Accuracy in spinal level determination, including transitional vertebrae: an ASReview supported systematic review.

purposeAccurate spinal level determination is crucial in spinal imaging and in interventional procedures of the spine. In particular, transitional vertebrae can significantly complicate accurate level identification. This systematic review aims to assess the accuracy of conventional and advanced methods for spinal level determination, with particular attention to thoracic and lumbar transitional vertebrae.

methodsA literature search, supported by the machine learning tool ASReview, was conducted in PubMed Web of Science and Embase for studies published between 1962 and 2026, leading to the inclusion of twenty-nine relevant studies. These were categorized into three groups: conventional methods (fluoroscopy, US, CT, MRI, n = 20), transitional vertebrae-specific studies (n = 4), and advanced methods (intra-operative navigation, algorithms and machine learning, n = 5).

resultsWith conventional methods, varying levels of accuracy were achieved with up to 100% accuracy reported in nine studies. In the studies involving transitional vertebrae no 100% was achieved. Inconsistencies in study designs, patient selection, and risk of bias were noted across the studies. This is considered representative for daily practice. The advanced methods showed an accuracy of 100% in most studies. Clinical recommendations of the existing literature on this topic are summarised.

conclusionsThis review highlights variability in radio-diagnostic use and level determination of spine procedures between medical specialists. Recognizing transitional vertebrae, clear intraoperative imaging, and collaboration between radiologists and interventionists are key. Preoperative MRI or CT scans should be compared with intra operative imaging and, a full spine image obtained if needed. Further research is needed to establish specialization transcending protocols for recognizing transitional anatomy during procedures. The advanced methods show promise, but their cost-effectiveness needs evaluation. Deep learning–based studies progress rapidly and have achieved near-perfect accuracy in detecting spinal anomalies; however, they are not yet applicable to routine clinical use.