are increasingly being detected; however, a prediction model for their rupture is rare. Random forest modeling was used to predict the rupture status of small middle cerebral artery (MCA) aneurysms with morphological features.

Methods

From January 2009 to June 2020, we retrospectively reviewed patients with small MCA aneurysms (<7 mm). The aneurysms were randomly split into training (70%) and internal validation (30%) cohorts. Additional independent datasets were used for the external validation of 78 small MCA aneurysms from another four hospitals. Aneurysm morphology was determined using computed tomography angiography (CTA). Prediction models were developed using the random forest and multivariate logistic regression.

Results

A total of 426 consecutive patients with 454 small MCA aneurysms (<7 mm) were included. A multivariate logistic regression analysis showed that size ratio (SR), aspect ratio (AR), and daughter dome were associated with aneurysm rupture, whereas aneurysm angle and multiplicity were inversely associated with aneurysm rupture. The areas under the receiver operating characteristic (ROC) curves (AUCs) of random forest models using the five independent risk factors in the training, internal validation, and external validation cohorts were 0.922, 0.889, and 0.92, respectively. The random forest model outperformed the logistic regression model (p = 0.048). A nomogram was developed to assess the rupture of small MCA aneurysms.

Conclusion

Random forest modeling is a good tool for evaluating the rupture status of small MCA aneurysms and may be considered for the management of small aneurysms.

Machine learning prediction of hematoma expansion in acute intracerebral hemorrhage

To examine whether machine learning (ML) approach can be used to predict hematoma expansion in acute intracerebral hemorrhage (ICH) with accuracy and widespread applicability, we applied ML algorithms to multicenter clinical data and CT findings on admission. Patients with acute ICH from three hospitals (n = 351) and those from another hospital (n = 71) were retrospectively assigned to the development and validation cohorts, respectively. To develop ML predictive models, the k-nearest neighbors (k-NN) algorithm, logistic regression, support vector machines (SVMs), random forests, and XGBoost were applied to the patient data in the development cohort. The models were evaluated for their performance on the patient data in the validation cohort, which was compared with previous scoring methods, the BAT, BRAIN, and 9-point scores. The k-NN algorithm achieved the highest area under the receiver operating characteristic curve (AUC) of 0.790 among all ML models, and the sensitivity, specificity, and accuracy were 0.846, 0.733, and 0.775, respectively. The BRAIN score achieved the highest AUC of 0.676 among all previous scoring methods, which was lower than the k-NN algorithm (p = 0.016). We developed and validated ML predictive models of hematoma expansion in acute ICH. The models demonstrated good predictive ability, showing better performance than the previous scoring methods.

Evaluation of aneurysm rupture risk based upon flowrate-independent hemodynamic parameters: a multi-center pilot study

BACKGROUND

Specifying generic flow boundary conditions in aneurysm hemodynamic simulations yields a great degree of uncertainty for the evaluation of aneurysm rupture risk. Herein, we proposed the use of flowrate-independent parameters in discriminating unstable aneurysms and compared their prognostic performance against that of conventional absolute parameters.

METHODS

This retrospective study included 186 aneurysms collected from three international centers, with the stable aneurysms having a minimum follow-up period of 24 months. The flowrate-independent aneurysmal wall shear stress (WSS) and energy loss (EL) were defined as the coefficients of the second-order polynomials characterizing the relationships between the respective parameters and the parent-artery flows. Performance of the flowrate-independent parameters in discriminating unstable aneurysms with the logistic regression, Adaboost, and support-vector machine (SVM) methods was quantified and compared against that of the conventional parameters, in terms of sensitivity, specificity, and area under the curve (AUC).

RESULTS

In discriminating unstable aneurysms, the proposed flowrate-independent EL achieved the highest sensitivity (0.833, 95% CI 0.586 to 0.964) and specificity (0.833, 95% CI 0.672 to 0.936) on the SVM, with the AUC outperforming the conventional EL by 0.133 (95% CI 0.039 to 0.226, p=0.006). Likewise, the flowrate-independent WSS outperformed the conventional WSS in terms of the AUC (difference: 0.137, 95% CI 0.033 to 0.241, p=0.010).

