Three-dimensional morphological analysis of intracranial aneurysms: a fully automated method for aneurysm sac isolation and quantification

Assessing accuracy and consistency in intracranial aneurysm sizing: human expertise vs. artificial intelligence

Intracranial aneurysms (IAs) are a common vascular pathology and are associated with a risk of rupture, which is often fatal. Aneurysm growth of more than 1 mm is considered a surrogate of rupture risk, therefore, this study presents a comprehensive analysis of intracranial aneurysm measurements utilizing a dataset comprising 358 IA from 248 computed tomography angiography (CTA) scans measured by four junior raters and one senior rater. The study explores the variability in sizing assessments by employing both human raters and an Artificial Intelligence (AI) system. Our findings reveal substantial inter- and intra-rater variability among junior raters, contrasting with the lower intra-rater variability observed in the senior rater. Standard deviations of all raters were above the threshold for IA growth (1 mm). Additionally, the study identifies a systemic bias, indicating a tendency for human experts to measure aneurysms smaller than the AI system. Our findings emphasize the challenges in human assessment while also showcasing the capacity of AI technology to improve the precision and reliability of intracranial aneurysm assessments, especially beneficial for junior raters. The potential of AI was particularly evident in the task of monitoring IA at various intervals, where the AI-based approach surpassed junior raters and achieved performance comparable to senior raters.

A deep learning-based automatic system for intracranial aneurysms diagnosis on three-dimensional digital subtraction angiographic images

BACKGROUND

Intracranial aneurysms (IAs) are a life-threatening disease. Their rupture can lead to hemorrhagic stroke. Most studies applying deep learning for the detection of aneurysms are based on angiographic images. However, critical diagnostic information such as morphology and aneurysm location are not captured by deep learning algorithms and still require manual assessments.

PURPOSE

Digital subtraction angiography (DSA) is the gold standard for aneurysm diagnosis. To facilitate the fully automatic diagnosis of aneurysms, we proposed a comprehensive system for the detection, morphology measurement, and location classification of aneurysms on three-dimensional DSA images, allowing automatic diagnosis without further human input.

METHODS

The system comprised three neural networks: a network for aneurysm detection, a network for morphology measurement, and a network for aneurysm location identification. A cross-scale dual-path transformer module was proposed to effectively fuse local and global information to capture aneurysms of varying sizes. A multitask learning approach was also proposed to allow an accurate localization of aneurysm neck for morphology measurement.

RESULTS

The cross-scale dual-path transformer module was shown to outperform other state-of-the-art network architectures, improving segmentation, and classification accuracy. The detection network in our system achieved an F2 score of 0.946 (recall 93%, precision 100%), better than the winning team in the Cerebral Aneurysm Detection and Analysis challenge. The measurement network achieved a relative error of less than 10% for morphology measurement, at the same level as human operators. Perfect accuracy (100%) was achieved on aneurysm location classification.

CONCLUSIONS

We have demonstrated that a comprehensive system can automatically detect, measure morphology and report the aneurysm location of aneurysms without human intervention. This can be a potential tool for the diagnosis of IAs, improving radiologists' performance and reducing their workload.

An Objective and Repeatable Sac Isolation Technique for Comparing Biomechanical Metrics in Abdominal Aortic Aneurysms

(1) Abdominal aortic aneurysm (AAA) biomechanics-based metrics often reported may be over/under-estimated by including non-aneurysmal regions in the analyses, which is typical, rather than isolating the dilated sac region. We demonstrate the utility of a novel sac-isolation algorithm by comparing peak/mean wall stress (PWS, MWS), with/without sac isolation, for AAA that were categorized as stable or unstable in 245 patient CT image sets. (2) 245 patient computed tomography images were collected, segmented, meshed, and had subsequent finite element analysis performed in preparation of our novel sac isolation technique. Sac isolation was initiated by rotating 3D surfaces incrementally, extracting 2D projections, curve fitting a Fourier series, and taking the local extrema as superior/inferior boundaries for the aneurysmal sac. The PWS/MWS were compared pairwise using the entire aneurysm and the isolated sac alone. (3) MWS, not PWS, was significantly different between the sac alone and the entire aneurysm. We found no statistically significant difference in wall stress measures between stable (n = 222) and unstable (n = 23) groups using the entire aneurysm. However, using sac-isolation, PWS (24.6 ± 7.06 vs. 20.5 ± 8.04 N/cm2; p = 0.003) and MWS (12.0 ± 3.63 vs. 10.5 ± 4.11 N/cm2; p = 0.022) were both significantly higher in unstable vs. stable groups. (4) Our results suggest that evaluating only the AAA sac can influence wall stress metrics and may reveal differences in stable and unstable groups of aneurysms that may not otherwise be detected when the entire aneurysm is used.

