Artificial Intelligence in Neurointerventions

Cerebrovascular disease covers a wide range of conditions that pose a significant risk of severe, potentially life-threatening sequelae largely due to ischemic and hemorrhagic stroke. Management involves a balance between alleviating the risk associated with the natural history of the disease process and avoiding deleterious iatrogenic complications associated with highly technical interventions. Therefore, prompt diagnosis and judicious decision-making are crucial for averting poor patient outcomes, making the field ideal for the application of artificial intelligence (AI) and machine learning.

This review aims to provide a comprehensive assessment of AI technologies utilized in cerebrovascular disease, focusing on diagnosis of and outcome prediction for acute ischemic stroke (AIS), intracranial aneurysms (IAs), and brain arteriovenous malformations (AVMs). An extensive literature review was conducted to identify recent studies employing the latest AI and machine learning techniques, and we present the clinical significance of AI in cerebrovascular disease, enabling practicing neurointerventionalists to grasp its current and future potential applications.

ISCHEMIC STROKE

Stroke Detection

The diagnosis of AIS hinges on timely and precise imaging techniques. Although noncontrast CT is typically the initial step in suspected stroke cases, its limited sensitivity restricts its utility mainly to ruling out intracerebral hemorrhage.¹ Conversely, MRI diffusion-weighted imaging (DWI) surpasses noncontrast CT with a sensitivity of at least 83% in detecting early AIS.¹ This method has not proved more effective than CT. However, challenges in MRI accessibility and availability lead to potential diagnostic delays. Hence, attempts have been made to integrate AI to improve time to diagnosis.

AI excels in detecting subtle findings imperceptible to human observers.^2-4 For example, Lu et al developed a deep-learning model comprising two deep convolutional neural networks (CNNs) to enhance the identification of AIS. The first CNN acted as a localization tool, outlining suspicious regions of infarction on preannotated CT scans. Subsequently, the output fed into the second CNN, a classification model that assigned a probability of AIS to each case. Following training, the model underwent both internal and external validation demonstrating robust performance across various metrics. Notably, it outperformed two experienced radiologists and enhanced their performance, highlighting the potential synergy between human expertise and machine assistance.⁴

Time of Stroke Onset

The time elapsed since the onset of stroke symptoms is pivotal in treatment decisions, yet this information isn’t always readily available, as up to 25% of AISs occur during sleep.^5-7 Thomalla et al demonstrated that combining DWI and T2-weighted fluid-attenuated inversion recovery (FLAIR) imaging can estimate stroke onset within 4 to 5 hours.⁸ However, identifying this mismatch is challenging even for experienced radiologists. Recent research suggests deep learning can enhance detection, outperforming human interpretation.^9,10 Polson et al developed a deep learning model to predict stroke onset time within the tissue plasminogen activator window and reached an area under the curve (AUC) of 0.814, which was significantly improved compared to a baseline of three neuroradiologists utilizing DWI-FLAIR mismatch.¹⁰

Large vessel occlusion (LVO) is a major reversible cause of AIS,^11-13 and rapid identification of LVO is crucial in initial stroke assessment. AI has become pivotal in this regard, leading to platforms like Rapid CTA (RapidAI), Viz LVO (Viz.ai), and aiOS (Aidoc) with commendable sensitivity and specificity.^14,15 Mobile stroke units (MSUs) are specialized ambulances equipped with basic CT capabilities; however, the quality can be suboptimal. Recent studies have demonstrated AI’s efficacy in detecting LVO using CTA scans from MSUs.¹⁶ Despite MSU CTA limitations in visualizing ischemic events due to image quality variations and early acquisition post–stroke onset, Czap et al developed a model that achieved an AUC of 0.84 on in-hospital CTAs and 0.80 on MSU CTAs.¹⁶ These findings highlight the potential of MSUs, coupled with proficient AI algorithms, to expedite acute diagnoses.

