Some artificial intelligence tools may currently be useful in orthopedic surgery and traumatology

Abstract

Artificial intelligence (AI) can help in diagnosing fractures and demonstrating effusions, dislocations, and focal bone lesions in both adult and pediatric aged individuals and also aid in early tumor discovery (bone osteosarcoma) and in robot-assisted surgery. A recent AI model [Mask R-CNN (region-based convolutional neural network)] has shown to be dependable for detecting surgical target zones in pediatric hip and periarticular infections, offering a more convenient and quicker alternative to conventional methods. It can help inexperienced physicians in pre-treatment evaluations, diminishing the risk of missed diagnosis and misdiagnosis. AI has some very interesting applications in orthopedic surgery, which orthopedic surgeons should be aware of and if possible use. Although some interesting advances have been made recently on AI in orthopedic surgery, its usefulness in clinical practice is still very limited. Ethical concerns, such as transparency in AI decision-making, data privacy, and the potential loss of human intuition cannot be forgotten. Besides, it is paramount to explore how to gain trust from both healthcare professionals and patients in the utilization of AI.

Key Words: Artificial intelligence; Orthopedic surgery; Pediatric orthopedics; Clinical applications

Core Tip: Artificial intelligence has some very interesting applications in orthopedic surgery (including pediatric orthopedics), which orthopedic surgeons (including pediatric orthopedic surgeons) should be aware of and if possible use. However, although some interesting advances have been made recently on Ai in orthopedic surgery, its usefulness in clinical practice is still very limited.

TO THE EDITOR

The article of Alomran et al[1] about awareness and perceptions among pediatric orthopedic surgeons of artificial intelligence (AI) is of great interest because it shows the absence of familiarity among pediatric orthopedic surgeons regarding AI, and advises better education and regulatory structures to ensure the secure incorporation of AI.

In their paper, Alomran et al[1] recall an article that found that emergency physicians might miss around 11% of acute pediatric fractures compared to specialist pediatric radiologists[2]. Alomran et al[1] also recall that some AI instruments, such as BoneView (Gleamer, Paris, France) can help in diagnosing fractures and uncovering effusion, dislocations, and focal bone lesions in both adult and pediatric individuals[3]. BoneView has shown a 30% decrease in undetected fractures and a 15% reduction in radiograph reading time, marking a considerable advancement over conventional imaging techniques[3].

It is also mentioned in the article by Alomran et al[1] another novelty [OrthoNext digital platform (Orthofix Medical Inc., Lewisville, TX, United States)] for preoperative surgical planning in pediatric orthopedics[4]. Finally, Alomran et al[1] recall other AI applications that can be employed in pediatric orthopedics, such as early bone tumor discovery (osteosarcoma) and robot-assisted surgery[4-6].

With the best intention to reinforce the message of the Alomran et al[1] article, using “artificial intelligence orthopedic surgery 2024” as keywords, I have reviewed the 992 articles published in PubMed from January 1, 2024 to December 26, 2024, finding 15 interesting ones.

In May 2024, Roberts et al[7] published a survey of patient acceptability of the employment of AI in the diagnosis of pediatric fractures. It was found that 76% of the parents preferred a nurse or doctor to evaluate their child's radiographs, 64% were happy for an AI program to diagnose their child's fracture, and 82% were happy with an AI program being utilized as an adjunct to a pediatric orthopedic surgeon's diagnosis.

In May 2024, Zech et al[8] mentioned that an openly attainable AI model substantially improved radiology and pediatric resident accuracy in uncovering pediatric upper limb fractures.

Regarding the possibility of using AI to predict results after some orthopedic surgery interventions, Jang et al[9] noted that most of the published articles had not been externally validated. Furthermore, these studies presented important methodological limitations, which prevent their use in clinical practice.

Concerning the safety of AI, Ghandour et al[10] have recently published (December 2024) a smartphone-based convolutional neural network (CNN) model as a screening tool that allows reliable detection of flatfoot and pes cavus, eliminating the need for assessment of these deformities by classical radiological study. It is an algorithm that integrates a deep CNN into a smartphone camera. In this case, safety seems to be very adequate since it allows not to perform X-rays with the radiation dose that they entail. It is essential that all AI applications are safe.

