Artificial Intelligence Enters the Hip Replacement Operating Room as VELYS Navigation System Launches in the U.S.

Total hip arthroplasty has long been a mature procedure, but it still depends heavily on intraoperative interpretation and positioning: pelvic angle, leg length, implant position. Every detail can affect postoperative range of motion, dislocation risk, and whether patients feel that “both legs are the same length.” In this setting, where millimeters and angles matter, artificial intelligence must be applied to concrete tasks if it is to be meaningful, rather than remaining in abstract ideas about diagnosis.

Johnson & Johnson’s DePuy Synthes announced on June 9 that VELYS Hip Navigation with AI Assistance has been commercially launched in the United States. The software is used during total hip arthroplasty to help automatically interpret images and identify anatomical landmarks, giving physicians a more consistent navigation reference when implanting an artificial hip joint.

According to the company, this AI-assisted function can automate image interpretation and landmark identification that previously had to be performed manually, meaning the identification of specific anatomical points in surgical images. These points are an important basis for calculating pelvic orientation, assessing cup position, and tracking changes in leg length; if the identification method is more stable, it could theoretically reduce differences between operators.

Data cited by DePuy Synthes show that the system can provide reproducible landmark identification and reduce workflow time by 57% compared with the manual process. This is a relatively concrete efficiency metric, but the information currently available publicly is still insufficient to assess the study design, sample size, comparison conditions, and clinical endpoints, so it should not be directly interpreted as meaning that patient outcomes have improved by the same magnitude.

More precisely, the core value of this type of surgical AI lies in reducing intraoperative cognitive burden and process variability, not in making medical decisions on its own. Surgeons still need to integrate the patient’s anatomical differences, bone quality, preoperative planning, and real-time tactile feedback; the navigation system provides measurement and alignment assistance, while final responsibility remains with the clinical team.

This launch also shows that medical AI is gradually extending from radiology and pathology interpretation into orthopedic surgical platforms. Compared with the uncertainty of large language models in medical question answering, the task of intraoperative navigation is narrower and its inputs are more controlled, making it easier to engineer and validate; but it likewise needs to confront differences caused by hospital equipment, surgical habits, patient body types, and image quality.

What truly needs to be added next is more transparent clinical data: how the system alerts users when AI identification fails, which populations or anatomical conditions are most prone to errors, and whether efficiency gains translate into fewer complications or more stable implant positioning. For the orthopedic operating room, the threshold for AI is not whether its name is novel, but whether it can continuously provide reliable answers that physicians can correct in a repetitive, high-pressure process that cannot be repeated.