To investigate whether the accuracy of artificial intelligence (AI)-assisted preoperative planning for total knee arthroplasty (TKA) is non-inferior to computer-assisted manual preoperative planning completed collaboratively by 3D engineers and orthopedic surgeons, and to evaluate the impact of age on the accuracy of AI-based planning.

This study adopted a retrospective, cross-sectional comparative design. Data were retrospectively collected from 127 patients who underwent TKA at the First Affiliated Hospital of Army Medical University between December 2022 and September 2023. They were divided into a ≤60 years group (40 to 60 years old, n=22) and an >60 years group (61 to 80 years old, n=105) based on age. All patients underwent both AI-assisted preoperative planning (AI planning) and computer-assisted manual preoperative planning (conventional planning). Hip-knee-shaft angle (HKS), femoral resection difference, tibial resection difference, femoral rotation angle, and prosthesis size data from AI planning, conventional planning, and actual intraoperative measurements, as well as time, manpower, and economic costs of AI planning and conventional planning were compared between the 2 planning methods. The noninferiority margin was set as resection angle difference ≤1°, resection thickness difference ≤1 mm, and prosthesis size difference ≤1.

Compared with the conventional planning, the AI planning showed a mean increase of 0.8°±1.2° in HKS (Cohen’s d=0.4, 95%CI: 0.2 to 0.7), a mean increase of 1.1±1.8 mm in femoral resection difference (Cohen’s d=0.5, 95%CI: 0.3 to 0.8), and a mean increase of 0.6±1.7 mm in tibial resection difference (Cohen’s d=0.3, 95%CI: 0.0 to 0.5). The lower limits of 95%CI for these above standardized mean differences were all greater than their corresponding negative non-inferiority margins. In comparison with the conventional planning, the AI planning in the ≤60 years group showed no significant differences in HKS, tibial resection difference, and femoral rotation angle (P>0.05), whereas in the >60 years group, only femoral resection difference of AI planning exceeded the non-inferiority margin (with a mean increase of 1.1 mm, P<0.05). For prosthesis size prediction, the AI planning achieved 73.2% femoral prosthesis and 87.4% tibial prosthesis predictions with size differences not exceeding 1, while conventional planning achieved 81.9% femoral prosthesis and 90.6% tibial prosthesis predictions with size differences not exceeding 1, with no significant difference between the 2 planning methods (P>0.05). The time consumed for AI planning was significantly shorter than that for conventional planning (27.4±4.1 vs 45.3±8.5 min, P<0.05). Moreover, AI planning required no additional labor and equipment costs, with no additional manpower or equipment costs (0 vs 3 000 Yuan/case).

AI-assisted preoperative planning for TKA s demonstrates non-inferior accuracy compared with conventional planning, with superior performance particularly in patients under 60 years, and significant advantages in time and economic costs.

To determine whether an intelligerct cloud platform follow-up model demonstrates superiority over conventional Methods in enhancing follow-up quality, reducing costs, and improving functional recovery following total knee arthroplasty(TKA).

A retrospective cohort study was conducted on 260 patients undergoing TKA at our hospital from October 2022 to September 2023. According to the different postoperative follow-up methods, they were divided int a cloud platform group (n=130) and a traditional group (n=130). The baseline data, pre- and postoperative knee function scores, and follow-up costs were collected and compared between the 2 groups.

The cloud platform group exhibited significantly higher rate of follow-up satisfaction, reduced resource utilization, shorter per-follow-up duration, fewer accompanying persons per visit, lower total follow-up expenditures, and decreased cumulative follow-up time when compared with the traditional group (P<0.05). At 6 weeks, 3, 6, and 12 months postoperatively, better range of motion (ROM) was observed in the cloud-based group than the traditional group (P<0.05), but no intergroup differences were seen in other knee function scores.

Cloud platform follow-up enhances early postoperative ROM (6 weeks to 12 months) and demonstrates marked cost-effectiveness when compared to the conventional mode. It represents a viable alternative for post-TKA rehabilitation surveillance.

To construct a prediction model for coronal malalignment of lower limb in middle-aged and young people in China based on body surface big data in order to provide a faster and more accurate tool for predicting the malalignment in clinical practice.

A cross-sectional trial was adopted on 915 patients with knee meniscus tears admitted to the Sports Medical Center of our hospital from May 2022 to December 2023. The coronal force line of lower limb was measured, and according to the lower limb force line grading standards, the patients were divided into neutral force line group and malalignment lower limb group, and assigned randomly into training set and validation set in a ratio of 7∶3. Seven indicators, such as gender, age, and body surface big data (including BMI, lower limb length, distance between both knee joints, distance between both ankle joints, and subcutaneous fat thickness) were used to analyze the training set to predict the value of malalignment force line. Logistic regression model and nomogram model were constructed to visualize our prediction model. Then calibration curves, receiver operating characteristic (ROC) curve, and decision curve analysis (DCA) were applied to evaluate the diagnostic efficacy of the constructed model.

In the training set of 640 cases, there were 299 males and 341 females, with a median age of 41.5 years old, and for the validation set, there are 275 patients, including 128 males and 147 females, with a median age of 41.0 years old. Significant differences were observed in above mentioned 7 indicators between the 2 groups in the training set (P<0.01). Based on the results of multiple logistic regression analysis, a prediction model for malalignment of lower limb was constructed, including BMI (24.31±3.58 kg/m², OR=1.12, 95%CI: 1.06~1.19, P<0.001), lower limb length [82.00 (78.00~87.00) cm, OR=0.95, 95%CI: 0.92~0.98, P=0.002], distance between both knee joints [30.00 (16.00~45.25) cm, OR=1.06, 95%CI: 1.05~1.07, P<0.001], distance between both ankle joint [23.00 (8.00~30.00) mm, OR=0.98, 95%CI: 0.96~1.00, P=0.078] and gender [man 299 (46.72%), OR=0.70, 95%CI: 0.46~1.06, P=0.089]. The area under the subject curve (AUC) value of our constructed model for predicting malalignment of lower limb was 0.808 and 0.770, respectively, in the training and validation sets.

Based on body surface big data, we primarily construct a prediction model for malalignment of lower limb for middle-aged and young people in China, which shows a good diagnostic performance on malalignment of lower limb.