Abstract
Dear Editor,
We read with great interest the review by Pontoh et al., which comprehensively presents the expanding role of artificial intelligence in orthopaedic research from preclinical development to clinical application. 1 The authors should be commended for clearly outlining the technological advances and future potential in this rapidly evolving field. However, a critical point that is not sufficiently emphasized in the review is the physician’s holistic evaluation process between diagnosis and treatment decision. This is because, in orthopaedic surgery, there is a clear difference between identifying a pathological finding and translating this finding into a surgical indication, where clinical judgment plays a key role.
Although radiological or biomechanical abnormalities can be clearly defined, the indication for intervention is often shaped by patient-specific factors such as comorbidities, functional expectations, socioeconomic conditions, and overall risk tolerance. It has been shown that even when radiological alignment is acceptable for nonoperative treatment according to established guidelines in distal radius fractures, there is significant variation in treatment decisions among surgeons. 2 This shows that clinical decisions cannot be fully standardized and may vary even when objective criteria are met. These differences reflect the multidimensional nature of clinical decision-making and highlight the important role of both patient and surgeon specific factors.
Current artificial intelligence models are highly effective in pattern recognition and prediction. However, they are inherently limited in capturing the complex context in which clinical decisions are made. The balance between what is technically possible and what is clinically appropriate for an individual patient requires multidimensional clinical judgment. This process is shaped by years of surgical experience and clinical knowledge. This difference is especially important in orthopaedic surgery where indications are not clearly defined and where clinical benefit must be carefully balanced against procedural risk.
As Bernstein noted, the value of an orthopaedic surgeon lies not only in processing information or imaging but also in determining indications through physical examination and applying this knowledge in practice. 3 An algorithm may identify a radiological pathology with high accuracy, but it cannot examine the patient or assess physical findings. At the same time, increasing direct access of patients to artificial intelligence tools may create medico-legal pressure on physicians, as responses generated without clinical context may influence clinical judgment. This may lead to new discussions on malpractice and responsibility in the future. Therefore, it is essential that artificial intelligence outputs do not replace final clinical decisions and that responsibility remains with the physician.
In this context, artificial intelligence-based decision support systems should complement rather than replace clinical judgment. As highlighted by Topol, the integration of artificial intelligence in medicine should strengthen, not weaken, the physician-patient relationship. 4 Preserving the individualized nature of surgical indication is essential for maintaining patient-centered care.
We congratulate the authors for their valuable contribution and believe that emphasizing these points will further strengthen the translational perspective of artificial intelligence in orthopaedics.