CONCLUSION

The flowrate-independent hemodynamic parameters surpassed their conventional counterparts in predicting the stability of aneurysms, which may serve as a promising set of hemodynamic metrics to be used for the prediction of aneurysm rupture risk when physiologically real vascular boundary conditions are unavailable.

Rupture Risk Assessment for Anterior Communicating Artery Aneurysms Using Decision Tree Modeling

Background

Although anterior communicating artery (ACoA) aneurysms have a higher risk of rupture than aneurysms in other locations, whether to treat unruptured ACoA aneurysms incidentally found is a dilemma because of treatment-related complications. Machine learning models have been widely used in the prediction of clinical medicine. In this study, we aimed to develop an easy-to-use decision tree model to assess the rupture risk of ACoA aneurysms.

Methods

This is a retrospective analysis of rupture risk for patients with ACoA aneurysms from two medical centers. Morphologic parameters of these aneurysms were measured and evaluated. Univariate analysis and multivariate logistic regression analysis were performed to investigate the risk factors of aneurysm rupture. A decision tree model was developed to assess the rupture risk of ACoA aneurysms based on significant risk factors.

Results

In this study, 285 patients were included, among which 67 had unruptured aneurysms and 218 had ruptured aneurysms. Aneurysm irregularity and vessel angle were independent predictors of rupture of ACoA aneurysms. There were five features, including size ratio, aneurysm irregularity, flow angle, vessel angle, and aneurysm size, selected for decision tree modeling. The model provided a visual representation of a decision tree and achieved a good prediction performance with an area under the receiver operating characteristic curve of 0.864 in the training dataset and 0.787 in the test dataset.

Conclusion

The decision tree model is a simple tool to assess the rupture risk of ACoA aneurysms and may be considered for treatment decision-making of unruptured intracranial aneurysms.

An Integrated Model Combining Machine Learning and Deep Learning Algorithms for Classification of Rupture Status of IAs

Background

The rupture risk assessment of intracranial aneurysms (IAs) is clinically relevant. How to accurately assess the rupture risk of IAs remains a challenge in clinical decision-making.

Purpose

We aim to build an integrated model to improve the assessment of the rupture risk of IAs.

Materials and Methods

A total of 148 (39 ruptured and 109 unruptured) IA subjects were retrospectively computed with computational fluid dynamics (CFDs), and the integrated models were proposed by combining machine learning (ML) and deep learning (DL) algorithms. ML algorithms that include random forest (RF), k-nearest neighbor (KNN), XGBoost (XGB), support vector machine (SVM), and LightGBM were, respectively, adopted to classify ruptured and unruptured IAs. A Pointnet DL algorithm was applied to extract hemodynamic cloud features from the hemodynamic clouds obtained from CFD. Morphological variables and hemodynamic parameters along with the extracted hemodynamic cloud features were acted as the inputs to the classification models. The classification results with and without hemodynamic cloud features are computed and compared.

Results

Without consideration of hemodynamic cloud features, the classification accuracy of RF, KNN, XGB, SVM, and LightGBM was 0.824, 0.759, 0.839, 0.860, and 0.829, respectively, and the AUCs of them were 0.897, 0.584, 0.892, 0.925, and 0.890, respectively. With the consideration of hemodynamic cloud features, the accuracy successively increased to 0.908, 0.873, 0.900, 0.926, and 0.917. Meanwhile, the AUCs reached 0.952, 0.881, 0.950, 0.969, and 0.965 eventually. Adding consideration of hemodynamic cloud features, the SVM could perform best with the highest accuracy of 0.926 and AUC of 0.969, respectively.

Conclusion

The integrated model combining ML and DL algorithms could improve the classification of IAs. Adding consideration of hemodynamic cloud features could bring more accurate classification, and hemodynamic cloud features were important for the discrimination of ruptured IAs.

