Background: Stereotactic irradiation of large or critically located arteriovenous malformations (AVMs) is a special challenge for clinicians and radiation physicists. To date, no comprehensive comparison of two linac-based radiosurgery systems used for hypofractionated radiotherapy of large AVMs was published. The aim of the study was to compare dose distributions between CyberKnife (CK) system and linac with a micro-multileaf collimator (L-mMLC) in high-grade or critically located cerebral AVMs.Methods: Two sets of plans made for 15 different patients with at least 95% target coverage were selected for comparisons. Conformity (CI), homogeneity (HI) and gradient score (GSI) indices, conformity index proposed by Lomax (CIL), conformation number (CN), quality of coverage (Q), volumes of brain receiving 12,10,8,6,4, and 2 Gy, minimum and maximum doses for critical structures in both treatment planning systems (TPS) were compared. Finally, the number of monitor units needed to deliver the prescribed dose was compared.Results: The mean minimum doses in the target volume were 93.3% (CK) and 90.7% (L-mMLC),p=n.s, maximum: 119.7 and 110%, respectively (p=0.004). The mean CI was 1.46 and 1.86, HI: 1.2, and 1.11, CIL 0.7, and 0.6, CN: 0.68 and 0.58 for CK and mMLC, respectively (p<0.05). The values of GSI and Q were not significantly different. The volumes of the brain receiving low doses (4 Gy and 2 Gy) were significantly lower in the CK system. The number of monitor units necessary to deliver the prescribed dose was significantly greater in case of the CK system.Conclusions: Better conformity can favor the CK system for treatment of large AVMs at the cost of higher maximum doses and worse homogeneity. L-mMLC is superior when shorter treatment time is required. Neither system can assure satisfying dose gradients outside large targets surrounded by numerous critical structures. [ABSTRACT FROM AUTHOR]