62 papers, newest 30 shown · updated 2026-09-26 · General Orthopaedics · All topics
The newest papers on artificial intelligence in orthopaedics from the orthopaedic journals The Reduction reads, newest first, each with a one-line summary. The list is drawn from the digest’s archive and refreshed every week.
Get new General Orthopaedics papers by email, free: subscribe to The Reduction, focused on your interests, on the days you pick.
A combined CNN and random forest AI model predicted hand surgeons' operative versus nonoperative recommendations for distal radius fractures with 87% accuracy using pre-reduction radiographs and clinical data, showing feasibility for point-of-care decision support.
Multicomponent machine learning approach using multiple registries improves osteosarcoma survival prediction accuracy across diverse populations versus single data-set models.
Machine-learning topic modeling of Spine Deformity's first decade shows growth in fusion outcomes, frailty/perioperative risk, and AI/LLM patient education research, while sagittal alignment and biomechanical topics have cooled.
In a structured hip preservation questionnaire, several large language models (ChatGPT, Gemini, Claude) matched or statistically outperformed expert clinicians in accuracy and consistency, suggesting growing potential as educational adjuncts.
Systematic review of 12 studies (5015 patients) finds machine learning applied to surface topography can estimate scoliosis curve severity with accuracy approaching radiographs, though curve-type classification remains less reliable.
Factors Associated With an LLM Arriving at the Same Diagnosis as a Musculoskeletal Specialist.
Large language models and clinicians disagreed on diagnoses 45% of the time; discordance reflected diagnostic ambiguity rather than patient psychological factors.
A deep learning model analyzing intraoperative arthroscopic images after rotator cuff repair predicted early tendon healing versus retear with high accuracy, potentially helping identify patients needing delayed rehabilitation.
A machine-learning-derived centromere gene signature predicts osteosarcoma prognosis and immune microenvironment status, while B3GALT4 was validated as a tumor suppressor that inhibits osteosarcoma cell proliferation and migration.
Do ChatGPT and Gemini provide accurate answers to patient questions about glenohumeral arthritis?
ChatGPT and Gemini align with AAOS glenohumeral arthritis guidelines only 50–75% of the time and frequently fabricate references, so clinicians and patients should use them cautiously.
The Basic Science of Large Language Models in Orthopaedic Surgery.
This review explains the underlying science of large language models like ChatGPT for orthopaedic surgeons, using clinical examples to illustrate retrieval versus reasoning failures so surgeons can critically evaluate AI tools.
Can Artificial Intelligence Assess Distal Radius Fracture Stability?
AI chatbots (ChatGPT, Claude) showed only fair-to-moderate agreement with hand surgeons in classifying distal radius fracture stability, and are not yet reliable for clinical triage.
This systematic review finds deep learning models show promise for predicting knee OA progression from imaging but suffer from inconsistent outcome definitions, limited external validation, and performance drop-off outside training datasets, limiting current clinical readiness.
An Introduction to Machine Learning for the Practicing Spine Surgeon.
This Clinical Spine Surgery article offers spine surgeons a practical primer on machine learning concepts, model design, and common pitfalls in interpreting AI-driven research.
Artificial Intelligence Scribes in Orthopaedic Surgery: A Narrative Review.
This narrative review evaluates AI documentation scribes for orthopaedic surgery, finding they offer promising EHR-integrated relief from charting burden but need specialty-specific validation before widespread adoption.
Large language models poorly predict ICD-10 codes (24% accuracy) but perform well for CPT codes (92%), requiring optimization before clinical use.
AI-based 3D MRI analysis identifies distinct anatomical patterns in ACL injury including flatter articular surfaces and increased sagittal inclinations.
A time-series/machine learning model using historical, environmental, and system data predicted daily orthopaedic trauma OR caseload far more accurately than rolling averages, offering a template to optimize OR scheduling and staffing.
Machine learning models using pre- and intraoperative data outperformed simple mean-based predictions in forecasting quality-of-life changes two years after adolescent idiopathic scoliosis surgery, supporting their potential use in patient counseling.
3D convolutional neural network accurately classified proximal humerus fractures and characterized greater tuberosity displacement and varus malalignment on CT, with performance comparable to surgeons for these key features.
Machine learning models using baseline patient data predicted post-TKA pain, function, satisfaction, healthcare utilization, mortality, and return to work with moderate-to-strong accuracy, potentially aiding preoperative counseling once externally validated.
Deep learning model predicts femoral mechanical axis from standard antero-posterior knee radiographs with 1.02-degree accuracy, outperforming linear regression and 6-degree varus methods.
Machine learning model predicts acromial and scapular spine fractures after reverse shoulder arthroplasty; medialized-distalized implants, cuff tear arthropathy, older age, and osteoporosis are strongest risk factors.
Machine learning-assisted MRI reveals novel vertebral body biomarkers in chronic low back pain patients including lower signal intensity, greater heterogeneity, and anterior displacement of structural mass.
AI-powered EMR tools like ambient scribes can reduce orthopaedic clinician workload and improve efficiency, though privacy, cost, and integration concerns remain unresolved.
Opinion piece on threats to research integrity from AI misuse, mega-journals, and impact-factor gaming, proposing transparency standards and reformed metrics to safeguard publishing quality.
Large language models achieved ~74% accuracy in pediatric orthopedics with high reading complexity and regional variability, suitable as supervised educational supplements but not independent decision tools.
Unsupervised machine learning identified four distinct biomechanical phenotypes in football players performing 90° change of direction, useful for ACL injury prevention targeting.
Machine learning identified distinct risk factors for moderate versus severe proximal junctional kyphosis after adult spinal deformity surgery: geometric stress for moderate disease and lordosis maldistribution for severe disease.
ChatGPT and Gemini provided reliable, high-quality theoretical information on minimally invasive bunion surgery but failed on actionability and were too complex for general readers.
Machine learning identified optimal age for total hip arthroplasty between 52.5-71.5 years, with lowest risk for readmission, revision, and mortality in this range.
This educational article explains AI, machine learning, deep learning, and cognitive computing concepts to help orthopaedic surgeons understand their potential clinical applications.
Home · All topics · Archive · Subscribe · FAQ
© 2026 The Reduction