Is AI 3D-printed PSI an accurate option for patients with developmental dysplasia of the hip undergoing THA?

Patients in both the PSI and control groups underwent CT-based 3D preoperative planning using AIHIP software (Version 3.0, Longwood Valley Technology, China), facilitated by two orthopedic surgeons. In contrast to other image processing software, which often involves numerous parameters and complex usability [5], AIHIP was designed to automate preoperative evaluations and simulate postoperative results. Convolutional neural networks (CNNs) were trained to automatically segment images, distinguishing between the femur and pelvis, and to identify key anatomical landmarks. The software then adjusted the pelvic orientation to a neutral stance using the bilateral anterior superior iliac spines and the pubic symphysis plane as references.

Before surgery, a detailed dysmorphic evaluation using established biomarkers was conducted. This evaluation first determined the position and size of the acetabular prosthesis, making adjustments to the acetabular cup's abduction and anteversion angles, as well as its coverage. The process then specified the position and size of the femoral prosthesis and automatically established the level of femoral resection. Simulations to forecast post-operative outcomes were performed, illustrating potential differences in leg length and modifications in femoral offset.

The PSI's design was preoperatively planned as illustrated in Fig. 1, using a three-dimensional planning system to simulate the PSI’s size, position, and orientation. The objective was to maximize the contact area while reducing surgical exposure. The design for each PSI received validation from two orthopedic surgeons, followed by the production of 3D models and surgical templates for the acetabulum and proximal femur according to the preoperative plan. The PSI comprised two components: one for guiding horizontal and directional cuts of the proximal femur, and another for directing the acetabulum's cup reaming and implant placement. The average duration from CT data acquisition to PSI completion was under 12 h, including CT processing within 1 h, THA planning and PSI simulation within 1 h, and 3D printing within 8 h.

Postoperative evaluation for each case involved assessing anterior–posterior X-rays of the pelvis, executed by a skilled technician with more than 10 years of experience in musculoskeletal imaging. This technician adjusted the tilt, rotation, and magnification to secure standard hip joint views. All postoperative imaging results were anonymized. Measurements were carried out by an experienced orthopedic resident with over 4 years of clinical experience. Before the study commenced, the resident underwent training in standardized radiographic measurements to ensure accuracy and consistency. The average of the resident's two measurements was utilized for statistical analysis. Follow-up X-ray examinations were conducted on the patients at least 4 weeks post-surgery.

The Lewinnek method [6, 7] was employed to measure the acetabular abduction angle and the acetabular anteversion angle. The abduction angle is defined by the angle between the line connecting the bilateral ischial tuberosities and the line through the center of the acetabular cup. The formula for the acetabular anteversion angle is given by: acetabular anteversion angle = ARC sine(a/b). The safe zone for implant placement, as defined by Meermans and Abdel et al. [8‐10] and supported by previous research, is an abduction angle of 40 ± 10° and an anteversion angle of 15 ± 10°. Leg length discrepancy (LLD) was identified as the variance in vertical distance from the inferior margin of the lesser trochanter to the bilateral teardrop lines. The difference in femoral offset (FO) was calculated as the discrepancy in vertical distance between the centers of the two femoral head prostheses and the proximal center axis of the femur [11].

Functional outcomes were assessed using the Harris hip score and the visual analog scale (VAS) for pain score preoperatively, at 3 days, and at 1, 3, and 12 months postoperatively.

Blood loss data were extracted from surgical records. The number of bone cuts was determined by the frequency of oscillating saw use for femur segmentation. The number of acetabular reaming iterations performed before finalizing implant selection was recorded. Hemoglobin (HGB) levels were measured preoperatively (7 days before surgery) and 3 days postoperatively. A decrease in HGB levels was identified as the difference between preoperative and postoperative measurements. The agreement rate between planned and actual implant placement was calculated, considering a variance of one implant model as acceptable.

All THA procedures were conducted by an experienced surgical team employing a uniform anterior approach technique, which was novel to the team in the context of anterior THA PSI guide usage. The standard direct anterior approach was adopted. Technical specifics included positioning patient’s supine under general anesthesia, with standard preparation and draping of the surgical site. Using the direct anterior route, the femoral neck was sequentially exposed. In the PSI group, the cutting guide was positioned between the femoral neck and anterior bony structures as determined by preoperative planning, guiding the femoral neck cut. Following femoral head removal and acetabular exposure, the superior joint capsule and labrum were excised, and the bone was delicately detached. The round ligament was severed, and the obturator foramen identified. Prosthetic design matched the femoral head and original acetabulum sizes. Reaming guide placement and assembly followed the "Lego principle," which involves completing the assembly and utilization of the guide structure within the patient's incision.. Guide pins, fixed in parallel, determined the anterior tilt angle for acetabular reaming based on their angle with the horizontal plane and the abduction angle from their orientation to the body axis. PSI application is depicted in Fig. 2 and 3. Subsequently, femoral preparation included proximal opening and medullary cavity broaching, with femoral implant anterior tilt angle aligned to the distal femoral canal line direction. The control group similarly prepared the acetabulum, maintaining a 40° abduction angle and a 15° anterior tilt angle during reaming, before implant insertion. Post-implantation, limb length was measured, and hip stability assessed, followed by layered incision closure. C-arm fluoroscopy verified implant positioning for both groups. The protheses used were Pinnacle Cup (DePuy Orthopaedics, Warsaw, IN), Corail Stem (DePuy Orthopaedics), routine procedure involved screws, excluding femoral head allografts, augments, or cement. The lead surgeon documented bone cuts and reaming instances. Standard perioperative care and patient education were uniformly provided to all patients.

Data analysis was conducted using SPSS version 22.0. Continuous data were represented as mean ± standard deviation. The independent t-test was used for comparisons between groups, while the paired t-test facilitated pre- and post-operative comparisons within groups. For categorical data, rates were provided, with the chi-square test or Fisher's exact test applied to intergroup comparisons. A significance level of α = 0.05 was established.

The average operation duration was 79.7 ± 20.8 min for the AIPSI group compared to 87.0 ± 26.6 min for the Control group (P = 0.235). The mean blood loss amounted to 405.7 ± 176.6 mL in the PSI group and 397.0 ± 188.3 mL in the control group (P = 0.855) (Table 3).

To assess postoperative functional outcomes, the Harris score and the Visual Analog Scale (VAS) pain score were evaluated preoperatively, 3 days postoperatively, and at 1, 3, and 12 months postoperatively (Table 4). There were no significant differences in preoperative Harris scores or VAS pain scores between the two groups. Three days postoperatively, the AIPSI group had an average Harris score of 61.3 ± 8.7, while the Control group scored 62.4 ± 12.2 (P = 0.668). At 12 months postoperatively, the AIPSI group's average Harris score was 95.5 ± 3.1, compared to 96.0 ± 2.0 for the Control group (P = 0.431). No significant differences were observed in VAS pain scores from 3 days to 12 months postoperatively.