Predicting Outcomes After Stroke

Clinicians navigating the complexities of stroke management often face the added challenge of addressing the concerns of patients’ families regarding both short- and long-term prognoses. Predicting these outcomes involves considering various factors, including clinical symptoms, radiologic findings, medical history, and available treatment options at the time of assessment. ASPECTS (Alberta Stroke Program Early CT Score) emerged as a valuable tool for forecasting outcomes following a middle cerebral artery stroke by quantifying ischemic changes using noncontrast head CT.¹⁷ Originally designed to identify candidates for thrombolytic therapy, ASPECTS has since aided decisions about which patients might benefit from mechanical thrombectomy.^18-20

Initially, manual ASPECTS interpretation was cumbersome and prone to variability due to human error.^21,22 However, the advent of commercially available AI software, such as Rapid ASPECTS (RapidAI) and e-ASPECTS (Brainomix), has automated this process, enhancing interrater agreement.^23-25 Additionally, several studies have developed models capable of generating ASPECTS scores from CT scans, showing strong agreement with scores derived from DWI.^26,27 These AI-driven advancements have surpassed the performance of experienced clinicians, suggesting AI’s potential role in managing AIS beyond diagnosis.

Furthermore, AI models have been deployed to forecast outcomes following AIS. AI has been leveraged to predict negative sequelae of stroke, including malignant cerebral edema and hemorrhagic transformation.^28-31 Such predictive tools may allow clinicians to gauge disease severity earlier in stroke patient management. Various studies have aimed to develop AI algorithms for predicting functional outcomes at different postdischarge stages,^32-34 indicating significant promise for prognostication after stroke.

INTRACRANIAL ANEURYSMS

IAs affect about 2% to 3% of adults and pose a risk of rupture with a high mortality rate. The purpose of intervention is to mitigate the risk of rupture in the future.^35-37 Current models for quantifying the risk of rupture rely on patient-specific risk factors such as hypertension and tobacco use and radiographic features such as aneurysm location and morphology; however, these methods remain imperfect. AI, especially deep learning algorithms, offers promise in enhancing aneurysm detection and prognostication.

Aneurysm Detection

Digital subtraction angiography (DSA) remains the gold standard for diagnosis due to its high sensitivity in detecting aneurysms of all sizes.³⁸ However, two-dimensional (2D) modalities such as CTA and MRA have several advantages, including being noninvasive, less expensive, and more widely available. Thus, work has gone into amplifying their sensitivity using AI. CNN classifiers have shown promising results, with reported sensitivity ranging from 79% to 100%.^39-43 Various algorithms such as RAG (retrieval-augmented generation) and U-Net have been examined. For example, Hainc et al used a commercially available AI system on 2D images, demonstrating a sensitivity of 79%.⁴⁰ With improvements, this application of AI may provide benefit to less experienced interventionalists or in resource-limited settings.

With regard to stratifying IA instability, Bizjak et al demonstrated the application of models in predicting IA growth based on morphologic features.⁴⁴ Utilizing CTA and MRA scans, they employed “deep shape” learning via the PointNet++ model to predict future aneurysm growth and rupture, achieving high sensitivity and accuracy.⁴⁴ Xiong et al developed an AI model using a support vector machine algorithm, which outperformed the commonly used PHASES score in predicting aneurysm rupture, identifying maximum size, location, and irregular shape of the IAs as major predictors.⁴⁵

The highest predictive capacity for assessing rupture risk in IAs emerges when algorithms integrate hemodynamic features and clinical information.^43,46-48 Chen et al found that hemodynamic features were more significant predictors than imaging alone, demonstrating comparable performance to logistic regression modeling.⁴⁶ Similarly, integrating hemodynamic features improved accuracy across metrics in various AI algorithms.⁴⁷

Predicting Outcome

AI has broadened its scope to not only predicting clinical outcomes but also forecasting occlusion rates after endovascular intervention. Various AI systems like elastic net and U-Net have been employed, yielding sensitivities from 75% to 98%.^49-52 Paliwal et al analyzed 84 internal carotid artery sidewall aneurysms treated with flow-diverting stents, incorporating AI algorithm parameters such as hemodynamics and morphology metrics,⁴⁹ achieving 90% accuracy for IA occlusion.

Jadhav et al used AI techniques to predict occlusion of bifurcation aneurysms using intrasaccular flow-diverting devices.⁵² Random forest modeling integrating clinical and imaging features exhibited the highest accuracy of 75.3% and sensitivity of 91.8%. Shiraz Bhurwani et al investigated radiographic outcomes at 6 months for patients treated with flow-diverting stents, achieving an average sensitivity of 0.92 but a specificity of 0.57.⁵⁰ The DIANES (IA diameter, indication, parent artery diameter ratio, neck ratio, side branch artery, and sex) score developed by Guédon et al achieved 89% sensitivity and 81% accuracy, emphasizing the role of imaging and clinical factors in predicting occlusion.⁵¹ Williams et al introduced the Aneurysm Occlusion Assistant, utilizing open-source libraries including Keras, TensorFlow, and scikit-learn, alongside angiographic parametric imaging and segmented DSA imaging, which predicted 6-month occlusion within 7 seconds post–device placement, with an accuracy of 0.84.⁵³ These studies highlight the potential for AI to influence procedure planning as well as decision-making during intervention.

