Intraoperative Computer Vision Integrated Interactive Fluoroscopy Correlates With Successful Femoroplasty on Clinic-Based Radiographs

摘要

? 2021 Arthroscopy Association of North AmericaPurpose: To quantitatively evaluate computer vision interface (CVI)-guided femoroplasty in the arthroscopic treatment of femoroacetabular impingement syndrome and compare those results with traditional unguided resections. Methods: Consecutive patients undergoing hip arthroscopy for femoroacetabular impingement syndrome between July 2019 and October 2019 were evaluated. Cases with CVI were identified along with controls, consisting of patients from the same study period who underwent surgery without the CVI and were balanced for age, sex, laterality, and preoperative alpha angles. Alpha angles were measured on pre- and postoperative clinic radiographs, as well as intraoperatively for the CVI group. Cam resections were quantified by measuring pre- and postresection alpha angles and compared between groups. The correlation between CVI views and office-based radiographs was assessed, and the 3 CVI views that best correlated with each of the 3 standard clinic radiographs were evaluated for accuracy and performance in detection of cam deformity with alpha angle ≥48° with the clinic-based films as the reference. Results: A total of 49 patients (51 hips) (average age, 28.7; 33 female patients) in the CVI group, and 51 patients (51 hips) (average age: 29.9; 35 female patients) in the control group. There were no significant differences between groups with respect to age, sex, laterality, or preoperative alpha angle (all P > .05). Significant alpha angle reduction occurred on all intraoperative and postoperative clinic views (all P < .01). The CVI views that best correlated with the clinic radiographs were 11:45 with the anteroposterior (ρ = 0.588, P = .0025), 12:30 with the Dunn lateral (ρ = 0.632, P = .0009), and 1:45 with the false-profile (ρ = 0.575, P = .0033). Greater reliability was observed with 12:30/Dunn (accuracy = 83.33%, P < .0001; sensitivity = 77.14%; specificity = 87.76%) and 1:45/false-profile (accuracy = 82.35%, P = .0051; sensitivity = 81.82%; specificity = 82.61%) than with 11:45/anteroposterior (accuracy = 69.15%, P = .0077; sensitivity = 56.10%; specificity = 79.25%). Conclusions: CVI-guided cam resection results in successful resection of proximal femur cam lesions and represents a femoroplasty templating method that does not require preoperative computed tomography imaging or additional invasive intraoperative referencing modules. The accuracy and adequacy of this resection was validated by comparison with routine clinic radiographs. Level of Evidence: Therapeutic Level III: retrospective comparative analysis.