The current clinical techniques for neuraxial needle placement in dogs are predominantly blind without prior knowledge of the depth required to reach the desired space. This study investigated the correlation and defined the relationship between easily obtainable external landmark variables in the dog; occipital–coccygeal length (OCL) and ilium wings distance (IWD), with the skin to epidural and intrathecal space distances using computed tomography (CT). The CT images of 86 dogs of different breeds were examined in this retrospective observational study. Images of dogs in sternal recumbency were optimized to the sagittal view. The distances between the skin and lumbosacral epidural space (LSE) and skin to sacrococcygeal space (SCE) were measured to the ligamentum flavum surrogate (LFS) line. The distance between the skin and the intrathecal space (ITS) was measured from the skin to the vertebral canal at the interlumbar (L5–L6) space. Measurements of the IWD and OCL were performed on dorsal and scout views, respectively. Linear regression equations and Pearson’s correlation coefficients were calculated between variables. Data were reported as mean (standard deviation). Significance was set as alpha < 0.05. After exclusion of four dogs, 82 CT scans were included. The depths were LSE 45 (15) mm, SCE 23 (10) mm, and ITS 50 (15) mm. There was a moderate correlation between OCL with LSE (=14.2 + OCL * 0.05 (r = 0.59, p < 0.0001)), and a strong correlation with ITS (=11.4 + OCL * 0.07 (r = 0.76, p < 0.0001)), while a very weak correlation was found with SCE (=14.0 + OCL * 0.02 (r = 0.27, p < 0.0584)). Similarly, with IWD, there was a moderate correlation with LSE (=10.8 + IWD * 0.56 (r = 0.61, p < 0.0001)), and strong correlation with ITS (=9.2 + IWD * 0.67 (r = 0.75, p < 0.0001)), while a weak correlation was found with SCE (=11.2 + IWD * 0.2 (r = 0.32, p < 0.0033)). Mathematical formulae derived from the multiple regression showed that the body condition score (BCS) improved the relationship between IWD and OCL and the LSE, SCE and ITS, while the addition of body weight was associated with multicollinearity. Further studies are required to determine the accuracy of the algorithms to demonstrate their ability for prediction in a clinical setting.