The overall goal of this study was to determine how different arterial partial pressure levels of oxygen (PaO2) and carbon dioxide (PaCO 2) affect brain MR images in isoflurane-anesthetized, mechanically ventilated dogs. More specifically the brain volume changes as measured from T1-weighted images and signal intensity changes as measured from susceptibility-weighted images (SWI) were investigated. Six healthy adult male dogs were used for this investigation, in a Latin square split plot design, where each dog served as its own control. Anesthesia was induced with intravenous propofol (6.24 ± 0.6 mg/kg) and maintained with isoflurane at an end tidal concentration of 1.7%. Ventilation parameters and inspired O2 concentration were modified to achieve six different combinations of PaCO2 (20, 40 and 80 mmHg) and PaO2 (100 and 500 mmHg), which were studied in a randomized order in each dog. Two MRI sequences were obtained in each dog for each combination of PaCO 2 and PaO2 after a 5-minute stabilization period, including 3D-T1 and SWI scans. Total brain volume (TBV) and the lateral ventricles volume (LVV) were measured from T1-weighted images. The SWI signal intensity was measured in 6 regions of interest including the thalamus, diencephalic gray matter, diencephalic white matter, medulla oblongata, cerebellar gray matter and cerebellar white matter. Physiological parameters were analysed with an analysis of variance for repeated measures. Total brain volume was analyzed using an analysis of variance, while LVV and signal intensity data were analyzed with an analysis of covariance. Post-hoc Tukey adjustment was made for paired comparisons between combinations. Significance was set at a 'p' value <= 0.05. No significant differences in TBV were found between treatments. The LVV was significantly greater with hypocapnia/normoxemia compared to hypercapnia/normoxemia, and with hypocapnia/hyperoxemia compared to hypercapnia/hyperoxemia (mean ± SD) (127 ± 43 mm3 and 132 ± 43 mm 3, respectively). The SWI signal intensity increased significantly with hypercapnia/normoxemia and hypercapnia/hyperoxemia compared to hypocapnia/normoxemia in the thalamus (8.8% ± 7.4% and 14.8% ± 10.2%, respectively) and diencephalic gray matter (8% ± 6.9% and 13.8% ± 9.6%). The SWI signal intensity was significantly higher during eucapnia/hyperoxemia than during hypocapnia/normoxemia and hypocapnia/hyperoxemia in the thalamus (14% ± 9.4% and 9% ± 4.2%, respectively), diencephalic gray matter (13.6% ± 8.9% and 5.5% ± 4%) and diencephalic white matter (8.2% ± 6.6% and 4.4% ± 2.9%). Based on these results, we conclude that brain MRI images can be affected with varying levels of PaO2 and PaCO2, especially with extreme values, during isoflurane anesthesia in dogs. Therefore, when using anesthetized animals, the level of PaO2 and PaCO2 should be controlled and accounted for in studies measuring brain volumes or when functional MRI studies are carried out.