Artificial intelligence in spine surgery: a scoping review.

BACKGROUND

The integration of artificial intelligence (AI) into spinal surgery is gaining attention due to its potential to expand evidence-based medicine and provide personalized care. However, its application in day-to-day surgical practice is still in the developmental stage. This scoping review aims to map the landscape of AI applications in spinal surgery, draw current frontiers, and identify gaps in the literature.

METHODS

Following PRISMA guidelines, a scoping review was conducted using PubMed and Cochrane databases up to January 2024. Included studies described AI models or validated AI applications in spinal surgery. A wide range of data was extracted, including objectives, outcomes, model architectures, validation techniques, the type of disease, the institutions involved, and journals.

RESULTS

The United States led contributions (32%), followed by China (18%), Europe (15%), Japan (9%), and South Korea (9%). These publications reached journals with an average two-year impact factor of 3.4. Resource sharing was limited: 4 studies provided self-service applications, seven shared data, and 17 offered code access. The studies addressed diverse spinal pathologies, led by degenerative (24%) and oncological (19%) conditions. Deep learning methods dominated, alongside non-deep learning models. Validation was reported in 76% of studies, both internal (67%), or external (19%). An exhaustive table of all the articles' details is available in the supplementary material.

CONCLUSION

The lack of rigorous external validation and restricted access to AI models and datasets limits AI's widespread adoption. To bridge this gap, stronger collaboration across disciplines, greater transparency in model development, and a concerted effort to ensure that validated models are made publicly accessible are needed.