DOI: 10.1177/10225536261504098 ISSN: 1022-5536

AI-powered preoperative planning combined with robotic-assisted execution in total knee arthroplasty: A propensity score–matched analysis of prosthesis sizing accuracy, perioperative outcomes, and early functional recovery

Teng Zhang, Yingdong Hu, Yuyu Fan, Zerui Sun, Jiaxing Wang, Hongxing Song, Qi Yao
Show PDF Cite

Background

Accurate prosthesis sizing and precise surgical execution are critical to the success of total knee arthroplasty (TKA). This study evaluated whether the integration of artificial intelligence (AI)–powered preoperative planning with robotic-assisted execution improves prosthesis sizing accuracy, perioperative parameters, and early functional recovery compared with conventional techniques.

Methods

In this retrospective cohort study, 144 patients undergoing primary unilateral TKA were included (AI-Robotic: 69; Conventional: 75). Propensity score matching (1:1) using eight baseline covariates yielded 56 matched pairs. The primary outcome was prosthesis size prediction accuracy. Secondary outcomes included osteotomy time, blood loss, radiographic alignment, and functional recovery assessed by the Visual Analog Scale (VAS), knee range of motion (ROM), and the Hospital for Special Surgery (HSS) score. Matched pairs were analyzed using paired t-tests and McNemar’s test, with doubly robust regression as a sensitivity analysis.

Results

After matching, all baseline covariates were well balanced (standardized mean differences < 0.1). The AI-Robotic group demonstrated significantly higher femoral component sizing accuracy (75.0% vs. 58.9%, P = 0.008), whereas the improvement in tibial component accuracy did not reach statistical significance (82.1% vs. 69.6%, P = 0.146). Osteotomy time was significantly shorter in the AI-Robotic group (−2.26 min, 95% CI −3.98 to −0.55; P = 0.011), as were intraoperative blood loss (−27.4 mL, 95% CI −48.2 to −6.5; P = 0.011) and total postoperative blood loss (−123.6 mL, 95% CI −221.1 to −26.1; P = 0.014). The lateral tibial component angle (LTC) was more precisely controlled in the AI-Robotic group (+0.76°, 95% CI 0.11 to 1.41; P = 0.022), while other radiographic parameters did not differ. No statistically significant differences were observed between groups in VAS pain scores, ROM, or HSS scores at any time point up to six months.

Conclusions

These findings suggest that the integration of AI-powered preoperative planning with robotic-assisted TKA was associated with improved femoral component sizing, shorter osteotomy time, reduced blood loss, and finer control of tibial sagittal alignment. However, these procedural advantages did not translate into measurable differences in early postoperative pain, range of motion, or functional scores. The clinical value of this integrated technology may lie primarily in optimizing the technical execution of TKA rather than in accelerating short-term functional recovery.

More from our Archive