Sage Journals: Discover world-class research

Abstract

Generative artificial intelligence (AI) is a powerful class of machine learning that moves beyond simply analysing data to actually creating new and original content, such as medical images or clinical text. The use of generative AI is varied in orthopaedic surgery. Generative AI moves us from one-size-fits-all surgical planning to highly personalised surgical blueprints for each patient’s unique anatomy and condition. While generative AI in surgery is new, it can provide real-time intelligent help to a surgeon’s skill and decision-making. Most practitioners see the use of AI as a tool to improve diagnosis and treatment, with some expressing their concern that it will conversely worsen diagnosis and treatment. With its use and potential, the use of generative AI currently should be supervised and validated, as it has been shown that sometimes the generated content does not reference to any actual source. Policies and economic values are also detrimental to the integration of AI technologies in clinical orthopaedics. Ethical issues, practitioners view and perspective, and the high overall cost of AI technology use, are among the barriers that may emerge. This comprehensive review addresses the opportunities, challenges, and future direction of integrating generative AI in orthopaedic surgery.

Keywords

artificial intelligence generative AI orthopaedic surgery

Introduction

Generative artificial intelligence (AI) is a powerful class of machine learning that moves beyond simply analysing data to actually creating new and original content, such as medical images, clinical text or 3D anatomical models. Orthopaedics is a specifically promising field for its application, as clinical practice can heavily relies on imaging, surgical planning, decision making and patient-specific solutions. These are all areas where generative AI may add its value. For instance, creating 3D models from standard x-rays, improving resolution of CT or MRI images, or generating virtual reality images of a patients’ data are applications relevant to diagnosis, surgical preparation, and rehabilitation.¹

This generative power is harnessed through several sophisticated technologies.¹ A worldwide study by Familiari et al. found that 25.1% of orthopaedic practitioner used AI in their practice. Most view it as beneficial, especially in referencing literature and aiding communication.² AI and natural language processing have been utilised in evaluating treatment success. Floyd et al. developed an AI model evaluating clinical notes to evaluate outcomes on treatment after proximal humerus fracture, which provide them with the analysed patient’s outcome based on each of their flexible and responsive individual treatment goals. With AI models helping practitioners achieving tasks as they intended, the modifiable nature of this technology allows flexibility, however guidelines and standards, as well as accuracy and validity test are still need to be implemented to ensure the quality of outcome.³ However, adoption of the use of AI remains limited by uncertainty about accuracy, ethical considerations, and cost. To fully realise its premise, generative AI necessitate validation and standardisation before routine application or integration into clinical practice.

Technically, generative AI employs methods such as adversarial training, probabilistic encoding, iterative noise reduction, and large language processing. These methods are made into models, known as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Diffusion Models and Large Language Models (LLMs), which require trainings to generate desired contents. While the details might be different, the common feature is the ability of synthesising new and realistic data rather than solely predicting the outcomes. These capabilities distinguish generative models from predictive AI, which is primarily concerned with forecasting clinical risks or prognosis of a procedure or disease.^4,5

This creative ability highlights a critical difference from traditional predictive AI. Predictive models look to the past to forecast the future, answering the clinical question, “What is the most likely outcome?” In contrast, generative models invent new possibilities, asking, “What is an optimal solution we could create?” This is more than a technical distinction; it signals a potential evolution in medicine from a reactive stance on risk management to a proactive approach centred on personalised solution design.⁶

To support clarity for readers less familiar with AI, Table 1 summarises key terms and provides examples specific to orthopaedic surgery.

Table 1.

Key terms in artificial intelligence with examples in orthopaedic surgery.

Term	Definition	Example application in orthopaedic surgery
Artificial intelligence (AI)	Broad field of computer systems designed to perform tasks requiring human intelligence.	AI-based decision support suggesting treatment for massive rotator cuff tears (repair vs reverse shoulder arthroplasty).
Machine learning (ML)	Subset of AI where algorithms learn patterns from data to make predictions or decisions.	Predicting postoperative stiffness after arthroscopic capsular release using patient and intraoperative factors.
Deep learning (DL)	A type of ML using layered neural networks to analyse complex data such as images or speech.	Automated detection of glenoid bone loss or cuff tears on MRI/CT with convolutional neural networks.
Predictive AI	Uses past data to forecast outcomes.	Estimating risk of dislocation after reverse shoulder arthroplasty based on implant positioning and anatomy.
Generative AI	Creates new data rather than only analysing existing data.	Generating 3D digital twins of scapula and humerus to optimise implant placement in shoulder arthroplasty, or synthetic MRI datasets for cuff tear training.