The influence of aneurysm morphology on the volume of hemorrhage after rupture

OBJECTIVE

Factors determining the risk of rupture of intracranial aneurysms have been extensively studied; however, little attention is paid to variables influencing the volume of bleeding after rupture. In this study the authors aimed to evaluate the impact of aneurysm morphological variables on the amount of hemorrhage.

METHODS

This was a retrospective cohort analysis of a prospectively collected data set of 116 patients presenting at a single center with subarachnoid hemorrhage due to aneurysmal rupture. A volumetric assessment of the total hemorrhage volume was performed from the initial noncontrast CT. Aneurysms were segmented and reproduced from the initial CT angiography study, and morphology indexes were calculated with a computer-assisted approach. Clinical and demographic characteristics of the patients were included in the study. Factors influencing the volume of hemorrhage were explored with univariate correlations, multiple linear regression analysis, and graphical probabilistic modeling.

RESULTS

The univariate analysis demonstrated that several of the morphological variables but only the patient's age from the clinical-demographic variables correlated (p < 0.05) with the volume of bleeding. Nine morphological variables correlated positively (absolute height, perpendicular height, maximum width, sac surface area, sac volume, size ratio, bottleneck factor, neck-to-vessel ratio, and width-to-vessel ratio) and two correlated negatively (parent vessel average diameter and the aneurysm angle). After multivariate analysis, only the aneurysm size ratio (p < 0.001) and the patient's age (p = 0.023) remained statistically significant. The graphical probabilistic model confirmed the size ratio and the patient's age as the variables most related to the total hemorrhage volume.

CONCLUSIONS

A greater aneurysm size ratio and an older patient age are likely to entail a greater volume of bleeding after subarachnoid hemorrhage.

Reliability and accuracy assessment of morphometric measurements obtained with software for three-dimensional reconstruction of brain aneurysms relative to cerebral angiography measures

OBJECTIVE

To analyze the reliability and accuracy of morphological measurements of software employed to three-dimensionally reconstruct aneurysms and vessels (VMTKlab, version 1.6.1,) with computed tomography angiography (CTA) as the source of images. Agreement with measurements from three-dimensional digital subtraction angiography (3 D-DSA) was evaluated.

METHODS

We evaluated 40 patients presenting with aneurysmal subarachnoid hemorrhage (aSAH). We analyzed four main variables of the aneurysm morphology: absolute height (size), neck (maximum neck width), perpendicular height, and maximum width. The CTA images were uploaded to the software and then segmented to reconstruct the aneurysm. This new method was compared to the current gold standard-3D reconstruction of pretreatment cerebral angiography. We used intraclass correlation coefficient (ICC) and Bland-Altman plot analyses to evaluate the agreement between these methods.

RESULTS

The ICCs obtained for absolute height, neck, perpendicular height, and maximum width were 0.85, 0.57, 0.85, and 0.89, respectively. This implied good agreement except for the neck of the aneurysm (moderate agreement). Bland-Altman plots are presented for the four indexes. The average of the differences was not significant in terms of absolute height, perpendicular height, and maximum width indicating good agreement. However, it was significant for the neck of the aneurysm.

CONCLUSIONS

We report good agreement between the values generated using VMTKlab and cerebral angiography for three of the four main variables. Discrepancies in neck diameter are not surprising and its underestimation with a traditional delineation from cerebral angiography has been reported before.

Artificial Intelligence in the Management of Intracranial Aneurysms: Current Status and Future Perspectives

Intracranial aneurysms with subarachnoid hemorrhage lead to high morbidity and mortality. It is of critical importance to detect aneurysms, identify risk factors of rupture, and predict treatment response of aneurysms to guide clinical interventions. Artificial intelligence has received worldwide attention for its impressive performance in image-based tasks. Artificial intelligence serves as an adjunct to physicians in a series of clinical settings, which substantially improves diagnostic accuracy while reducing physicians' workload. Computer-assisted diagnosis systems of aneurysms based on MRA and CTA using deep learning have been evaluated, and excellent performances have been reported. Artificial intelligence has also been used in automated morphologic calculation, rupture risk stratification, and outcomes prediction with the implementation of machine learning methods, which have exhibited incremental value. This review summarizes current advances of artificial intelligence in the management of aneurysms, including detection and prediction. The challenges and future directions of clinical implementations of artificial intelligence are briefly discussed.