ARTERIOVENOUS MALFORMATIONS

Brain AVMs are congenital lesions that are characterized as an abnormal tangle of connected arteries and veins without intervening capillary beds. They most commonly present with intracerebral hemorrhage with high rates of morbidity and mortality. Management options include surgical resection, stereotactic radiosurgery, and/or endovascular embolization. Similar to IAs, the decision to treat unruptured lesions is centered around predicting the lifetime risk of rupture, which remains an imperfect art and has resulted in several AI applications.

Saggi et al employed AI algorithms, including random forest models and gradient-boosted decision trees, to predict hemorrhage risk in 189 pediatric AVM patients, identifying smaller AVM sizes, left-sided AVMs, and concurrent arterial aneurysms as predictors not discernible through conventional methods.⁵⁴ In adults, Oermann et al developed a three-dimensional surface AI algorithm accurately forecasting adverse events postradiosurgery in 1,810 AVM patients, with comparisons to established scoring systems yielding AUC values from 0.6 to 0.7.⁵⁵ Jiao et al investigated AI-based indicators for postoperative motor deficits in AVM resection recipients, achieving the highest AUC of 0.88 compared to the Spetzler-Martin grading scale.^56,57

CONCLUSION

Prediction and risk stratification are at the heart of diagnosis and treatment of cerebrovascular disease, making it ideal for the adoption of AI. Over the last 5 to 10 years, significant work has been devoted to the development of AI technology applied specifically to cerebrovascular disease. Despite its relative nascency, some AI software has already been implemented in the clinical setting underlining the potential of AI in this realm.

As we gather more robust data to input, AI algorithms will continue to improve and impact management of cerebrovascular pathology, resulting in improved cost-effectiveness and, more importantly, patient outcomes. It is only a matter of time before AI is ubiquitously used in the neurointerventional sphere and revolutionizes clinical care of cerebrovascular disease.

1. Chalela JA, Kidwell CS, Nentwich LM, et al. Magnetic resonance imaging and computed tomography in emergency assessment of patients with suspected acute stroke: a prospective comparison. Lancet. 2007;369:293-298. doi: 10.1016/S0140-6736(07)60151-2

2. Nishio M, Koyasu S, Noguchi S, et al. Automatic detection of acute ischemic stroke using non-contrast computed tomography and two-stage deep learning model. Comput Methods Programs Biomed. 2020;196:105711. doi: 10.1016/j.cmpb.2020.105711

3. Kaothanthong N, Atsavasirilert K, Sarampakhul S, et al. Artificial intelligence for localization of the acute ischemic stroke by non-contrast computed tomography. PloS One. 2022;17:e0277573. doi: 10.1371/journal.pone.0277573

4. Lu J, Zhou Y, Lv W, et al. Identification of early invisible acute ischemic stroke in non-contrast computed tomography using two-stage deep-learning model. Theranostics. 2022;12:5564-5573. doi: 10.7150/thno.74125

5. Lago A, Geffner D, Tembl J, et al. Circadian variation in acute ischemic stroke: a hospital-based study. Stroke. 1998;29:1873-1875. doi: 10.1161/01.str.29.9.1873

6. Fink JN, Kumar S, Horkan C, et al. The stroke patient who woke up: clinical and radiological features, including diffusion and perfusion MRI. Stroke. 2002;33:988-993. doi: 10.1161/01.str.0000014585.17714.67

7. Serena J, Dávalos A, Segura T, et al. Stroke on awakening: looking for a more rational management. Cerebrovasc Dis Basel Switz. 2003;16:128-133. doi: 10.1159/000070592

8. Thomalla G, Cheng B, Ebinger M, et al. DWI-FLAIR mismatch for the identification of patients with acute ischaemic stroke within 4·5 h of symptom onset (PRE-FLAIR): a multicentre observational study. Lancet Neurol. 2011;10:978-986. doi: 10.1016/S1474-4422(11)70192-2