Opportunities in orthopaedic clinical practice

Generative AI models create new synthetic data by learning from given data sets.^4,7 The use of generative AI is varied in orthopaedic surgery.^7,8 Most of the current clinically applicable use of AI are robotic-assisted surgery, prediction measurements and communication aid.^3,7,9 Table 2 shows previous studies regarding the use of generative AI in orthopaedic surgery.

Table 2.

Generative artificial intelligence use in orthopaedics.

Author	Year	Title	Subject	Methods	Outcome	Clinical applicability
Hernigou et al.⁹	2021	Digital twins, artificial intelligence, machine learning technology to identify a real personalised motion axis of the tibiotalar joint for robotics in total ankle arthroplasty	20 ankle digital twins (originated from 3-D models of 5 patients)	Analysing ankle kinematics in digital twins with the help of AI and machine learning	Several instantaneous centres of rotation occur as the tibia moves across the dome of the talus, with 6 mm maximum distance between two different centres and 20° possible axes variation of different ankles	Clinically adopted
Ahn et al.¹⁰	2023	High-resolution knee plain radiography image synthesis using style generative adversarial network adaptive discriminator augmentation	10,000 anteroposterior (AP) radiography of the knee	Image generation with deep convolutional GAN (DCGAN) and style GAN adaptive discriminator augmentation (StyleGAN2-ADA)	34%–50% classification accuracy of the experts. 2.96 fréchet inception distance, no significant difference in the principal component analysis between real and generated images	Early experimental
Agharia et al.¹¹	2024	The ability of artificial intelligence tools to formulate orthopaedic clinical decisions in comparison to human clinicians: An analysis of ChatGPT 3.5, ChatGPT 4, and bard	8 clinical cases with 97 questions	ChatGPT 3.5: July 20 version; ChatGPT 4: July 20 version; and bard: V2023.07.13. Analysis on clinical cases	ChatGPT 3.5, ChatGPT 4.0 and bard with 7.2%, 0.0% and 3.1% refusal rate; 40.2%, 68.0% and 45.4% proportion of response	Early experimental
Isleem et al.¹²	2024	Can generative artificial intelligence pass the orthopaedic board examination?	301 Orthopaedic board exam style-questions	ChatGPT 3 was used to answer the question, in a new chat for each question	60.8% correct answer	Early experimental
Kim et al.¹³	2024	Clinical validation of enhanced CT imaging for distal radius fractures through conditional Generative adversarial networks (cGAN)	CT scans of 8 distal radius fracture patients	cGAN was used to improve the resolution of 3 mm CT images to match that of 1 mm images (CT enhancement)	Quantitative: cGAN with training (method 2) generated higher quality resolution, in PSNR, MSE metrics (85.1% vs 14.9%), and SSIM (63.6% vs 36.4%) compared to untrained GAN (method 1). Qualitative: Method 2 (grade 2.2) showed better quality images compared to method 1 (grade 2.7)	Promising
Miskiewicz et al.¹⁴	2024	Enhancing access to orthopaedic education: Exploring the potential of generative artificial intelligence (AI) in improving health literacy on rotator cuff injuries	PEMs about RCI from 25 orthopaedic hospitals	ChatGPT was asked to rewrite the PEMs by incorporating several key guideline recommendations made by the CDC and NIH to an eighth-grade reading level	The revised PEMs exhibited a more concise format, with an average word count of 312.2 ± 65.73, showing a marked decrease in length	Early experimental
Khosravi et al.⁷	2024	Analysing racial differences in imaging joint replacement registries using generative artificial intelligence: Advancing orthopaedic data equity	480,407 Pelvic radiographs from large institutional registry of total hip arthroplasty patients	Denoising diffusion probabilistic models generated radiographs conditioned on demographic and imaging characteristics	African american patients, when compared to white patients, demonstrated decreased interacetabular distance (GAC: 0.83; p value <0.001)	Clinically adopted
Familiari et al.²	2025	Exploring artificial intelligence in orthopaedics: A collaborative survey from the ISAKOS young professional task force	211 orthopaedic surgeons	Online survey of feedback on the use of AI in the orthopaedic field	52.1% lack of differentiation between AI, ML, DL. 25.1% using AI. 95% leveraging AI technologies.	n/a
Floyd et al.³	2025	Using artificial intelligence to develop a measure of orthopaedic treatment success from clinical notes	868 clinical notes of patients with treated PHF	Develop a clinical text classifier using AI approaches to identify patient’s treatment outcome	One of the deep learning models (Bio-ClinicalBERT) achieve the highest accuracy (0.87 ± 0.04), amongst the bag-of-words models, the highest accuracy was achieved by XGB classifier (0.84 ± 0.03).	Clinically adopted
Zhao et al.¹⁵	2025	Conditional GAN performs better than orthopaedic surgeon in virtual reduction Of femoral neck fracture	208 pre-surgical CT scans with significantly displaced femoral fracture (garden type III to IV)	Virtual reduction of femoral fracture using trained cGAN	Pauwels angles of reduction by cGAN was more stable (p < 0.001) than manual reduction. Garden indices in manual reduction was significantly lower than cGAN (p = 0.002). The satisfied reduction rate of cGAN was significantly lower than manual reduction (88.372% vs 53.448%, p < 0.001)	Early experimental
Nian et al.¹⁶	2025	Pediatric supracondylar humerus and diaphyseal femur fractures: A comparative analysis of chat generative pretrained transformer and google gemini recommendations versus american academy of orthopaedic surgeons clinical practice guidelines	2 pediatric fracture-specific CPGs available from AAOS (13 evidence-based recommendations)	Each recommendation was converted into a final chatbot prompt, then provided to ChatGPT-4.0, ChatGPT-3.5, and google gemini	ChatGPT-4.0, ChatGPT-3.5, and google gemini. ChatGPT-4.0 provided 11 accurate responses (84.6%), ChatGPT-3.5 provided 9 accurate responses (69.2%), and google gemini provided 11 accurate responses (84.6%). There was no significant difference in accuracy (P = 0.533) among the 3 chatbots	Promising
Eravsar et al.¹⁷	2025	Is ChatGPT a more academic source than google searches for patient questions about hip arthroscopy? An analysis of the most frequently asked questions	Response to the questions containing the terms “HA,” “HA surgery,” “hip labral repair,” or “femoroacetabular impingement”	ChatGPT’s responses compared to google searches for hip arthroscopy	ChatGPT provided significantly more academic references than google web search (93% vs. 46%, p-value <0.05).	Early experimental

