3D-Printed Patient-Specific Instruments-Combined AI Virtual Preoperative Planning-Assisted Medial Fixed-Bearing Unicompartmental Knee Arthroplasty

Summary

Here, we present a protocol to enhance the precision of fixed-bearing unicompartmental knee arthroplasty. It uses a digital workflow, combining 3D-printed patient-specific instruments with preoperative simulation to improve alignment accuracy, increase surgical safety, and enable personalized, reproducible outcomes.

Abstract

Unicompartmental knee arthroplasty (UKA) is widely recognized as an effective treatment for advanced knee osteoarthritis. However, its dependency on surgical experience and two-dimensional (2D) radiographs introduces significant challenges for less-experienced surgeons and for patients with femoral or tibial bone deformities. To enhance surgical precision and reduce prosthesis malposition-related complications, our team integrated three-dimensional (3D) printing technology for patient-specific instrumentation (PSI) with artificial intelligence (AI)-based preoperative simulation planning, thereby addressing the limitations of traditional approaches. This study presents a comprehensive digital workflow for fixed-bearing UKA that aims to achieve precise prosthetic alignment, improve surgical efficiency and safety, reduce operative time, and enable personalized prosthesis positioning with consistent and reproducible outcomes. The protocol consists of five critical steps: 1. Strict patient selection based on fixed-bearing UKA indications; 2. Comprehensive imaging acquisition, including CT scans of the affected knee, anteroposterior and lateral radiographs, and full-length weight-bearing X-rays in DICOM format; 3. AI-assisted preoperative planning with detailed surgical simulation reports; 4. Fabrication of customized 3D-printed cutting guides; 5. Precise osteotomy execution with intraoperative verification. By implementing this standardized digital approach, we demonstrate how integrating AI and 3D printing can optimize UKA outcomes through enhanced reproducibility, accuracy, and patient-specific customization.

Introduction

In recent years, the deep integration of digital technologies with orthopedic science has driven a paradigm shift in clinical practice-transitioning from traditional empirical and generalized methods to precision-based, personalized, and fully digital diagnostic and therapeutic strategies¹. Within this transformative landscape, 3D printing technology, as a flagship application of digital innovation, has emerged as a pivotal enabler for achieving patient-specific customization and highly precise surgical interventions in diverse orthopedic procedures^2,3. Osteoarthritis (OA), recognized as the most common degenerative joint disease worldwide, exhibits a strong correlation between its prevalence and population aging, with epidemiological studies demonstrating that accelerating global demographic aging has established OA as a major cause of chronic pain and locomotor dysfunction in middle-aged and elderly populations⁴. Regarding treatment options for medial compartment osteoarthritis (MOA), UKA presents distinct clinical advantages over total knee arthroplasty (TKA), including better preservation of native knee anatomy and physiological function, a lower incidence of postoperative complications, and significantly shorter rehabilitation periods.

The successful implementation of medial UKA is critically dependent on precise lower limb alignment reconstruction, where key challenges include variability in osteotomy execution, soft tissue balancing, and component alignment. Technically demanding aspects such as prosthesis selection and osteotomy positioning remain predominantly dependent on the surgeon's visual estimation and manual techniques, resulting in substantial operator-dependent variability, with performance discrepancies exceeding 30% among surgeons with different levels of experience⁵. Furthermore, patient-specific anatomical variations coupled with the surgeon's learning curve for new prosthesis systems frequently contribute to inconsistent clinical outcomes in UKA procedures, highlighting the need for greater standardization in surgical techniques⁶. The integration of AI planning and 3D-printed PSI directly addresses these challenges by enabling patient-specific alignment planning, automated guide design, and submillimeter osteotomy accuracy, thereby reducing dependence on surgical experience and improving reproducibility. Recent advances in 3D-printed PSI for UKA have further demonstrated these advantages⁷. Extensive research in recent years has shown that 3D digital imaging design combined with 3D printing technology enables the development of personalized surgical navigation systems for knee arthroplasty, where computed tomography-magnetic resonance imaging (CT-MRI) multimodal image fusion-based 3D-printed PSI significantly enhances the accuracy of limb alignment correction, osteotomy volume control, tibial plateau sizing, and femoral condylar prosthesis positioning^8,9,10.