Machine Learning for Hepatocellular Carcinoma Segmentation at MRI: Radiology In Training

A 68-year-old woman with a history of hepatocellular carcinoma underwent conventional transarterial chemoembolization. Manual tumor segmentation on images, which can be used to assess disease progression, is time consuming and may suffer from interobserver reliability issues. The authors present a how-to guide to develop machine learning algorithms for fully automatic segmentation of hepatocellular carcinoma and other tumors for lesion tracking over time.

Construction and Evaluation of Multiple Radiomics Models for Identifying the Instability of Intracranial Aneurysms Based on CTA

Background and Aims

Identifying unruptured intracranial aneurysm instability is crucial for therapeutic decision-making. This study aims to evaluate the role of Radiomics and traditional morphological features in identifying aneurysm instability by constructing and comparing multiple models.

Materials and Methods

A total of 227 patients with 254 intracranial aneurysms evaluated by CTA were included. Aneurysms were divided into unstable and stable groups using comprehensive criteria: the unstable group was defined as aneurysms with near-term rupture, growth during follow-up, or caused compressive symptoms; those without the aforementioned conditions were grouped as stable aneurysms. Aneurysms were randomly divided into training and test sets at a 1:1 ratio. Radiomics and traditional morphological features (maximum diameter, irregular shape, aspect ratio, size ratio, location, etc.) were extracted. Three basic models and two integrated models were constructed after corresponding statistical analysis. Model A used traditional morphological parameters. Model B used Radiomics features. Model C used the Radiomics features related to aneurysm morphology. Furthermore, integrated models of traditional and Radiomics features were built (model A+B, model A+C). The area under curves (AUC) of each model was calculated and compared.

Results

There were 31 (13.7%) patients harboring 36 (14.2%) unstable aneurysms, 15 of which ruptured post-imaging, 16 with growth on serial imaging, and 5 with compressive symptoms, respectively. Four traditional morphological features, six Radiomics features, and three Radiomics-derived morphological features were identified. The classification of aneurysm stability was as follows: the AUC of the training set and test set in models A, B, and C are 0.888 (95% CI 0.808–0.967) and 0.818 (95% CI 0.705–0.932), 0.865 (95% CI 0.777–0.952) and 0.739 (95% CI 0.636–0.841), 0.605(95% CI 0.470–0.740) and 0.552 (95% CI 0.401–0.703), respectively. The AUC of integrated Model A+B was numerically slightly higher than any single model, whereas Model A+C was not.

Conclusions

A radiomics and traditional morphology integrated model seems to be an effective tool for identifying intracranial aneurysm instability, whereas the use of Radiomics-derived morphological features alone is not recommended. Radiomics-based models were not superior to the traditional morphological features model.

Intracranial Aneurysm Rupture Risk Estimation With Multidimensional Feature Fusion

The rupture of aneurysms is the main cause of spontaneous subarachnoid hemorrhage (SAH), which is a serious life-threatening disease with high mortality and permanent disability rates. Therefore, it is highly desirable to evaluate the rupture risk of aneurysms. In this study, we proposed a novel semiautomatic prediction model for the rupture risk estimation of aneurysms based on the CADA dataset, including 108 datasets with 125 annotated aneurysms. The model consisted of multidimensional feature fusion, feature selection, and the construction of classification methods. For the multidimensional feature fusion, we extracted four kinds of features and combined them into the feature set, including morphological features, radiomics features, clinical features, and deep learning features. Specifically, we applied the feature extractor 3D EfficientNet-B0 to extract and analyze the classification capabilities of three different deep learning features, namely, no-sigmoid features, sigmoid features, and binarization features. In the experiment, we constructed five distinct classification models, among which the k-nearest neighbor classifier showed the best performance for aneurysm rupture risk estimation, reaching an F2-score of 0.789. Our results suggest that the full use of multidimensional feature fusion can improve the performance of aneurysm rupture risk assessment. Compared with other methods, our method achieves the state-of-the-art performance for aneurysm rupture risk assessment methods based on CADA 2020.

