MAGNETIC RESONANCE IMAGING PROCESSING AND ANALYSIS FOR THE DIAGNOSIS AND MANAGEMENT OF CEREBRAL ARTERIOVENOUS MALFORMATIONS

The internship student will be based at ESME's Ivry-sur-Seine site (engineering school) under the joint supervision of Dr Yasmina LEROUL CHENOUNE (ESME associate professor, HDR, Université Paris Est), Dr Mounir LAHLOUH (ESME associate professor) and Dr Frédéric CLARENÇON (neuroradiologist, Head of Interventional Neuroradiology Department, La Pitié Salpêtrière Hospital). Regular follow-up meetings will be organized either at La Pitié Salpêtrière Hospital or at ESME.

Brain arteriovenous malformations (bAVMs) are anomalous vasculatures that comprise supplying arteries, draining veins, and nidus ^[1]. The pathogenesis of bAVMs has not been completely elucidated ^[2]. The clinical manifestations of bAVMs include intracranial hemorrhage, seizure, headache, and other neurological deficits ^[1]. The diagnosis and treatment involve multidisciplinary team collaboration where radiology, anesthesiology, neurosurgery, and interventional neuroradiology are participating, but it is still controversial and the prognosis of bAVMs patients is overall unsatisfactory ^[3].

Earlier investigations have revealed correlations between the specific angioarchitecture details of bAVMs and several crucial aspects in a clinical context, including analysis of rupture risk ^[4], formulation of treatment plans ^[5], assessment of patient outcomes ^[6], and the provision of follow-up care ^[7]. While there exist many modalities to evaluate the angioarchitecture of bAVMs, they present limitations due to the complexity of the vascular pathologies and the high-flow characteristics ^[8]. The concept of four-dimensional digital subtraction angiography (4D-DSA) was commercially available in 2015 ^[9], it enables viewing the vasculature from any desired angle at every stage of vascular filling and emptying and shows superior spatial/temporal resolution compared to traditional modalities ^{[9, 10]}. However, there is still a lack of high-quality studies to show its effectiveness and further help improve its performance.

Moreover, the most common risk of bAVMs is intracranial hemorrhage, which causes high mortality and morbidity ^[11]. Currently, the prediction for the rupture of bAVMs relies on subjective empirical judgments as well as simple scoring systems ^[12], which are far from precision. The development of radiomics and its application in neurovascular diseases could address this problem. Previous studies have shown the possibility of prediction for the rupture of bAVMs using radiomics ^[13], but this issue needs to be validated and further explored using multiparametric approaches.

Objectives

This study will focus on evaluating and enhancing three conventional rating systems (R₂eD AVM^[14], ARI^[15], VALE^[16]), developing and validating models with MRI-derived radiomic features to predict the risk of rupture of bAVMs. The retrospective database is covering a period of 20 years and includes patients form different French hospital centers. All patients with untreated bAVMs in our database have undergone MRI surveillance, categorizing them into ruptured and unruptured groups based on follow-up data.

Methods

• Systematic review of the literature on MRI for the angioarchitectural analysis of bAVMs
• Systematic review of the literature to explore the application of radiomics in the clinical management of bAVMs (diagnosis, differentiation of hematomas, seizure prediction, rupture prediction...)
• Radiomics model development to extract shaperelated, first-order, and higher-order radiomic features from multimodality sequences: T2-weighted imaging (T2WI), time-of-flight MR angiography (TOF-MRA), and susceptibility-weighted imaging (SWI)
• A supervised radiomics model development and validation using machine learning or deep learning algorithms, with the goal of assessing its effectiveness in predicting bAVMs rupture compared to traditional scoring systems.

References

[1] Chen, C.J., et al., Brain arteriovenous malformations: A review of natural history, pathobiology, and interventions. Neurology, 2020. 95(20):917-927.

[2] Steiger, H.J., Recent progress understanding pathophysiology and genesis of brain AVM-a narrative review. Neurosurg Rev, 2021. 44(6):3165-3175.

[3]  Magro, E., et al., The Treatment of Brain AVMs Study (TOBAS): an all-inclusive framework to integrate clinical care and research. J Neurosurg, 2018. 128(6):1823-1829.

[4] Sahlein, D.H., et al., Features predictive of brain arteriovenous malformation hemorrhage: extrapolation to a physiologic model. Stroke, 2014. 45(7):1964-70.

[5] Luo, C.B., et al., Fistula components of brain arteriovenous malformations: angioarchitecture analysis and embolization prior to gamma-knife surgery. J Chin Med Assoc, 2013. 76(5):277-81.

[6] Zipfel, G.J., et al., Do the Morphological Characteristics of Arteriovenous Malformations affect the result of radiosurgery? J Neurosurg, 2004.

[7] Jin, H., et al., Interval angioarchitectural evolution of brain arteriovenous malformations following rupture. J Neurosurg, 2018. 131(1):96-103.

[8] Raman, A., et al., A Systematic Review Comparing Digital Subtraction Angiogram With Magnetic Resonance Angiogram Studies in Demonstrating the Angioarchitecture of Cerebral Arteriovenous Malformations. Cureus, 2022. 14(6):e25803.

[9] Davis, B., et al., 4D digital subtraction angiography: implementation and demonstration of feasibility. AJNR Am J Neuroradiol, 2013. 34(10):1914-21.

[10] Sandoval-Garcia, C., et al., 4D DSA a new technique for arteriovenous malformation evaluation: a feasibility study. J Neurointerv Surg, 2016. 8(3):300-4.

[11]  C, S., et al., Epidemiology and Natural History of Arteriovenous Malformations. Neurosurg Focus, 2001.

[12] Spetzler, R.F. and N.A. Martin, A Proposed Grading System for Arteriovenous Malformations. J Neurosurg, 1986.

[13] Zhang, S., et al., CT Angiography Radiomics Combining Traditional Risk Factors to Predict Brain Arteriovenous Malformation Rupture: a Machine Learning, Multicenter Study. Translational Stroke Research, 2023.

[14] Feghali J, Yang W, Xu R, Liew J, McDougall CG, Caplan JM, et al. R(2)eD AVM Score. Stroke. 2019;50(7):1703-10.

[15] Mosteiro A, Pedrosa L, Torne R, Rodriguez-Hernandez A, Amaro S, Reyes LA, et al. Venous tortuosity as a novel biomarker of rupture risk in arteriovenous malformations: ARI score. J Neurointerv Surg. 2022;14(12):1220-1225. doi: 10.1136/neurintsurg-2021-018181.

[16] Chen Y, Han H, Meng X, Jin H, Gao D, Ma L, et al. Development and Validation of a Scoring System for Hemorrhage Risk in Brain Arteriovenous Malformations. JAMA Netw Open. 2023; 6(3):e231070.

Candidates with a master’s degree in biomedical engineering, Computer Science, Mathematics, Physics, or related fields are encouraged to apply. They should have skills in image processing and analysis. The candidate should have knowledge of computer vision and have already conducted experiments with learning approaches. An interest in other disciplines and medical applications would be an advantage. Good programming skills are essential (Matlab, Python). Strong communication and writing skills, especially in English, are valuable assets. This project requires strong communication, teamwork, as well as the ability to fit into a multidisciplinary, academic and clinical environment.