AI: artificial intelligence; RCI: rotator cuff injury; cGAN: conditional generative adversarial nertworks; PEM: patient educational materials; CDC: Centers for Disease Control and Prevention; NIH: National Institutes of Health (NIH); GAC: Gwet’s AC1; ML: machine learning; DL: deep learning; PHF: proximal humerus fracture; HA: hip arthroplasty.

Preoperative phase: Enhancing diagnostic precision and surgical planning

Generative imaging is changing how orthopaedic surgeons see and understand patient anatomy. One of the most powerful applications is the ability to create a “digital twin” of a patient’s anatomy which is a rich interactive tool for diagnosis and planning. This starts with improving and transforming standard medical images.⁵ Using models like GANs and diffusion models you can turn low resolution or low dose scans into high quality detailed images. This improves diagnostic accuracy by showing subtle problems that would otherwise be missed and reduces patient radiation exposure without losing image quality. These models can also algorithmically reconstruct parts of an image that are missing or unclear and give a more complete picture for clinical evaluation.¹⁸

A major innovation with big clinical benefits is the ability to automatically create 3D anatomical models from standard 2D images. For example, studies have used GAN-based models to generate accurate 3D models of the human spine from a pair of regular X-rays. This gives surgeons a detailed volumetric view of complex anatomy that was previously only possible with more expensive and higher radiation methods like CT scans. This could make advanced anatomical visualisation more widely available especially in places with only basic X-ray equipment.¹⁹ Beyond improving existing images, generative models can create entirely new synthetic X-rays. Research has shown that GANs can generate high resolution synthetic knee X-rays that show the stages of arthritis and are so realistic that even surgeons and expert radiologists can’t tell them apart from real images, with accuracy percentage between 34% - 50% according to Ahn et al.¹⁰ These synthetic images can be used to expand limited datasets for research, train new clinicians and develop and test other AI diagnostic tools without risking patient privacy.^10,20