How Machine Learning is Powering Neuroimaging to Improve Brain Health

This report presents an overview of how machine learning is rapidly advancing clinical translational imaging in ways that will aid in the early detection, prediction, and treatment of diseases that threaten brain health. Towards this goal, we aresharing the information presented at a symposium, "Neuroimaging Indicators of Brain Structure and Function - Closing the Gap Between Research and Clinical Application", co-hosted by the McCance Center for Brain Health at Mass General Hospital and the MIT HST Neuroimaging Training Program on February 12, 2021. The symposium focused on the potential for machine learning approaches, applied to increasingly large-scale neuroimaging datasets, to transform healthcare delivery and change the trajectory of brain health by addressing brain care earlier in the lifespan. While not exhaustive, this overview uniquely addresses many of the technical challenges from image formation, to analysis and visualization, to synthesis and incorporation into the clinical workflow. Some of the ethical challenges inherent to this work are also explored, as are some of the regulatory requirements for implementation. We seek to educate, motivate, and inspire graduate students, postdoctoral fellows, and early career investigators to contribute to a future where neuroimaging meaningfully contributes to the maintenance of brain health.

Detection and analysis of cerebral aneurysms based on X-ray rotational angiography - the CADA 2020 challenge

The Cerebral Aneurysm Detection and Analysis (CADA) challenge was organized to support the development and benchmarking of algorithms for detecting, analyzing, and risk assessment of cerebral aneurysms in X-ray rotational angiography (3DRA) images. 109 anonymized 3DRA datasets were provided for training, and 22 additional datasets were used to test the algorithmic solutions. Cerebral aneurysm detection was assessed using the F2 score based on recall and precision, and the fit of the delivered bounding box was assessed using the distance to the aneurysm. The segmentation quality was measured using the Jaccard index and a combination of different surface distance measures. Systematic errors were analyzed using volume correlation and bias. Rupture risk assessment was evaluated using the F2 score. 158 participants from 22 countries registered for the CADA challenge. The U-Net-based detection solutions presented by the community show similar accuracy compared to experts (F2 score 0.92), with a small number of missed aneurysms with diameters smaller than 3.5 mm. In addition, the delineation of these structures, based on U-Net variations, is excellent, with a Jaccard score of 0.92. The rupture risk estimation methods achieved an F2 score of 0.71. The performance of the detection and segmentation solutions is equivalent to that of human experts. The best results are obtained in rupture risk estimation by combining different image-based, morphological, and computational fluid dynamic parameters using machine learning methods. Furthermore, we evaluated the best methods pipeline, from detecting and delineating the vessel dilations to estimating the risk of rupture. The chain of these methods achieves an F2-score of 0.70, which is comparable to applying the risk prediction to the ground-truth delineation (0.71).

Low Shear Stress at Baseline Predicts Expansion and Aneurysm-Related Events in Patients With Abdominal Aortic Aneurysm

BACKGROUND

Low shear stress has been implicated in abdominal aortic aneurysm (AAA) expansion and clinical events. We tested the hypothesis that low shear stress in AAA at baseline is a marker of expansion rate and future aneurysm-related events.

METHODS

Patients were imaged with computed tomography angiography at baseline and followed up every 6 months >24 months with ultrasound measurements of maximum diameter. From baseline computed tomography angiography, we reconstructed 3-dimensional models for automated computational fluid dynamics simulations and computed luminal shear stress. The primary composite end point was aneurysm repair and/or rupture, and the secondary end point was aneurysm expansion rate.