9. Zhu H, Jiang L, Zhang H, et al. An automatic machine learning approach for ischemic stroke onset time identification based on DWI and FLAIR imaging. NeuroImage Clin. 2021;31:102744. doi: 10.1016/j.nicl.2021.102744

10. Polson JS, Zhang H, Nael K, et al. Identifying acute ischemic stroke patients within the thrombolytic treatment window using deep learning. J Neuroimaging Off J Am Soc Neuroimaging. 2022;32:1153-1160. doi: 10.1111/jon.13043

11. Smith WS, Lev MH, English JD, et al. Significance of large vessel intracranial occlusion causing acute ischemic stroke and TIA. Stroke. 2009;40:3834-3840. doi: 10.1161/STROKEAHA.109.561787

12. Smith WS, Tsao JW, Billings ME, et al. Prognostic significance of angiographically confirmed large vessel intracranial occlusion in patients presenting with acute brain ischemia. Neurocrit Care. 2006;4:14-17. doi: 10.1385/ncc:4:1:014

13. Malhotra K, Gornbein J, Saver JL. Ischemic strokes due to large-vessel occlusions contribute disproportionately to stroke-related dependence and death: a review. Front Neurol. 2017;8:651. doi: 10.3389/fneur.2017.00651

14. Murray NM, Unberath M, Hager GD, Hui FK. Artificial intelligence to diagnose ischemic stroke and identify large vessel occlusions: a systematic review. J Neurointerv Surg. 2020;12:156-164. doi: 10.1136/neurintsurg-2019-015135

15. Karamchandani RR, Helms AM, Satyanarayana S, et al. Automated detection of intracranial large vessel occlusions using Viz.ai software: experience in a large, integrated stroke network. Brain Behav. 2023;13:e2808. doi: 10.1002/brb3.2808

16. Czap AL, Bahr-Hosseini M, Singh N, et al. Machine learning automated detection of large vessel occlusion from mobile stroke unit computed tomography angiography. Stroke. 2022;53:1651-1656. doi: 10.1161/STROKEAHA.121.036091

17. Barber PA, Demchuk AM, Zhang J, Buchan AM. Validity and reliability of a quantitative computed tomography score in predicting outcome of hyperacute stroke before thrombolytic therapy. ASPECTS Study Group. Alberta Stroke Programme Early CT Score [published erratum appears in Lancet. 2000;355:2170]. Lancet. 2000;355:1670-1674. doi: 10.1016/s0140-6736(00)02237-6

18. Yoo AJ, Zaidat OO, Chaudhry ZA, et al. Impact of pretreatment noncontrast CT Alberta Stroke Program Early CT Score on clinical outcome after intra-arterial stroke therapy. Stroke. 2014;45:746-751. doi: 10.1161/STROKEAHA.113.004260

19. Goyal M, Demchuk AM, Menon BK, et al. Randomized assessment of rapid endovascular treatment of ischemic stroke. N Engl J Med. 2015;372:1019-1030. doi: 10.1056/NEJMoa1414905

20. Powers WJ, Derdeyn CP, Biller J, et al. 2015 American Heart Association/American Stroke Association focused update of the 2013 guidelines for the early management of patients with acute ischemic stroke regarding endovascular treatment: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2015;46:3020-3035. doi: 10.1161/STR.0000000000000074

21. Wardlaw JM, Farrall AJ, Perry D, et al. Factors influencing the detection of early CT signs of cerebral ischemia: an internet-based, international multiobserver study. Stroke. 2007;38:1250-1256. doi: 10.1161/01.STR.0000259715.53166.25

22. Menon BK, Puetz V, Kochar P, Demchuk AM. ASPECTS and other neuroimaging scores in the triage and prediction of outcome in acute stroke patients. Neuroimaging Clin N Am. 2011;21:407-423, xii. doi: 10.1016/j.nic.2011.01.007

23. Nagel S, Sinha D, Day D, et al. e-ASPECTS software is non-inferior to neuroradiologists in applying the ASPECT score to computed tomography scans of acute ischemic stroke patients. Int J Stroke. 2017;12:615-622. doi: 10.1177/1747493016681020