Personalised surgical blueprints

Generative AI moves us from one-size-fits-all surgical planning to highly personalised surgical blueprints for each patient’s unique anatomy and condition. By processing a patient’s CT or MRI, AI algorithms can quickly and automatically create detailed, interactive 3D models of bones and surrounding tissues. These models let surgeons see complex fractures, tumours or deformities from any angle, measure important parameters and develop a precise individual surgical plan.²¹ The 3D models created by AI are the foundation for surgical simulations. Using virtual reality (VR) or augmented reality (AR), surgeons can “rehearse” a procedure on a patient’s digital twin. They can practice surgical techniques, try different approaches, anticipate problems and optimise implant placement all in a risk-free virtual environment. This is especially valuable for complex procedures like joint replacement, bone tumor removal and corrective spinal surgery where it has been shown to increase surgical success and reduce operating time^22,23 Generative AI can create custom orthopaedic implants. By analysing a patient’s specific anatomical data – bone density, geometry and movement patterns – AI algorithms can design prosthetics and implants that fit like a glove.²⁴ This level of personalisation will increase implant lifespan, patient comfort and better functional outcomes.^24–26

Conditional GAN (cGAN) utilises deep learning, where the creation of a target image from an input image is trained through an algorithm to construct a realistic visualisation, with a better quality from the traditional raw data-based approaches.^13,15 A study by Kim et al. showed the use of trained model of cGAN results a better quality in terms of resolution and performance, compared to the untrained model of GAN. However, the database used for training affects the outcome quality, and analysing unseen data may become its weakness in generating the intended images. Producing high quality scans with this technology provides better understanding on patient’s clinical anatomy, which in turn will results in better assessment and surgical outcomes.¹³ In a study by Zhao et al., virtual reduction of femoral fracture by trained cGAN provide better satisfactory rate compared to manual reduction by orthopaedic surgeon. Previously prepared virtual reduction of fracture may act as a target goal in the actual open reduction, and may be embedded in virtual reality glasses, assisting operation in real time.¹⁵

Intraoperative phase: Augmenting surgical execution

While generative AI in surgery is new, it can provide real time intelligent help to a surgeon’s skill and decision making. Here we focus on dynamic adjustments and better navigation. AI-powered systems can improve surgical navigation by combining preoperative 3D models with real time intraoperative imaging (like fluoroscopy or ultrasound). This creates a live augmented view of the surgical field, so the surgeon can track their instruments more precisely to the patient’s anatomy. In robotic assisted surgery which is becoming more common in orthopaedics, AI algorithms can analyse data from sensors and cameras to adjust the surgical plan in real time. This allows the system to account for small changes in patient position or tissue movement, so the robot can be safer and more accurate in actions like cutting bone or placing screws. These systems can also provide real time decision support, for example by analysing intraoperative data to predict the stability of an implant and suggest to the surgeon.^21,27

In the recent time, AI usage is not preferentially seen to significantly help in improving surgical outcome, although the use of AI robotics for surgical procedure are emerging, and more commonly, practitioners have been using AI to support diagnosis and intraoperative assessment based on image.² AI technology had been successfully utilised by Hernigou et al. in analysing 3-D models to determine patient’s precise and individual ankle real axis of motion, which aids the process of axial alignment of talar implant in total ankle arthroplasty.⁹ Kim et al. suggested the use of cGAN to increase the image quality of the recently emerging C-arm CT, which resolution and performance are often inadequate in terms of clinical use.¹³

Postoperative phase: Personalising recovery and rehabilitation

After surgery, generative AI can create more personalised and closely monitored recovery plans to optimise outcomes and reduce complications. Instead of a one-size-fits-all protocol for everyone, AI can create a dynamic schedule of exercises tailored to an individual’s needs, range of motion and muscle strength. By looking at patient-specific data – type of surgery, pre-op functional levels and real-time data from wearable sensors, generative AI can create custom rehab programs. AI-guided gait analysis using cameras or wearable sensors can track a patient’s progress.^28,29 However, these rehabilitation sequences still require supervision by physicians.³⁰ By looking at movement data these systems can predict functional problems and recovery speed, so therapists can adjust the rehab plan in real time.^28,29