Automated Cutting Plane Positioning for Intracranial Aneurysm Quantification

OBJECTIVE

Aneurysm rupture risk can be assessed by its morphologic and hemodynamics features extracted based on angiographic images. Feature extraction entails aneurysm isolation, typically by manually positioning a cutting plane (MCP). To eliminate intra- and inter-rater variabilities, we propose automatic cutting plane (ACP) positioning based on the analysis of vascular surface mesh.

METHODS

Innovative Hough-like and multi-hypothesis-based detection of aneurysm center, parent vessel inlets, and centerlines were proposed. These were used for initialization and iterative ACP positioning by geometry-inspired cost function optimization. For validation and baseline comparison, we tested MCP and manual neck curve-based isolation. Isolated aneurysm morphology was characterized by size, dome height, aspect ratio, and nonsphericity index.

RESULTS

Methods were applied to 55 intracranial saccular aneurysms from two sites, involving 3-D digital subtraction angiography, computed tomography angiography, and magnetic resonance angiography modalities. Isolation based on ACP resulted in smaller average inter-curve distances (AICDs), compared to those obtained by MCP. One case had AICD higher than 1.0 mm, while 90% of cases had AICD 0.5 mm. Intra- and inter-rater AICD variability of manual neck curves was higher compared to MCP, validating its robustness for clinical purposes.

CONCLUSION

The ACP method achieved high accuracy and reliability of aneurysm isolation, also confirmed by expert visual analysis. So extracted morphologic features were in good agreement with MCP-based ones, therefore, ACP has great potential for aneurysm morphology and hemodynamics quantification in clinical applications.

SIGNIFICANCE

The novel method is angiographic modality agnostic; it delivers repeatable isolation important in follow-up aneurysm assessment; its performance is comparable to MCP; and re-evaluation is fast and simple.

Early Results in Flow Diverter Sizing by Computational Simulation: Quantification of Size Change and Simulation Error Assessment

BACKGROUND

Sizing of flow diverters (FDs) stent in the treatment of intracranial aneurysms is a challenging task due to the change of stent length after implantation.

OBJECTIVE

To quantify the size change and assess the error in length prediction in 82 simulated FD deployments.

METHODS

Eighty-two consecutive patients treated with FDs were retrospectively analyzed. Implanted FD length was measured from angiographic images and compared to the nominal sizes of the implanted device. Length change was obtained by subtracting the nominal length from the real length and dividing by the nominal length. Implanted devices were simulated on 3-dimensional models of each patient. Simulation error was obtained by subtracting real length from simulated length and dividing by the real length of the FD. Subanalysis was done using ANOVA. Statistical significance was set to P < .05, and bootstrap resampling was used.

RESULTS

When assessing the length change of the FD after implantation, changes of 30% in average and up to 80% with reference to the nominal length of the device were observed. The simulation results showed a lower error of 3.52% in average with a maximum of 30%. Paired t-test showed nonsignificant differences between measured and real length (P = .07, with the mean of differences at 0.45 mm, 95% confidence interval [-0.950 0.038]).

CONCLUSION

Nominal length is not an accurate sizing metric when choosing the size of an FD irrespective of the brand and manufacturer. Good estimation of the final length of the stent after deployment as expressed by an error of 3.5% in average.

Computer-Assisted Three-Dimensional Morphology Evaluation of Intracranial Aneurysms

OBJECTIVE

Precise morphologic evaluation is important for intracranial aneurysm (IA) management. At present, clinicians manually measure the IA size and neck diameter on 2-dimensional (2D) digital subtraction angiographic (DSA) images and categorize the IA shape as regular or irregular on 3-dimensional (3D)-DSA images, which could result in inconsistency and bias. We investigated whether a computer-assisted 3D analytical approach could improve IA morphology assessment.

METHODS

Five neurointerventionists evaluated the size, neck diameter, and shape of 39 IAs using current and computer-assisted 3D approaches. In the computer-assisted 3D approach, the size, neck diameter, and undulation index (UI, a shape irregularity metric) were extracted using semiautomated reconstruction of aneurysm geometry using 3D-DSA, followed by IA neck identification and computerized geometry assessment.