24. Goebel J, Stenzel E, Guberina N, et al. Automated ASPECT rating: comparison between the Frontier ASPECT Score software and the Brainomix software. Neuroradiology. 2018;60:1267-1272. doi: 10.1007/s00234-018-2098-x

25. Chan N, Sibtain N, Booth T, et al. Machine-learning algorithm in acute stroke: real-world experience. Clin Radiol. 2023;78:e45-e51. doi: 10.1016/j.crad.2022.10.007

26. Albers GW, Wald MJ, Mlynash M, et al. Automated calculation of Alberta Stroke Program Early CT Score: validation in patients with large hemispheric infarct. Stroke. 2019;50:3277-3279. doi: 10.1161/STROKEAHA.119.026430

27. Kuang H, Najm M, Chakraborty D, et al. Automated ASPECTS on noncontrast CT scans in patients with acute ischemic stroke using machine learning. AJNR Am J Neuroradiol. 2019;40:33-38. doi: 10.3174/ajnr.A5889

28. Yu Y, Guo D, Lou M, et al. Prediction of hemorrhagic transformation severity in acute stroke from source Pperfusion MRI. IEEE Trans Biomed Eng. 2018;65:2058-2065. doi: 10.1109/TBME.2017.2783241

29. Foroushani HM, Hamzehloo A, Kumar A, et al. Quantitative serial CT imaging-derived features improve prediction of malignant cerebral edema after ischemic stroke. Neurocrit Care. 2020;33:785-792. doi: 10.1007/s12028-020-01056-5

30. Choi JM, Seo SY, Kim PJ, et al. Prediction of Hemorrhagic Transformation after Ischemic Stroke Using Machine Learning. J Pers Med. 2021;11:863. doi: 10.3390/jpm11090863

31. Hoffman H, Wood JS, Cote JR, et al. Machine learning prediction of malignant middle cerebral artery infarction after mechanical thrombectomy for anterior circulation large vessel occlusion. J Stroke Cerebrovasc Dis. 2023;32:106989. doi: 10.1016/j.jstrokecerebrovasdis.2023.106989

32. Forkert ND, Verleger T, Cheng B, et al. Multiclass support vector machine-based lesion mapping predicts functional outcome in ischemic stroke patients. PloS One. 2015;10:e0129569. doi: 10.1371/journal.pone.0129569

33. Ozkara BB, Karabacak M, Hamam O, et al. Prediction of functional outcome in stroke patients with proximal middle cerebral artery occlusions using machine learning models. J Clin Med. 2023;12:839. doi: 10.3390/jcm12030839

34. Dragoș HM, Stan A, Pintican R, et al. MRI radiomics and predictive models in assessing ischemic stroke outcome—a systematic review. Diagnostics (Basel). 2023;13:857. doi: 10.3390/diagnostics13050857

35. Rinkel GJ, Djibuti M, Algra A, van Gijn J. Prevalence and risk of rupture of intracranial aneurysms: a systematic review. Stroke. 1998;29:251-256. doi: 10.1161/01.str.29.1.251

36. Vlak MH, Algra A, Brandenburg R, Rinkel GJ. Prevalence of unruptured intracranial aneurysms, with emphasis on sex, age, comorbidity, country, and time period: a systematic review and meta-analysis. Lancet Neurol. 2011;10:626-636. doi: 10.1016/S1474-4422(11)70109-0

37. UCAS Japan Investigators, Morita A, Kirino T, et al. The natural course of unruptured cerebral aneurysms in a Japanese cohort. N Engl J Med. 2012;366:2474-2482. doi: 10.1056/NEJMoa1113260

38. Thompson BG, Brown RD, Amin-Hanjani S, et al. Guidelines for the management of patients with unruptured intracranial aneurysms: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke. 2015;46:2368-2400. doi: 10.1161/STR.0000000000000070

39. Duan H, Huang Y, Liu L, et al. Automatic detection on intracranial aneurysm from digital subtraction angiography with cascade convolutional neural networks. Biomed Eng Online. 2019;18:110. doi: 10.1186/s12938-019-0726-2

40. Hainc N, Mannil M, Anagnostakou V, et al. Deep learning based detection of intracranial aneurysms on digital subtraction angiography: a feasibility study. Neuroradiol J. 2020;33:311-317. doi: 10.1177/1971400920937647

