Can artificial intelligence in spine imaging affect current practice? Practical developments and their clinical status

Abstract

Background: As artificial intelligence (AI) increases its footprint in spine imaging, gauging the clinical relevance of new developments poses an increasingly difficult challenge, especially given the majority of developments reflect experimental or early work. With this summary of available AI tools, focusing on those in clinical use, the benefits of AI in spine imaging are explained for radiologists and surgeons to understand the current state of and potentially the decision to adopt AI in clinical practice.

Methods: Through a narrative review of publications relating to "artificial intelligence" and "spine imaging" in the PubMed database, this article provides an update on AI applications in spine imaging being utilized in current clinical practice.

Results: Current applications of AI in spine imaging range from deep learning image reconstruction and denoising, spine segmentation and biometry, radiological report generation, surgical outcomes prediction, surgical planning, to intraoperative assistance. Developments in deep learning reconstruction (DLR) are most mature and demonstrate improvements to imaging speed and interpretability compared to non-AI alternatives. While clinical implementations exist in other use cases, their performance remains either an area of active investigation or comparable to the level of a human.

Conclusions: Uses of AI in spine imaging span multiple applications with early clinical implementation in most areas, suggesting a promising future ahead.

Keywords: Artificial intelligence; Augmented reality; Image reconstruction; Radiology; Segmentation; Spine imaging; Surgical planning.

Fig. 1

Comparative examples of Siemens DeepResolve DLR (left) versus conventional MR imaging (right) on sagittal (A) cervical spine T2-weighted turbo spin echo sequences and (B) lumbar spine T1-weighted turbo spin echo sequences. In both cases, DeepResolve produced images with twice the resolution in the plane of acquisition (3 × 3 × 3 mm versus 6 × 6 × 3 mm in [A] and 4 × 4 × 4 mm versus 9 × 9 × 4 mm in [B]) with similar acquisition times (0:44 minutes vs. 0:38 minutes in [A] and 1:09 minuted versus 0:57 minutes in [B]). Of note, there was a 2 year gap between images for each patient owing to the introduction of DLR.

Fig. 2

Processing flows for spine imaging analysis platforms.

Fig. 3

Example of preoperative planning of pedicle screw placement from Betram et al. [29] (reprinted under CC BY-NC license [https://creativecommons.org/licenses/by-nc/4.0/]) with screenshots of Brainlab Spine Planning suite showing recommended pedicle screw trajectories.

Fig. 4

Examples of (A) automated grading of lumbar spinal canal stenosis from Tumko et al. [15] (reprinted under CC BY license) and (B) automated diagnosis of various conditions in the lumbar spine from Jamaludin et al. [27]. Full description of license can be found at https://creativecommons.org/licenses/by/4.0/.