RESULTS

The size and neck diameter measured using the manual 2D approach were smaller than computer-assisted 3D measurements by 2.01 mm (P < 0.001) and 1.85 mm (P < 0.001), respectively. Applying the definitions of small IAs (<7 mm) and narrow-necked IAs (<4 mm) from the reported data, interrater variation in manual 2D measurements resulted in inconsistent classification of the size of 14 IAs and the necks of 19 IAs. Visual inspection resulted in an inconsistent shape classification for 23 IAs among the raters. Greater consistency was achieved using the computer-assisted 3D approach for size (intraclass correlation coefficient [ICC], 1.00), neck measurements (ICC, 0.96), and shape quantification (UI; ICC, 0.94).

CONCLUSIONS

Computer-assisted 3D morphology analysis can improve accuracy and consistency in measurements compared with manual 2D measurements. It can also more reliably quantify shape irregularity using the UI. Future application of computer-assisted analysis tools could help clinicians standardize morphology evaluations, leading to more consistent IA evaluations.

Management of Cerebral Aneurysm Descriptors based on an Automatic Ostium Extraction

We present a framework to manage cerebral aneurysms. Rupture risk evaluation is based on manually extracted descriptors, which is time-consuming. Thus, we provide an automatic solution by considering several questions: How can expert knowledge be integrated? How should meta data be defined? Which interaction techniques are needed for data exploration.

Interrater Agreement in the Radiologic Characterization of Ruptured Intracranial Aneurysms Based on Computed Tomography Angiography

OBJECTIVE

To determine interrater agreement in the initial radiologic characterization of ruptured intracranial aneurysms based on computed tomography angiography (CTA) with special emphasis on the rater's level of experience.

METHODS

One junior and one senior rater of 5 high-volume neurovascular tertiary centers evaluated anonymized CTA images of 30 consecutive patients with aneurysmal subarachnoid hemorrhage. Each rater described location, side, size, and morphology in a standardized manner. Interrater variability was analyzed using intraclass correlation and Fleiss' kappa analysis.

RESULTS

There was a high level of agreement for location (κ = 0.76, 95% confidence interval [CI] 0.74-0.79), side (κ = 0.95, CI 0.91-0.99), maximum diameter (intraclass correlation coefficient [ICC] 0.81, CI 0.70-0.90), and dome (ICC 0.78, CI 0.66-0.88) of intracranial aneurysms. In contrast, a lower level of agreement was observed for aneurysms' neck diameter (ICC 0.39, CI 0.28-0.58), the presence of multiple aneurysms (κ = 0.35, CI 0.30-0.40), and aneurysm morphology (blister κ = 0.11, CI -0.05 to 0.07; fusiform κ = 0.54, CI 0.48-0.60; multilobular, κ = 0.39 CI 0.33-0.45). The interrater agreement in the senior rater group was greater than in the junior rater group.

CONCLUSIONS

Interrater agreement confirms the benefit of CTA as initial diagnostic imaging in ruptured intracranial aneurysms but not for aneurysm morphology and presence of multiple aneurysms. A trend towards greater interrater agreement between more experienced raters was noticed.

Factors affecting formation and rupture of intracranial saccular aneurysms

Unruptured intracranial aneurysms represent a decisional challenge. Treatment risks have to be balanced against an unknown probability of rupture. A better understanding of the physiopathology is the basis for a better prediction of the natural history of an individual patient. Knowledge about the possible determining factors arises from a careful comparison between ruptured versus unruptured aneurysms and from the prospective observation and analysis of unbiased series with untreated, unruptured aneurysms. The key point is the correct identification of the determining variables for the fate of a specific aneurysm in a given individual. Thus, the increased knowledge of mechanisms of formation and eventual rupture of aneurysms should provide significant clues to the identification of rupture-prone aneurysms. Factors like structural vessel wall defects, local hemodynamic stress determined also by peculiar geometric configurations, and inflammation as trigger of a wall remodeling are crucial. In this sense the study of genetic modifiers of inflammatory responses together with the computational study of the vessel tree might contribute to identify aneurysms prone to rupture. The aim of this article is to underline the value of a unifying hypothesis that merges the role of geometry, with that of hemodynamics and of genetics as concerns vessel wall structure and inflammatory pathways.