Regarding the reliability of AI, Sarantopoulos et al[11] have published the important possibilities that AI has in the management of infectious diseases (diagnosis, surveillance, outbreak detection, and treatment). However, they have also mentioned the substantial limitations and challenges that AI still has (data privacy concerns, potential biases, and ethical dilemmas). For AI to gain ground in the field of orthopedic surgery, it is essential that AI applications be reliable.

In June 2024, Liu et al[12] published a model (Mask region-based CNN) of AI-assisted magnetic resonance imaging for the identification of surgical target zones in pediatric hip and periarticular infections. The model recognized osteomyelitis with an accuracy of 0.976 and abscess with an accuracy of 0.957, both statistically better than the four orthopedic surgeons involved in the study. The Mask R-CNN model was dependable for detecting surgical target zones in pediatric hip and periarticular infections. It can help unexperienced physicians in pre-treatment evaluations, diminishing the risk of missed diagnosis and misdiagnosis.

In September 2024, Hasei et al[13] developed a high-performance AI model to uncover osteosarcoma (a primary malignant bone tumor) from radiographs. It is important to remind that osteosarcoma substantially affects children and young adults. The utilization of this AI model could help physicians in early osteosarcoma detection.

In October 2024, Husarek et al[14] demonstrated good diagnostic accuracy across most commercially available AI fracture detection solutions and anatomical areas, with the highest performance accomplished when utilized in conjunction with human evaluation. However, diagnostic accuracy was lower for spine and rib fractures.

Regarding the current state of AI adoption in clinical practice there are some challenges faced by orthopedic surgeons in integrating AI tools into their workflows and the barriers to extensive implementation such as cost, training, and regulatory obstacles. The article by Jang et al[9] focused on the possibility of using AI to predict clinically significant outcomes after some orthopedic surgery interventions (total joint arthroplasties, spine and sport medicine).

Therefore, it is essential to remedy the methodological limitations mentioned above in order to make the application of AI a true reality in clinical practice.

Abi-Rafeh et al[15] have analyzed the potential usefulness of AI for plastic surgeons. They proposed several applications with relevance to different target audiences, including attending plastic surgeons, trainees/educators, researchers/scholars, and patients. However, they also identified important limitations of Chat Generative Pre-Trained Transformer.

Ethical issues (transparency in AI decision-making, data privacy) and public acceptance (potential loss of human touch) are critical to the increasing use of AI in orthopedic surgery.

AI systems in medical fields have significant limitations: They can only be as efficacious as the data upon which they are trained. Insufficient and unrepresentative datasets can result in models that produce false or clinically inconsistent information. For instance, studies have demonstrated that AI models frequently strive with low confidence levels and insufficient validation across diverse populations, highlighting the necessity for larger and more representative datasets to improve predictive accuracy.

Furthermore, the performance of AI algorithms can vary substantially depending on the quality and quantity of training data. Many AI applications still face challenges in matching the diagnostic accuracy of experienced clinicians. This inconsistency emphasizes the importance of addressing data limitations to ensure that AI systems improve rather than impede clinical decision-making[9,11].

Admitting these limitations is paramount for fostering a realistic comprehension of how AI can be efficaciously integrated into clinical workflows while lessening possible risks related to over-reliance on technology.

The articles included in this study have been selected because they show a good balance between the advantages and disadvantages of AI systems in orthopedic surgery. The fact that the information published on pediatrics orthopedics has been commented on is due to the fact that an article on this topic was recently published in the World Journal of Orthopedics[1].

As previously mentioned, the article by Jang et al[9] studied the possibility of using AI to predict results after some orthopedic surgery interventions (total joint arthroplasties, spine and sport medicine).

In conclusion, the article by Alomran et al[1], together with the information contained in this Letter to the Editor, demonstrate that AI has some very interesting applications in orthopedic surgery (including pediatric orthopedics), which orthopedic surgeons (including pediatric orthopedic surgeons) should be aware of and if possible use. However, although some interesting advances have been made recently on AI in orthopedic surgery, its usefulness in clinical practice is still very limited. Ethical concerns, such as transparency in AI decision-making, data privacy, and the potential loss of human touch cannot be forgotten. Besides, it is paramount to explore how to gain trust from both healthcare professionals and patients in the utilization of AI. The current literature does not appear to show that AI systems are used very frequently by orthopedic surgeons.

ACKNOWLEDGEMENTS

My sincere thanks to Leonard A. Valentino, MD, Rush University, Chicago, Illinois, United States, for editing the English of this manuscript.