RESULTS

We included 295 patients with median AAA diameter of 49 mm (interquartile range, 43-54 mm) and median follow-up of 914 (interquartile range, 670-1112) days. There were 114 (39%) aneurysm-related events, with 13 AAA ruptures and 98 repairs (one rupture was repaired). Patients with low shear stress (<0.4 Pa) experienced a higher number of aneurysm-related events (44%) compared with medium (0.4-0.6 Pa; 27%) and high (>0.6 Pa; 29%) shear stress groups (P=0.010). This association was independent of known risk factors (adjusted hazard ratio, 1.72 [95% CI, 1.08-2.73]; P=0.023). Low shear stress was also independently associated with AAA expansion rate (β=+0.28 mm/y [95% CI, 0.02-0.53]; P=0.037).

CONCLUSIONS

We show for the first time that low shear stress (<0.4 Pa) at baseline is associated with both AAA expansion and future aneurysm-related events. Aneurysms within the lowest tertile of shear stress, versus those with higher shear stress, were more likely to rupture or reach thresholds for elective repair. Larger prospective validation trials are needed to confirm these findings and translate them into clinical management.

Classifying Ruptured Middle Cerebral Artery Aneurysms With a Machine Learning Based, Radiomics-Morphological Model: A Multicentral Study

Objective

Radiomics and morphological features were associated with aneurysms rupture. However, the multicentral study of their predictive power for specific-located aneurysms rupture is rare. We aimed to determine robust radiomics features related to middle cerebral artery (MCA) aneurysms rupture and evaluate the additional value of combining morphological and radiomics features in the classification of ruptured MCA aneurysms.

Methods

A total of 632 patients with 668 MCA aneurysms (423 ruptured aneurysms) from five hospitals were included. Radiomics and morphological features of aneurysms were extracted on computed tomography angiography images. The model was developed using a training dataset (407 patients) and validated with the internal (152 patients) and external validation (73 patients) datasets. The support vector machine method was applied for model construction. Optimal radiomics, morphological, and clinical features were used to develop the radiomics model (R-model), morphological model (M-model), radiomics-morphological model (RM-model), clinical-morphological model (CM-model), and clinical-radiomics-morphological model (CRM-model), respectively. A comprehensive nomogram integrating clinical, morphological, and radiomics predictors was generated.

Results

We found seven radiomics features and four morphological predictors of MCA aneurysms rupture. The R-model obtained an area under the receiver operating curve (AUC) of 0.822 (95% CI, 0.776, 0.867), 0.817 (95% CI, 0.744, 0.890), and 0.691 (95% CI, 0.567, 0.816) in the training, temporal validation, and external validation datasets, respectively. The RM-model showed an AUC of 0.848 (95% CI, 0.810, 0.885), 0.865 (95% CI, 0.807, 0.924), and 0.721 (95% CI, 0.601, 0.841) in the three datasets. The CRM-model obtained an AUC of 0.856 (95% CI, 0.820, 0.892), 0.882 (95% CI, 0.828, 0.936), and 0.738 (95% CI, 0.618, 0.857) in the three datasets. The CRM-model and RM-model outperformed the CM-model and M-model in the internal datasets (p < 0.05), respectively. But these differences were not statistically significant in the external dataset. Decision curve analysis indicated that the CRM-model obtained the highest net benefit for most of the threshold probabilities.

Conclusion

Robust radiomics features were determined related to MCA aneurysm rupture. The RM-model exhibited good ability in classifying ruptured MCA aneurysms. Integrating radiomics features into conventional models might provide additional value in ruptured MCA aneurysms classification.

Artificial Neural Network for Predicting the Safe Temporary Artery Occlusion Time in Intracranial Aneurysmal Surgery

BACKGROUND

Temporary artery clipping facilitates safe cerebral aneurysm management, besides a risk for cerebral ischemia. We developed an artificial neural network (ANN) to predict the safe clipping time of temporary artery occlusion (TAO) during intracranial aneurysm surgery.

METHOD

We devised a three-layer model to predict the safe clipping time for TAO. We considered age, the diameter of the right and left middle cerebral arteries (MCAs), the diameter of the right and left A1 segment of anterior cerebral arteries (ACAs), the diameter of the anterior communicating artery, mean velocity of flow at the right and left MCAs, and the mean velocity of flow at the right and left ACAs, as well as the Fisher grading scale of brain CT scans as the input values for the model.