To address the inherent limitations of conventional UKA techniques, our research team has developed an innovative approach by integrating 3D-printed PSI with AI-based virtual preoperative planning to assist in medial fixed-bearing UKA. By establishing a fully digitized surgical workflow for fixed-platform UKA, critical technological advancements have been achieved, including (1) preoperative biomechanical simulation using AI-driven dynamic modeling and (2) submillimeter osteotomy precision guided by 3D-printed navigation modules.

Protocol

This research received ethical approval from the First Hospital of Hebei Medical University Research Ethics Committee. Informed consent was obtained from all individual participants included in the study.

1. Patient selection and preoperative imaging

1. Select patients with anteromedial compartment osteoarthritis or medial femoral condyle osteonecrosis.
2. Confirm an intact anterior cruciate ligament (ACL), competent medial collateral ligament (MCL), and lateral compartment cartilage with Outerbridge grade ≤ II changes.
3. Exclude patients with inflammatory arthritis, varus deformity varus deformity >15°, flexion contracture >15°, or knee flexion <90°.
4. Acquire a thin-slice CT scan (≤1.0 mm) extending from 15 cm proximal to the femoral condyles to 15 cm distal to the tibial plateau.
5. Obtain complementary full-length, weight-bearing radiographs and dedicated knee views for comprehensive preoperative evaluation.

2. CT data segmentation

1. Open Mimics software v2.1 and import the CT DICOM data of the femur, patella, tibia, and fibula.
2. Use the Threshold command to adjust the standardized threshold range for bone (minimum value: 175) and filter out non-bone voxels from the image (Figure 1A).
3. Use the Region Grow command to extract bones in separate regions (Figure 1B).
4. Use the Edit Masks tool to add and fill any unselected skeletal areas (Figure 1C).
5. Convert the Masks into mesh (grid) data (Figure 1D).
6. Export the STL bone models for all relevant structures and save as backup files (Figure 1E).

3. AI-guided virtual prosthesis templating and surgical simulation

1. Open Creo Parametric software and import the STL model data of the femur, patella, tibia, and fibula.
2. Analyze the femoral anatomy by identifying the condylar line, distal femoral surface, posterior condylar surface, mechanical axis, and anatomical axis (Figure 2A-B).
3. Determine the patient-specific femoral angles (e.g., femoral varus and femoral external rotation) based on individual anatomical characteristics (Figure 2A-B).
4. Determine the patient-specific tibial angles (e.g., tibial varus, tibial retroversion, and tibial external rotation) according to the 3D anatomical model (Figure 2C-D).
5. Based on each patient's bone morphology, select the appropriate prosthesis components: femoral component (size 1-6) positioned 2-3 mm from the femoral cortical edge; tibial component (size 1-6) avoiding edge overhang or prosthesis conflict; and polyethylene insert (size 9-13 mm) (Figure 2E-F).
6. Verify implant alignment and gap balance to achieve 0° flexion and 90° extension gaps within a tolerance of ±1 mm.
7. Generate a preoperative planning report summarizing component sizing, alignment parameters, and virtual simulation results.
   NOTE: The virtual simulation for this preoperative plan was completed in an average duration of less than 2 h on a standard clinical workstation. This process typically involved 3 to 5 iterations to refine component positioning and achieve optimal soft-tissue balance, ensuring a reproducible and efficient workflow.

