Table 1.

Extracted characteristics from the studies included in the review.

Characteristic	Description
(1) Author's data	•Country of the authors' affiliations: the affiliation country from the majority of the authors or the corresponding author •Authors' fields of expertise: health fields, data science fields, or both •Status with industry: if one of more author was affiliated with an industrial partner
(2) Year	•The year it was published based on Medline, IEEE Xplore, Web of Science, or Embased databases
(3) Study type and design	•Study type: classification of primary studies into basic, clinical and epidemiological research; and subclassification into interventional or observational [1] •Study design: classification of studies into retrospective or prospective nature and further categorization into study design being cross-sectional, cohort, descriptive, case-control, or case-series type study designs using a described classification algorithms [4]
(4) Area of spine care focus	•Clinical application type: either diagnostic tools, clinical decision support tools, automated clinical/instrumentation assessment, clinical outcome prediction, combined or others •Studied anatomy: either cervical, thoracic, lumbar, sacral, combiner •Studied disease: if there was specific pathology that the study targeted •Studied surgery: if there were particular surgeries or procedures that the study targeted •Study imaging modality: either standard radiograph, dual-energy x-ray absorptiometry (DEXA), CT, MRI including which sequence(s), US, interventional imaging (fluoroscopy, O-arm, guided navigation), combined, or others
(5) Number of subjects and images included	•Overall size: number of subjects, imagesa, or both included in the complete study •Overall diseased subjects: number of patients, images, or both included in the general study diagnosed with at least one spine condition •Categorized size: overall subjects, images, or both included in the general study binned into categories of <100, 100-1000, 1000-10000, 10000-100000, and >100000
(6) Size of the dataset used for DL development and validation	•Dataset size: number of subjects, images, or both included in the DL development phase (when applicable) •Dataset diseased subjects: number of patients, images, or both included in the DL development phase diagnosed with at least one spine condition (when applicable) •Categorized dataset size: dataset subjects: number of patients, images, or both included in the DL development phase (when applicable) binned into categories of <100, 100–1,000, 1,000–10,000, 10,000–100,000, and >100,000
(7) Origin of the dataset	•Either single-center, multicentric, public registry or dataset, synthetic images, or combined
(8) Whether the dataset is publicly available, part of a registry, or institutional data	•Availability and information of origin and access of datasets from publicly available and part of a clinical registry or database datasets
(9) DL method and architecture used	•DL methodology: either CNN, long short-term memory networks (LSTM), Recurrent neural network (RNN), Generative Adversarial Networks (GAN), Radial basis function networks (RBFN), Multilayer Perceptrons (MLP), Self-organizing maps (SOM), deep belief network (DBM), restricted Boltzmann machines (RBM), or other •DL task type: classification, regression, segmentation, object detection, image generation, or other •DL architecture: architecture and backbone family (ex: DenseNet, VGGs, etc) •Number of pipeline(s) and DL architecture(s) used or tested •Other ML techniques used or tested in the study
(10) DL training and validation	•Training: split of the dataset into training, validation, testing, and use of cross-validation •External validation: if external validation of the completed DL pipeline was performed
(12) Evaluation of performances	•Performance metrics used to validate their pipeline •If they used external data to validate their pipeline