41. Jin H, Geng J, Yin Y, et al. Fully automated intracranial aneurysm detection and segmentation from digital subtraction angiography series using an end-to-end spatiotemporal deep neural network. J Neurointerv Surg. 2020;12:1023-1027. doi: 10.1136/neurintsurg-2020-015824

42. Zeng Y, Liu X, Xiao N, et al. Automatic diagnosis based on spatial information fusion feature for intracranial aneurysm. IEEE Trans Med Imaging. 2020;39:1448-1458. doi: 10.1109/TMI.2019.2951439

43. Ou C, Qian Y, Chong W, et al. A deep learning-based automatic system for intracranial aneurysms diagnosis on three-dimensional digital subtraction angiographic images. Med Phys. 2022;49:7038-7053. doi: 10.1002/mp.15846

44. Bizjak Ž, Pernuš F, Špiclin Ž. Deep shape features for predicting future intracranial aneurysm growth. Front Physiol. 2021;12:644349. doi: 10.3389/fphys.2021.644349

45. Xiong W, Chen T, Li J, et al. Interpretable machine learning model to predict rupture of small intracranial aneurysms and facilitate clinical decision. Neurol Sci. 2022;43:6371-6379. doi: 10.1007/s10072-022-06351-x

46. Chen G, Lu M, Shi Z, et al. Development and validation of machine learning prediction model based on computed tomography angiography-derived hemodynamics for rupture status of intracranial aneurysms: a Chinese multicenter study. Eur Radiol. 2020;30:5170-5182. doi: 10.1007/s00330-020-06886-7

47. Chen HY, Elmer J, Zafar SF, et al. Combining transcranial Doppler and EEG data to predict delayed cerebral ischemia after subarachnoid hemorrhage. Neurology. 2022;98:e459-e469. doi: 10.1212/WNL.0000000000013126

48. Detmer FJ, Lückehe D, Mut F, et al. Comparison of statistical learning approaches for cerebral aneurysm rupture assessment. Int J Comput Assist Radiol Surg. 2020;15:141-150. doi: 10.1007/s11548-019-02065-2

49. Paliwal N, Jaiswal P, Tutino VM, et al. Outcome prediction of intracranial aneurysm treatment by flow diverters using machine learning. Neurosurg Focus. 2018;45:E7. doi: 10.3171/2018.8.FOCUS18332

50. Shiraz Bhurwani MM, Waqas M, Podgorsak AR, et al. Feasibility study for use of angiographic parametric imaging and deep neural networks for intracranial aneurysm occlusion prediction. J Neurointerv Surg. 2020;12(7):714-719. doi: 10.1136/neurintsurg-2019-015544

51. Guédon A, Thépenier C, Shotar E, et al. Predictive score for complete occlusion of intracranial aneurysms treated by flow-diverter stents using machine learning. J Neurointerv Surg. 2021;13:341-346. doi: 10.1136/neurintsurg-2020-016748

52. Jadhav A, Kashyap S, Bulu H, et al. Towards automatic prediction of outcome in treatment of cerebral aneurysms. AMIA Annu Symp Proc. 2022;2022:570-579.

53. Williams KA, Podgorsak AR, Bhurwani MMS, et al. The Aneurysm Occlusion Assistant, an AI platform for real time surgical guidance of intracranial aneurysms. Proc SPIE Int Soc Opt Eng. 2021;11601:116010V. doi: 10.1117/12.2581003

54. Saggi S, Winkler EA, Ammanuel SG, et al. Machine learning for predicting hemorrhage in pediatric patients with brain arteriovenous malformation. J Neurosurg Pediatr. 2022;30:203-209. doi: 10.3171/2022.4.PEDS21470

55. Oermann EK, Rubinsteyn A, Ding D, et al. Using a machine learning approach to predict outcomes after radiosurgery for cerebral arteriovenous malformations. Sci Rep. 2016;6:21161. doi: 10.1038/srep21161

56. Jiao Y, Zhang J, Yang X, et al. Artificial intelligence-assisted evaluation of the spatial relationship between brain arteriovenous malformations and the corticospinal tract to predict postsurgical motor defects. AJNR Am J Neuroradiol. 2023;44:17-25. doi: 10.3174/ajnr.A7735

57. Spetzler RF, Martin NA. A proposed grading system for arteriovenous malformations. J Neurosurg. 1986;65:476-483. doi: 10.3171/jns.1986.65.4.0476