Generative language model such as ChatGPT can help answer patient’s postoperative concern. Dubin et al. found that ChatGPT and arthroplasty trained nurse provide similarly appropriate answer regarding common total hip arthroplasty postoperative questions. Most patients do not trust ChatGPT in answering their postoperative question, however when the answers are given blinded, they were more comfortable in the answers given by ChatGPT.³¹ Another study by Gök et al. compared ChatGPT-4.0 and DeepSeek-V3 performance in answering commonly asked questions regarding total knee arthroplasty, and it was found that ChatGPT-4.0 provide a more descriptive and detailed content in its answers, even though they are similar in scientific accuracy and understandability.³²

Transforming the landscape of orthopaedic research and education

LLMs are now trained to generate sophisticated human-like answers to prompts.^17,33 ChatGPT, when compared to google search engine, generates more academic responses (93% vs 46%).¹⁷ Even when given prompt in accordance to the American Academy of Orthopaedic Surgeons clinical practice guidelines (AAOS CPGs), Nian et al. found that the newest LLMs at that time can deliver responses with no significant difference to the existing guidelines.¹⁶ Holland et al. evaluates abstracts constructed by ChatGPT-4.0 and ChatGPT-3.5 compared to the ones written by a surgical resident and a senior author where surgeon reviewers found that the AI-generated abstracts are subjectively ranked better than abstracts written by human. Current trustworthy uses of these chatbots are limited, they showed their ability in translation and organising hospital electronic medical records but often spreads misinformation from falsified unknown-sourced data and making biased analysis. These LLMs are lacking creativity as they are dependent on the given prompt, often requiring multiple modification until satisfying answers are generated. Validation should be a compulsory step if LLMs are utilised to generate scientific text.³³

Navigating the challenges and limitations

Generative AI may perform artificial hallucination while generating contents, which concerns the possibility of the technologies providing its user with fabricated information. Generative LLMs for example, often delivers their answers with definitive statements, misleading users while the answer does not actually based on any actual reference material.¹² Quality of the learning source and different learning models determines the quality of analysis and eventually generated outcomes, often falters when encountering unseen data. Different intent of generative AI requires different algorithm model, which affects the type of source material usually comprises of datasets. Technique regularisation, improvement of datasets, and transfer learning could be the potential solver for this problem.¹³

Concerns regarding the ethics of AI application presents on the social and moral aspects. The instance of generative AI language model in giving incorrect answers that sounds plausibly correct is one of the factors that hinder the use of AI, making it unreliable to use in practice.³⁴ Different models of AI, even in analysing the same dataset, can exert different quality of outcome as shown by Floyd et al., hence algorithm bias, data quality, and interpretation should also be considered in integrating AI in clinical practice.^3,35

Policies and economic values are also detrimental in the integration of AI technologies in clinical orthopaedics.³⁴ Availability of AI technologies are affected by cost and regulation. Insurance policy commonly does not cover the use of AI technologies.² The value of AI technologies in the market keeps expanding year-to-year, mirroring how the developments are constantly being made.¹²

The reliability of generative AI technology shows inconsistency, which should not be seen if they are to be implemented in daily practice.¹¹ With a popular model of generative AI, ChatGPT, Agharia et al. found that the model’s ability in generating decision for orthopaedic patients are non-consistent compared to practitioner.^11,36 A similar study by Isleem et al. found that ChatGPT were only able to correctly answer 60.8% of the 301 orthopaedic board exam-style question, which is similar to the score of interns and junior resident, with limitation in answering complex question which requires intuition.¹² Practitioner’s will in utilising AI is also a factor in AI adoption. Training is needed in integrating novel technology to clinical practice, which often expensive and time-consuming, possibly making it unappealing to practitioners with older age. Knowledge regarding AI technologies is not well understood by some, with practitioners showing confusion regarding the difference of AI, machine learning (ML) and deep learning (DL).²

Future directions: Toward an autonomous and ethical orthopaedic ecosystem

Most practitioner sees the use of big data as a tool to improve diagnosis and treatment, a minority expressed their concern that it will conversely worsen diagnosis and treatment. The view on the future of AI in orthopaedics is still ambiguous, where its potential in image-based diagnosis, especially in osteoarthritis, and patient outcome prediction are eagerly expected, while regulation, cost–value concerns, and scarcity of evidence-based applications remain barriers.² Questions of legal liability remain unresolved, as accountability for poor outcomes involving AI-assisted planning is unclear, and bias in datasets risks perpetuating inequities if underrepresented populations are not adequately included. AI learning and algorithm can be embedded in various types of technologies. The implementation of generative AI in low-quality imaging machine was suggested to be able to enhance image quality, potentially lowering cost from the need of repetitive retake scans required for diagnosis confirmation.¹³ Together, these challenges emphasize why enthusiasm for generative AI is tempered by caution within the orthopaedic ecosystem.