CFD: computational fluid dynamics or confounding factor dissemination? The role of hemodynamics in intracranial aneurysm rupture risk assessment

Image-based computational fluid dynamics holds a prominent position in the evaluation of intracranial aneurysms, especially as a promising tool to stratify rupture risk. Current computational fluid dynamics findings correlating both high and low wall shear stress with intracranial aneurysm growth and rupture puzzle researchers and clinicians alike. These conflicting findings may stem from inconsistent parameter definitions, small datasets, and intrinsic complexities in intracranial aneurysm growth and rupture. In Part 1 of this 2-part review, we proposed a unifying hypothesis: both high and low wall shear stress drive intracranial aneurysm growth and rupture through mural cell-mediated and inflammatory cell-mediated destructive remodeling pathways, respectively. In the present report, Part 2, we delineate different wall shear stress parameter definitions and survey recent computational fluid dynamics studies, in light of this mechanistic heterogeneity. In the future, we expect that larger datasets, better analyses, and increased understanding of hemodynamic-biologic mechanisms will lead to more accurate predictive models for intracranial aneurysm risk assessment from computational fluid dynamics.

RADStation3G: a platform for cardiovascular image analysis integrating PACS, 3D+t visualization and grid computing

RADStation3G is a software platform for cardiovascular image analysis and surgery planning. It provides image visualization and management in 2D, 3D and 3D+t; data storage (images or operational results) in a PACS (using DICOM); and exploitation of patients' data such as images and pathologies. Further, it provides support for computationally expensive processes with grid technology. In this article we first introduce the platform and present a comparison with existing systems, according to the platform's modules (for cardiology, angiology, PACS archived enriched searching and grid computing), and then RADStation3G is described in detail.

Performance assessment of isolation methods for geometrical cerebral aneurysm analysis

Geometrical aneurysm quantification is considered an important topic for the study of aneurysm formation, growth, risk of rupture and also in treatment planning. Usually, quantification involves aneurysm isolation, consisting in the operation of detecting the boundary between the aneurysm dome and its feeding arteries. This operation is sometimes performed manually, but it is a tedious task, subject to user variability. To obtain reproducible measurements, automatic techniques have been proposed. In this paper, we compare different aneurysm isolation techniques, two automatic and one manual-based on a cutting plane. All of them are compared against the results obtained by manual delineations of 26 real cases. We show from the results that automatic methods have good performance, providing results similar to manual methods in average. We also show that automatic methods improve reproducibility compared to direct measurements performed on volume rendering views. Each automatic method presents strengths and weaknesses in particular cases such as small aneurysms, aneurysms with multiple parent vessels or terminal aneurysms, but their reproducibility makes them suitable for robust population studies. Finally, based on this study, we have proposed a criterion that allows to use a combination of the two methods studied and that outperforms each of them individually.

AngioLab--a software tool for morphological analysis and endovascular treatment planning of intracranial aneurysms

Determining whether and how an intracranial aneurysm should be treated is a tough decision that clinicians face everyday. Emerging computational tools could help clinicians analyze clinical data and make these decisions. AngioLab is a single graphical user interface, developed on top of the open source framework GIMIAS, that integrates some of the latest image analysis and computational modeling tools for intracranial aneurysms. Two workflows are available: Advanced Morphological Analysis (AMA) and Endovascular Treatment Planning (ETP). AngioLab has been evaluated by a total of 62 clinicians, who considered the information provided by AngioLab relevant and meaningful. They acknowledged the emerging need of these type of tools and the potential impact they might have on the clinical decision-making process.

Automatic neck plane detection and 3D geometric characterization of aneurysmal sacs

Geometric indices defined on intracranial aneurysms have been widely used in rupture risk assessment and surgical planning. However, most indices employed in clinical settings are currently evaluated based on two-dimensional images that inevitably fail to capture the three-dimensional nature of complex aneurysmal shapes. In addition, since measurements are performed manually, they can suffer from poor inter and intra operator repeatability. The purpose of the current work is to introduce objective and robust techniques for the 3D characterization of intracranial aneurysms, while preserving a close connection to the way aneurysms are currently characterized in clinical settings. Techniques for automatically identifying the neck plane, key aneurysm dimensions, shape factors, and orientations relative to the parent vessel are demonstrated in a population of 15 sidewall and 15 terminal aneurysms whose surface has been obtained by two trained operators using both level-set segmentation and thresholding, the latter reflecting typical clinical practice. Automatically-identified neck planes are shown to be in concordance with those manually positioned by an expert neurosurgeon, and automatically-derived geometric indices are shown to be largely insensitive to segmentation method or operator. By capturing the 3D nature of aneurysmal sacs and by minimizing observer variability, our approach allows large retrospective and prospective studies on aneurysm geometric risk factors to be performed using routinely acquired clinical images.