4. Design and production of patient-specific instruments

1. Use AI algorithms to automatically generate cutting guide designs from the finalized 3D plan. The patient-specific instrument (PSI) designs are initially generated using an automated AI-driven workflow.
   ​NOTE: The core algorithms, based on Artificia Neural Networks (ANNs) for anatomical feature recognition, analyze the 3D bone model, predefined osteotomy planes, and key anatomical landmarks¹¹. This system automatically generates the core functional elements of the guides, including the osteotomy slot positions, the bone-contacting surface ("fit-to-bone" geometry), and the overall external support structure, optimizing for parameters such as resection depth, guide stability, and contact area.
2. Perform a mandatory manual validation and editing step following the AI generation. Inspect the proposed designs meticulously in the planning software. If any issues are identified, such as insufficient contact with the cortical bone or potential soft tissue impingement, the relevant parameters (e.g., contact surface offset, pin trajectory) must be manually adjusted.
3. Design a tibial cutting guide for coronal and axial resections, incorporating three fixation pinpoints.
4. Design femoral guide I for distal resection and femoral guide II for posterior cuts and peg-hole drilling (Figure 3).
5. Fabricate all guides using medical-grade polyamide 12 via additive manufacturing using Selective Laser Sintering (SLS) technology. The printing orientation was optimized to minimize the need for support structures on critical bone-contacting surfaces, typically orienting the guide's base at a 15^\circ15 angle relative to the build platform.
6. Post-processing:
   1. Unpacking and depowdering: Remove the parts carefully from the build chamber
   2. Shot blasting: Perform shot blasting to achieve a smooth surface finish
   3. Quality inspection: Verify dimensional accuracy against the digital model
   4. Ultrasonic cleaning: Clean the guidesin deionized water to remove any residual powder
   5. Drying: Dry the guides and pack them for terminal sterilization via autoclaving at 132°C for 4 min prior to surgery.

5. Surgical procedure

1. Verify patient identity and procedure with the patient and against medical records in the preoperative holding area. Confirm signed informed consent.
2. Induce general endotracheal anesthesia. Administer prophylactic intravenous antibiotics (cefuroxime 1.5 g) 30 min prior to skin incision. Additionally, administer 1 g intravenous tranexamic acid prior to tourniquet inflation to minimize surgical blood loss.
3. Position the patient supine on the operating table. Place a nonsterile tourniquet high on the proximal thigh. Secure a lateral post at the level of the tourniquet to stabilize the femur during flexion. Pad and secure the contralateral leg. Prep and drape the operative leg in the standard sterile fashion, then, cover the foot and ankle with a sterile stockinette.
4. Perform a medial parapatellar arthrotomy while preserving the medial meniscotibial ligament.
5. Remove osteophytes that may impair guide apposition using rongeurs and osteotomes. Perform all bone resections (tibial and femoral) using an oscillating saw. The removal of osteophytes and the bone cuts are performed under direct visualization without intraoperative imaging and are guided by the preoperative plan and anatomical landmarks to ensure precise component positioning and soft tissue balance.
6. Position and fix the tibial cutting guide using 2.0 mm Kirschner wires (K-wires). Insert the K-wires under direct visualization with their trajectory aligned with the orientation of the guide pin holes.
   NOTE: Fluoroscopy is not used for guide alignment; placement is based on direct anatomical exposure and adherence to preoperative planning landmarks.
7. Perform the vertical tibial cut using a posteriorly inclined oscillating saw blade.
8. Perform the horizontal tibial cut according to the preoperatively planned posterior slope.
9. Position and fix femoral guide I, then execute the distal femoral resection along the preplanned resection plane.
10. Position and fix femoral guide II, then execute the posterior chamfer cuts and drill the fixation peg holes.
11. Trial the prosthesis components and assess the implant positioning, joint stability, and ligament balance through the range of motion.

6. Intraoperative verification and validation

1. Protect the medial collateral ligament (MCL) with a Hohmann retractor throughout the procedure.
2. Using digit calipers, verify each osteotomy surface by comparing it with the preoperative 3D-printed bone model to confirm geometric accuracy.
   NOTE: The geometry and dimensions of each osteotomy surface are verified by direct comparison with the preoperative 3D-printed bone model. Digital calipers were utilized to perform precise quantitative measurements (e.g., resection thickness, condylar dimensions) to a precision of 0.1 mm, ensuring the achieved bone cuts matched the preoperative plan.
3. Measure flexion and extension gaps under physiological loading conditions, ensuring the difference is <1 mm.
4. Assess patellar tracking dynamically through the full flexion-extension arc to confirm central tracking and prevent malalignment.
5. Document the final implant position using intraoperative radiographs and confirm absence of cortical overhang and restoration of the joint line.