Conclusion

Generative AI is frequently used in the orthopaedic settings. With continuous development to the technologies, the use of generative AI will potentially aid practice, even during surgical procedures. Robotic-assisted surgery, predictive measurements, and communication aid is the currently clinically applicable generative AI technology. The potential of generative AI in synthesising new data in imaging is promising, however supervision and validation are still needed. There are also concerns regarding the ethical, legality, reliability, and cost of generative AI. Future studies in AI should focus on enhancing the accuracy, quality, and clinical applicability of generative AI.

Footnotes

ORCID iDs

Felly Liu

Joshua Edward Hananto

Siti Zainab Bani Purwana

Erica Kholinne

Ismail Hadisoebroto Dilogo

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Templin

Perez

Sylvia

, et al. Addressing 6 challenges in generative AI for digital health: a scoping review. PLOS digital health 2024; 3(5): e0000503.

Familiari

Saithna

Martinez-Cano

, et al. Exploring artificial intelligence in orthopaedics: a collaborative survey from the ISAKOS young professional task force. J Exp Orthop 2025; 12(1): e70181.

Floyd

Almeldien

Smith

, et al. Using artificial intelligence to develop a measure of orthopaedic treatment success from clinical notes. Front Digit Health 2025; 7: 1523953.

Kunze

. Generative artificial intelligence and musculoskeletal health care. HSS J 2025; 21: 15563316251335334.

Musalamadugu

Kannan

. Generative AI for medical imaging analysis and applications. Future Medicine AI 2023; 1(2): FMAI5.

Sharma

Lysenko

Jia

, et al. Advances in AI and machine learning for predictive medicine. J Hum Genet 2024; 69: 487–497.

Khosravi

Rouzrokh

Erickson

, et al. Analyzing racial differences in imaging joint replacement registries using generative artificial intelligence: advancing orthopaedic data equity. Arthroplast Today 2024; 29: 101503.

Kumar Yadav

Joshi

Tiwari

, et al. From data to precision: the transformative role of AI and machine learning in modern orthopaedic practice. J Clin Orthop Trauma 2025; 69: 103101.

Hernigou

Olejnik

Safar

, et al. Digital twins, artificial intelligence, and machine learning technology to identify a real personalized motion axis of the tibiotalar joint for robotics in total ankle arthroplasty. Int Orthop 2021; 45(9): 2209–2217.

10.

Ahn

Choi

, et al. High‐resolution knee plain radiography image synthesis using style generative adversarial network adaptive discriminator augmentation. J Orthop Res 2023; 41(1): 84–93.

11.

Agharia

Szatkowski

Fraval

, et al. The ability of artificial intelligence tools to formulate orthopaedic clinical decisions in comparison to human clinicians: an analysis of ChatGPT 3.5, ChatGPT 4, and bard. J Orthop 2024; 50: 1–7.

12.

Isleem

Zaidat

Ren

, et al.

Can generative artificial intelligence pass the orthopaedic board examination?

J Orthop 2024; 53: 27–33.

13.

Kim

Ryu

Keum

, et al. Clinical validation of enhanced CT imaging for distal radius fractures through conditional generative adversarial networks (cGAN). PLoS One 2024; 19(8 August): e0308346.

14.

Miskiewicz

Perez

Capotosto

, et al. Enhancing access to orthopedic education: exploring the potential of generative artificial intelligence (AI) in improving health literacy on rotator cuff injuries. Cureus 2024; 16: e72833.

15.

Zhao

Mei

Wang

, et al. Conditional GAN performs better than orthopedic surgeon in virtual reduction of femoral neck fracture. BMC Muscoskelet Disord 2025; 26(1): 687.

16.

Nian

Umesh

Simpson

, et al. Pediatric supracondylar humerus and diaphyseal femur fractures: a comparative analysis of chat generative pretrained transformer and google gemini recommendations versus American academy of orthopaedic surgeons clinical practice guidelines. J Pediatr Orthop 2025; 45(4): e338–e344.

