Efficiency assessment of intelligent patient-specific instrumentation in total knee arthroplasty: a prospective randomized controlled trial.

BACKGROUND: In total knee arthroplasty (TKA), the practical use of patient-specific instrumentation (PSI) has been reported previously with both advantage and disadvantage. The application of artificial intelligent (AI) forces overwhelmingly development of medical industries, while the impact of AI on PSI efficiency remains unknown. Thus, this study aimed to assess the efficiency of Intelligent-PSI (i-PSI) in TKA, compared with the conventional instrumentation-TKA (CI). METHODS: 102 late-stage OA patients who met inclusive criteria were recruited in this prospective randomized controlled trial and separated into two groups (i-PSI vs. CI). In both groups, an AI preoperative planning engine was applied for surgery decision making. In CI group, conventional instrumentation was applied for bony resection, while resection of i-PSI group was completed with i-PSI. A convolutional neural network was applied to automatically process computer tomography images and thus produced i-PSI. With the help of three-dimension printing, the workflow of production was largely simplified. AI-driven preoperative planning guided resection and alignment decisions. Resection measurement, perioperative radiography and perioperative clinical outcomes were analyzed to verify efficiency of i-PSI. RESULTS: In resection outcomes, smaller deviation of lateral and medial distal femoral resection were found in i-PSI group than CI group (P = 0.032 and 0.035), while no difference was found in other resection planes. In radiography outcomes, postoperative coronal alignments of i-PSI group, including postoperative Hip-knee-ankle axis (HKA) (P = 0.025), postoperative HKA outliners (P = 0.042), Femoral coronal alignment (FCA) (P = 0.019) and Joint line convergence angle (JLCA) (P = 0.043) showed closer to neutral position than CI group. Moreover, Femoral sagittal alignment (FSA) of i-PSI group showed closer to neutral position than CI group(P = 0.005). No difference was found in other alignments. In clinical outcomes, i-PSI group seemed to cost more surgical time than CI group (P = 0.027), while others showed no differences between the two groups. CONCLUSION: Intelligent Patient-specific Instrumentation in TKA achieved simplified production flow than conventional PSI, while also showed more accurate resection, improved synthesis position and limb alignment than conventional instrumentation. Above all, this study proved that i-PSI being an applicable and promising tool in TKA.