Results

A 69-year-old male patient presented with bilateral knee osteoarthritis, who had previously undergone mobile-bearing unicompartmental knee arthroplasty on the right side two years ago with satisfactory postoperative recovery. Due to an intramedullary nailing procedure performed for femoral shaft fracture on the left knee two decades ago, conventional surgical instrumentation could not be properly inserted owing to the altered femoral medullary canal anatomy, thereby posing significant risks of prosthesis malalignment in standard unicompartmental knee replacement. Consequently, the surgical team has proposed an innovative approach utilizing 3D-printed patient-specific instruments combined with AI-powered virtual preoperative planning to facilitate medial fixed-bearing unicompartmental knee arthroplasty. Comprehensive preoperative imaging assessments have been completed, including weight-bearing full-length radiographs of bilateral lower limbs, standard weight-bearing anteroposterior and lateral knee radiographs, varus/valgus stress radiographs, and CT scans of the affected knee. The high-resolution CT scans with 1 mm slice thickness were performed, covering a 15 cm range from the distal femur to the proximal tibia to ensure DICOM data quality for accurate 3D reconstruction. After thorough evaluation, the patient's condition was confirmed to meet the diagnostic criteria for anteromedial osteoarthritis (Figure 4).

The AI-based preoperative planning, including 3D modeling and surgical simulation with a fixed-bearing unicompartmental knee prosthesis, was completed within 120 min (Figure 5). CT-derived anatomical data was utilized to determine optimal implant positioning and resection parameters (Table 1).

Three customized 3D-printed surgical guides (tibial cutting block, femoral distal/posterior condyle guide) were fabricated over approximately 6 h based on the preoperative plan, ensuring millimeter-level accuracy in translating digital designs to physical instruments.

The PSI-guided proximal tibial osteotomy was validated intraoperatively, confirming coronal/sagittal alignment matched preoperative planning. Sequential femoral resections were executed using two PSI guides. All bone cuts were verified intraoperatively for dimensional accuracy and alignment conformity (Figure 6).

Final weight-bearing X-rays demonstrated excellent tibiofemoral component positioning and limb alignment (mechanical axis restoration), with <1° deviation from preoperative digital plans, confirming successful execution of patient-specific instrumentation strategy (Table 1 and Figure 7). This integrated workflow highlights the synergy between AI-driven planning (120 min), rapid PSI production (6 h), and precise surgical execution-achieving sub-3mm/3°accuracy thresholds as evidenced by postoperative imaging.

Figure 1: CT data segmentation. (A) Use the Threshold command to adjust the threshold range (minimum value 175) and the threshold out of the bones in the image. (B) Use the Region Grow command to extract bones in different regions. (C) Use the Edit Masks tool to add and fill the unselected skeletal area. (D) Complete the conversion of Masks into grid data. (E) Export STL model data of bones in different parts for backup. Please click here to view a larger version of this figure.

Figure 2: AI-Guided virtual prosthesis templating and surgical simulation (A-B) Analysis of the femur anatomy by identifying the condylar line, distal femoral surface, posterior condyle surface, mechanical axis, and anatomical axis using Creo Parametric. (C-D) Anatomical angles of patients were determined according to the characteristics (tibial varus, tibial retroversion angle, tibial external rotation, etc.). (E-F) According to the patient's bone characteristics, the type of prosthesis for femoral component was selected (size:1-6) Less than 2-3mm from the edge of the femoral cortex, tibial component (size:1-6) avoiding edge of tibial component sagging or conflict of prostheses and polyethylene (size: 9-13 mm) insert. Please click here to view a larger version of this figure.

Figure 3: Patient-specific instrumentation design specifications (A-C) The patient-specific designs obtained by the AI-algorithm. 
(1, 2, 3, 9, 10) Kirschner Wire (K-wire) Fixation Hole; (4) Depth-Limited K-wire Guide Hole; (5) Vertical Osteotomy Guide Slot; (6) Pressurization Point; (7) Horizontal Osteotomy Guide Channel; (8) Distal Femoral Resection Slot; (11) Descending Osteotomy Guide Slot; (12) Guide Post Hole; (13) Posterior Femoral Condylar Osteotomy Guide Slot. Please click here to view a larger version of this figure.

Figure 4: Radiographs of the patient's legs (A) Weight-bearing full-length radiographs of bilateral lower limbs; (B-C) Standard weight-bearing anteroposterior and lateral knee radiographs; (D-E) Varus/valgus stress radiographs; (F): 3D reconstruction of knee joint models Please click here to view a larger version of this figure.