17.

Eravsar

Aydin

Eryilmaz

, et al. Is ChatGPT a more academic source than google searches for patient questions about hip arthroscopy? An analysis of the Most frequently asked questions. Journal of ISAKOS 2025; 12: 100892.

18.

Guermazi

Omoumi

Tordjman

, et al. How AI May transform musculoskeletal imaging. Radiology 2024; 310(1): e230764.

19.

Yang

Lin

Wang

, et al. Generative adversarial network (GAN) for automatic reconstruction of the 3D spine structure by using simulated bi-planar X-ray images. Diagnostics 2022; 12(5): 1121.

20.

Segal

Rubin

, et al. Evaluating the clinical realism of synthetic chest X-Rays generated using progressively growing GANs. SN Comput Sci 2021; 2(4): 321.

21.

Kumar

Sporn

Ong

, et al. Integrating artificial intelligence in orthopedic care: advancements in bone care and future directions. Bioengineering 2025; 12(5): 513.

22.

Ghaednia

Fourman

Lans

, et al. Augmented and virtual reality in spine surgery, current applications and future potentials. Spine J 2021; 21(10): 1617–1625.

23.

Henderson

Van Schuyver

Economopoulos

, et al. The use of artificial intelligence for orthopedic surgical backlogs such as the one following the COVID-19 pandemic: a narrative review. JBJS Open Access 2024; 9(3): e24.

24.

Tripathi

Hajra

Mulukutla

, et al. Artificial intelligence in biomedical engineering and its influence on healthcare structure: current and future prospects. Bioengineering 2025; 12(2): 163.

25.

Filardi

. The evolution and impact of customized implants with intricate designs. In: Vaishya

Ghosh

(eds). Biomaterials in Orthopaedics & Trauma: Current Status and Future Trends in Revolutionizing Patient Care [Internet]. Springer Nature Singapore, 2025, pp. 291–307.

26.

Luo

. Toward fully automated personalized orthopedic treatments: innovations and interdisciplinary gaps. Bioengineering (Basel) 2024; 11: 817.

27.

Smits

Cotic

Hinterwimmer

, et al. Potenzial von Wearable-Technologie in der Knieendoprothetik. Ortho 2024; 53(11): 858–865.

28.

Liao

Lee

Chiang

, et al. Postoperative rehabilitation after anterior cruciate ligament reconstruction through telerehabilitation with artificial intelligence brace during COVID-19 pandemic. J Clin Med 2023; 12(14): 4865.

29.

Yoshikawa

Kanzaki

Matsuno

, et al. Gait parameters and postoperative pain following total ankle arthroplasty using wearable motion sensors: a cross-sectional study. Cureus 2025; 17(2): e78392.

30.

Ganti

VKAT

Pandugula

Polineni

TNS

, et al. Transforming sports medicine with deep learning and generative AI: personalized rehabilitation protocols and injury prevention strategies for professional athletes. Review of Contemporary Philosophy 2023; 22(1): 3868–3882.

31.

Dubin

Bains

DeRogatis

, et al. Appropriateness of frequently asked patient questions following total hip arthroplasty from ChatGPT compared to arthroplasty-trained nurses. J Arthroplast 2024; 39(9): S306–S311.

32.

Gök

Pamuk

Serttaş

, et al. ChatGPT-4.0 or DeepSeek-V3? Comparative analysis of answers to the Most frequently asked questions by total knee replacement candidate patients. Medicine 2025; 104(34): e43951.

33.

Holland

Lorenz

Cavanagh

, et al. Comparison of medical research abstracts written by surgical trainees and senior surgeons or generated by large language models. JAMA Netw Open 2024; 7(8): e2425373.

34.

Farina

Lavazza

. Ethical considerations and policy interventions concerning the impact of generative AI tools in the economy and in society. AI and ethics. AI Ethics 2025; 5(1): 737–745.

35.

Tariq

Gill

Hussain

. Evaluating the potential of artificial intelligence in orthopedic surgery for value-based healthcare. International Journal of Multidisciplinary Sciences and Arts 2023; 2: 1–9.

36.

Moritz

Romeike

Stosch

Tolks

. Generative AI (gAI) in medical education: Chat-GPT and co. GMS J Med Educ 2023;40(4):Doc54.