A Study Of Preoperative Planning For Total Hip Arthroplasty Based On Propensity Score Matching

Objective: To compare the accuracy of artificial intelligence(AI)planning and two-dimensional planning in total hip arthroplasty(THA)based on propensity score matching(PSM),so as to provide a theoretical basis for the selection of preoperative planning methods for THA and a new direction for the individualized and precise treatment of patients.Method: A total of 312 patients who underwent unilateral primary total hip arthroplasty in Department of Joint Surgery,The First Affiliated Hospital of Dalian Medical Universityfrom January 2018 to August 2022 were retrospectively collected.Among them,125 patients used AI planning and 187 patients used two-dimensional planning before operation.The basic information and surgical data of the patients were collected,including age,gender,body mass index(BMI),diagnosis,affected side,American Society of Anesthesiologists(ASA)score,and comorbidities.The AI planning used the AI HIP system for 3D planning of CT data,and the X-ray film template for 2D planning.The preoperative planning and intraoperative actual acetabular and femoral prosthesis sizes of the two groups were collected.In order to reduce the influence of the difference of covariates between the two groups on the results,the propensity score matching method was used,the caliper value was set to 0.01,and the patients in the AI planning group and the two-dimensional planning group were matched according to the ratio of 1:1.After matching,105 patients in AI planning and 105 patients in twodimensional planning were obtained.The accuracy of AI planning and two-dimensional planning was analyzed by comparing the preoperative planning and the actual implant size during operation,and the influencing factors were explored.Results:1.Before the propensity score matching,there were 125 patients in the AI planning group and 187 patients in the two-dimensional planning group.There was no statistical difference between the two groups except BMI(P<0.05).2.After the propensity score matching,105 pairs of patients were successfully matched between the two groups,and all covariates reached a balance,with no statistical significance(P > 0.05);3.Complete accuracy was defined as the same model of preoperative planning and actual implantation,coincidence was defined as within one model difference,and noncoincidence was defined as two or more models difference.The results of preoperative planning of acetabular prosthesis: the complete accuracy rate of AI planning group was57.1%,and the coincidence rate was 94.3%.The complete accuracy rate of twodimensional planning group was 41.0%,and the coincidence rate was 76.2%.The results of femoral prosthesis preoperative planning: the complete accuracy rate was 53.3%and the coincidence rate was 91.4% in the AI planning group,and the complete accuracy rate was 34.3% and the coincidence rate was 69.5% in the two-dimensional planning group.Compared with two-dimensional planning,AI planning was obviously more accurate in predicting the size of acetabular and femoral prosthesis,and the implant difference between them was statistically significant(completely accurate:acetabular side P=0.027,femoral side P = 0.007;conformance: acetabular side P<0.001,femoral side P < 0.001);4.Acetabular prosthesis: The difference between the preoperative planning model and the actual implantation model of acetabular prosthesis in AI planning group was smaller than that in two-dimensional planning group(P < 0.05);The correlation between preoperative planning model and actual implantation model of acetabular prosthesis in AI planning group was higher than that in two-dimensional planning group(AI: r = 0.852,P < 0.001;two dimensions: r=0.759,P < 0.001);Femoral prosthesis : The difference between the preoperative planning model and the actual implantation model of femoral prosthesis in AI planning group was smaller than that in two-dimensional planning group(P < 0.05);The correlation between preoperative planning model and actual implantation model of femoral prosthesis in AI planning group was higher than that in two-dimensional planning group(AI: r=0.783,P< 0.001;Two dimensions: r=0.712,P<0.001).5.Univariate analysis showed that different preoperative diagnosis may lead to the influence of AI planning on the complete accuracy of acetabular prosthesis model prediction(P<0.05).Binary logistic regression analysis showed that the complete accuracy of planning acetabular prosthesis model with AI in patient with femoral neck fracture and osteonecrosis of the femoral head was higher than that in patient with developmental dysplasia of the hip(P<0.05).Conclusion: Using PSM method to balance the covariates of two groups of patients can reduce the bias caused by mixed factors.In the preoperative planning of acetabular prosthesis and femoral prosthesis for the first total hip replacement,the accuracy of AI planning method is higher than that of two-dimensional planning method,and the results of AI planning are closer to the actual implantation results.Diagnosis is the influence factor of AI planning to predict the complete accuracy of acetabular prosthesis model.Compared with patients with developmental dysplasia of the hip,AI planning has more accurate results for patients with femoral neck fracture and osteonecrosis of the femoral head.