Frozen cadaver specimens from three dogs were used to create a sectional anatomic atlas of the sacroiliac region. Frozen/thawed cadaver specimens from 12 dogs were used to develop an ultrasound-guided sacroiliac joint injection technique. Accuracy of the technique was tested in 15 additional canine cadaver specimens, using injectate containing blue dye and iodinated contrast medium. Sonoanatomic landmarks for consistently identifying a caudodorsal window into the canine sacroiliac joint space included the L7-S1 articular process joints, ilial wing, sacral wing, sacral lamina, and median sacral crest. Accuracy of ultrasound-guided sacroiliac joint injection was not significantly affected by operator, but was affected by the tissue location targeted and the reference standard used for calculations. Accuracy of the technique was good for placing injectate into either the synchondrosis component, dorsal sacroiliac ligament or ventral sacroiliac ligament; fair to poor for placing injectate into the synovial component; and poor for placing injectate into all four sacroiliac soft tissue structures. Concurrent placement of injectate into extraarticular tissues occurred frequently. We conclude that ultrasound-guided sacroiliac joint injection is feasible for evaluation as a treatment method for lumbosacral region pain in dogs, but is not sufficiently accurate for localizing pain to the sacroiliac joint alone.