Figure 5: AI-guided 3D preoperative planning (A-B) Alignment relationship between tibial and femoral components; (C-F) Tibial component parameter planning; (G-J) Femoral component parameter planning Please click here to view a larger version of this figure.

Figure 6: Intraoperative bone cuts verification (A-B) Tibial Guide with orthogonal resection slots for vertical coronal and horizontal axial cuts; (C) Femoral Guide I providing distal resection control; (D) Femoral Guide II enabling simultaneous preparation of posterior chamfer cuts and dual-peg fixation bores. (E-H) Real-time quality assurance comparison of resected bone fragments to mirror-image preoperative models. Please click here to view a larger version of this figure.

Figure 7: Postoperative radiographic assessment  (A-B) Alignment Relationship Between Tibial and Femoral Components. (C-D) Postoperative Radiographic Assessment (All postoperative measurements were largely consistent with the preoperative plan).Please click here to view a larger version of this figure.

Table1 AI-Guided 3D Preoperative Planning and Postoperative Radiographic Assessment
Parameter	Target Value	Measurement value	Parameter	Target Value	Measurement value
Posterior tibial slope	6.0°	6.0°	Tibial prosthesis varus angle	3°	3.3°
Tibial component posterior slope angle	5°	5.5°	Femoral prosthesis external rotation angle	3°	4.5°
Tibial horizontal osteotomy thickness	6.0 mm	6.2 mm	Distal femoral osteotomy thickness	6.5 mm	6.5mm
Tibial prosthesis model/size	4#	4#	Femoral prosthesis model/size	5#	5#
Note: precise surgical execution-achieving sub-3mm/3°accuracy thresholds as evidenced by postoperative imaging.

Table 1: AI-guided 3D preoperative planning and postoperative radiographic assessment.

Discussion

The successful implementation of this digital UKA protocol hinges on several critical steps. First, high-quality, thin-slice (≤1.0 mm) CT data acquisition is paramount, as the fidelity of the subsequent 3D bone models and the precision of the PSI guides are directly dependent on it^8,12. Second, the AI-guided virtual planning stage is crucial for determining patient-specific alignment. This involves accurately identifying anatomical axes and landmarks (e.g., femoral condylar line, tibial torsion) to plan component positioning that respects individual anatomy and achieves balanced flexion/extension gaps within a ±1 mm tolerance^13,14. Finally, the intraoperative execution requires meticulous removal of osteophytes that could impede perfect bone-PSI contact, followed by stable fixation of each guide with K-wires before performing the definitive, guided osteotomies.

A key modification in our workflow is the use of AI algorithms to reconstruct soft tissue morphology and predict cartilage thickness from CT data alone, addressing the common limitation of CT-based PSI which cannot directly account for cartilage¹². If a guide does not sit perfectly intraoperatively, which is often due to residual osteophytes or soft tissue interposition, the protocol mandates immediate troubleshooting by re-checking and clearing the bone surfaces. Furthermore, we employ intraoperative verification by comparing resected bone fragments to 3D-printed, mirror-image preoperative models (Figure 6), providing immediate feedback on the accuracy of each cut.

Despite its advantages, this technique has inherent limitations. The primary limitation is its reliance on CT imaging, which involves radiation exposure and increased initial costs compared to conventional methods. While our AI software estimates soft tissue, it does not fully replace the data provided by MRI for precise cartilage mapping, which could theoretically lead to minor inaccuracies in gap balancing predictions¹⁵. Additionally, the process requires access to specialized software, 3D printing facilities, and adds a preoperative planning and guide fabrication time of approximately 6-8 h, which may not be feasible in all healthcare settings.

While cost-benefit analysis of 3D-PSI remains debated, our protocol's reduced operative time may offset initial expenses. Justin et al. found significant inconsistencies between preoperative digital planning and postoperative component alignment outcomes specifically in the sagittal and axial plane positioning angles of femoral components as well as the coronal plane angulation of tibial components¹⁶, which necessitates further validation through large-scale prospective cohort studies involving a minimum sample size of 200 cases utilizing a multicenter design with extended follow-up periods of at least 5 years.