RESULTS

This study included 125 patients: 105 patients from a retrospective cohort for training the model and 20 patients from a prospective cohort for validating the model. The output of the neural network yielded up to 960 s overall safe clipping time for TAO. The input values with the greatest impact on safe TAO were mean velocity of blood at left MCA and left ACA, and Fisher grading scale of brain CT scan.

CONCLUSION

This study presents an axillary framework to improve the accuracy of the estimated safe clipping time interval of temporary artery occlusion in intracranial aneurysm surgery.

Machine Learning-Based Prediction of Small Intracranial Aneurysm Rupture Status Using CTA-Derived Hemodynamics: A Multicenter Study

BACKGROUND AND PURPOSE

Small intracranial aneurysms are being increasingly detected while the rupture risk is not well-understood. We aimed to develop rupture-risk models of small aneurysms by combining clinical, morphologic, and hemodynamic information based on machine learning techniques and to test the models in external validation datasets.

MATERIALS AND METHODS

From January 2010 to December 2016, five hundred four consecutive patients with only small aneurysms (<5 mm) detected by CTA and invasive cerebral angiography (or surgery) were retrospectively enrolled and randomly split into training (81%) and internal validation (19%) sets to derive and validate the proposed machine learning models (support vector machine, random forest, logistic regression, and multilayer perceptron). Hemodynamic parameters were obtained using computational fluid dynamics simulation. External validation was performed in other hospitals to test the models.

RESULTS

The support vector machine performed the best with areas under the curve of 0.88 (95% CI, 0.85-0.92) and 0.91 (95% CI, 0.74-0.98) in the training and internal validation datasets, respectively. Feature ranks suggested hemodynamic parameters, including stable flow pattern, concentrated inflow streams, and a small (<50%) flow-impingement zone, and the oscillatory shear index coefficient of variation, were the best predictors of aneurysm rupture. The support vector machine showed an area under the curve of 0.82 (95% CI, 0.69-0.94) in the external validation dataset, and no significant difference was found for the areas under the curve between internal and external validation datasets (P = .21).

CONCLUSIONS

This study revealed that machine learning had a good performance in predicting the rupture status of small aneurysms in both internal and external datasets. Aneurysm hemodynamic parameters were regarded as the most important predictors.

CT Angiography-Based Radiomics for Classification of Intracranial Aneurysm Rupture

Background: Intracranial aneurysm rupture is a devastating medical event with a high morbidity and mortality rate. Thus, timely detection and management are critical. The present study aimed to identify the aneurysm radiomics features associated with rupture and to build and evaluate a radiomics classification model of aneurysm rupture. Methods: Radiomics analysis was applied to CT angiography (CTA) images of 393 patients [152 (38.7%) with ruptured aneurysms]. Patients were divided at a ratio of 7:3 into retrospective training (n = 274) and prospective test (n = 119) cohorts. A total of 1,229 radiomics features were automatically calculated from each aneurysm. The feature number was systematically reduced, and the most important classifying features were selected. A logistic regression model was constructed using the selected features and evaluated on training and test cohorts. Radiomics score (Rad-score) was calculated for each patient and compared between ruptured and unruptured aneurysms. Results: Nine radiomics features were selected from the CTA images and used to build the logistic regression model. The radiomics model has shown good performance in the classification of the aneurysm rupture on training and test cohorts [area under the receiver operating characteristic curve: 0.92 [95% confidence interval CI: 0.89-0.95] and 0.86 [95% CI: 0.80-0.93], respectively, p < 0.001]. Rad-score showed statistically significant differences between ruptured and unruptured aneurysms (median, 2.50 vs. -1.60 and 2.35 vs. -1.01 on training and test cohorts, respectively, p < 0.001). Conclusion: The results indicated the potential of aneurysm radiomics features for automatic classification of aneurysm rupture on CTA images.