Previous studies have demonstrated that unicompartmental knee arthroplasty offers significant advantages over total knee arthroplasty in terms of symptom relief, functional recovery, and postoperative complication control¹⁷. However, long-term follow-up data indicate a substantially lower implant survival rate for UKA compared to TKA, which may be attributed to inadequate surgical exposure due to the minimally invasive approach used in UKA, resulting in increased intraoperative errors in extramedullary alignment determination and reduced accuracy in femoral condylar intramedullary positioning¹⁸. Current evidence confirms that even minor malalignment or residual varus deformity significantly elevates the risk of future revision surgery in UKA patients, while postoperative valgus misalignment induces excessive stress concentration in the lateral compartment, accelerating cartilage degeneration in this region¹⁹. Consequently, meticulous preoperative three-dimensional planning coupled with submillimeter-level precision in intraoperative implant positioning constitutes the cornerstone of successful UKA procedures²⁰. Conventional analysis of knee radiographs, CT, and MRI scans often fails to accurately reconstruct the 3D structure of the lower limb or quantitatively assess cartilage wear in the medial compartment. Moreover, measurement errors appear unavoidable during surgery since tibial plateau alignment in T-UKA relies on subjective extramedullary guiding rods. Additionally, any deviations in intramedullary positioning of the distal femoral condyle, as well as misalignment of the guiding rod, may lead to adverse outcomes such as meniscal impingement, suboptimal implant tracking, prolonged surgical duration, increased blood loss, and heightened risk of intraoperative fat embolism²¹. Baldini et al. also noted a high prevalence (10%-20%) of T-UKA cases where the femoral component alignment deviated significantly from normative values²². Fei Gu et al. successfully integrated CT and MRI datasets to achieve precision minimally invasive prosthesis design while simultaneously reducing risks of spacer impingement and rotational malalignment¹⁵.

Our AI algorithm reduced preoperative planning time to 120 min - significantly faster than conventional manual methods (typically 3-6 h)²³. By automating tibial slope calculation (accounting for native posterior inclination) and femoral component positioning (with dynamic gap simulation), the system addresses the "learning curve effect" observed in freehand UKA techniques¹³. This is particularly valuable for surgeons transitioning from TKA to UKA²⁴. Our results align with prior studies reporting improved accuracy using PSI in UKA, while introducing novel AI-driven optimization of biomechanical parameters critical advancement given the documented correlation between malalignment and early implant failure^14,25.

The implementation of this digital workflow requires consideration of economic implications, learning curve, and accessibility. While initial costs are higher due to specialized software, 3D printing facilities, and AI integration, the reduction in operative time and potential for fewer revisions may offset these expenses in the long term. The learning curve for surgeons and technicians is moderate, with approximately 5-10 cases needed to achieve proficiency. However, the requirement for specialized infrastructure may limit accessibility in resource-limited settings. Future efforts should focus on cost reduction strategies and streamlined workflows to enhance broader clinical adoption.

The future applications of this integrated digital workflow are extensive. The underlying AI planning algorithm can be further refined with larger datasets to improve its predictive accuracy for soft-tissue balancing and clinical outcomes. This protocol can be adapted for other complex joint arthroplasties, such as patellofemoral arthroplasty or total ankle replacement, where patient-specific anatomy is paramount. Furthermore, this workflow serves as a foundational step towards the future implementation of augmented reality (AR) navigation, where the virtual plan could be projected directly onto the surgical field without the need for physical guides, potentially combining the accuracy of PSI with greater flexibility and reduced cost.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Creo Parametric 8.0.0.0	PTC	https://www.ptc.com/ en/products/creo	Surgical simulation guide output
3D-Printed Patient-Specific Instruments	Sikon Medical Co., Ltd. China	D250610	Anatomic Conformity & Precision Positioning; Optimized Limb Alignment;Surgical Efficiency; Minimally Invasive Approach?Enhanced Reproducibility
Femoral Component	Beijing Leadcom Biomedical Co., Ltd.	A5101-05lM	Resurfaces the arthritic femoral condyle (medial/lateral).
Mimics Research 21.0	Materialise	https://www.materialise. com/en/software/mimics-research-suite	image processing
Polyethylene Insert	Beijing Leadcom Biomedical Co., Ltd.	A5301-0410	Acts as an articulating surface between femoral and tibial components
Tibial Componen	Beijing Leadcom Biomedical Co., Ltd.	A5201-04LM	Replaces the worn tibial plateau and provides a base for  the polyethylene insert.