Abstract
To explore the application value of an artificial intelligence (AI)-based cartilage model evaluation system in the decision-making of knee joint replacement procedures (total knee arthroplasty [TKA] vs. unicompartmental knee arthroplasty [UKA]), and to compare the dynamic differences in postoperative functional recovery between the two procedures. Eighty patients with advanced knee osteoarthritis evaluated by the AI cartilage model were retrospectively included, with 40 patients in each of the TKA group and the UKA group. The AI model performed thin-slice MRI scans of the knee joints, selected the fat-suppressed three-dimensional isotropic intermediate-weight VISTA sequence to improve image identification, and imported the data into an MRI-based intelligent reconstruction system to achieve intelligent segmentation and reconstruction of MRI. It reconstructed 15 structures including the femur, tibia, patella, their cartilages, and ligaments. The surgical procedures were decided based on the cartilage damage. The Knee Society Score (KSS, including knee score and function score), Visual Analog Pain Score (VAS), and WOMAC Osteoarthritis Index (pain, stiffness, function dimensions) were recorded before surgery and 1, 3, 6, and 12 months after surgery in both groups. Simultaneously record the preoperative and postoperative hip–knee–ankle (HKA) angles and their differences for both patient groups, the location of preoperative pain, and the preoperative and 12-month postoperative knee range of motion (ROM) and their differences. There were no statistically significant differences in gender and average BMI between the unicompartmental knee replacement group and the total knee replacement group (P > 0.05); however, there was a statistically significant difference in average age between the two groups (P < 0.05). Within each group, there were statistically significant differences in all scores (KSS knee score, KSS function score, VAS score, WOMAC score) at 1 month, 3 months, 6 months, and 12 months after surgery compared with those before surgery (P < 0.05). In the inter-group comparison (comparison between unicompartmental knee replacement patients and total knee replacement patients), for the KSS knee score, P = 0.27 before surgery, P < 0.05 at 1 month after surgery; P = 0.46 at 3 months after surgery, P = 0.33 at 6 months after surgery, and P = 0.08 at 12 months after surgery. For the KSS function score, P < 0.05 before surgery, P < 0.05 at 1 month after surgery, P < 0.05 at 3 months after surgery, P = 0.73 at 6 months after surgery, and P = 0.76 at 12 months after surgery. For the VAS score, P = 0.51 before surgery, P < 0.05 at 1 month and 3 months after surgery, P = 0.82 at 6 months after surgery, and P = 0.80 at 12 months after surgery. For the WOMAC score, P < 0.05 at all time points before and after surgery. In the unicompartmental knee arthroplasty (UKA) group, the preoperative hip–knee–ankle (HKA) angle was 175.8° ± 1.4°, the postoperative HKA angle was 178.6° ± 1.0°, and the change was 2.9° ± 0.6°. In the total knee arthroplasty (TKA) group, the preoperative HKA angle was 175.1° ± 2.5°, the postoperative HKA angle was 179.2° ± 0.9°, and the change was 8.1° ± 1.8°. Comparison between the two groups showed no significant difference in the preoperative HKA angle (p = 0.12), whereas the postoperative HKA angle and the change from baseline were significantly different (both p < 0.05). In the UKA group, pain was located medially in 35 cases (87.5%) and laterally in 5 cases (12.5%). In the TKA group, pain was located medially in 25 cases (62.5%), laterally in 10 cases (25.0%), and in the patellofemoral joint in 5 cases (12.5%). For range of motion (ROM), the UKA group had a preoperative ROM of 105.8° ± 6.4°, a 12-month postoperative ROM of 130.8° ± 5.2°, and a change of 25.0° ± 1.5°. The TKA group had a preoperative ROM of 89.0° ± 6.8°, a 12-month postoperative ROM of 121.0° ± 4.5°, and a change of 32.0° ± 2.5°. Intergroup comparisons of preoperative, postoperative, and change values all showed statistically significant differences (all p < 0.05). The AI cartilage model can accurately distinguish the indications for TKA and UKA by quantifying the morphology of cartilage damage. UKA is more suitable for local cartilage damage in the medial compartment, and it has a significant advantage in early functional recovery; while TKA has better long-term stability in patients with extensive cartilage degeneration. This study provides an evidence-based basis for AI-assisted personalized selection of knee joint replacement procedures.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-026-36186-x.
Keywords: Orthopedic decision-making using artificial intelligence, Cartilage biomechanical modeling, Total knee arthroplasty, Unicompartmental knee arthroplasty
Subject terms: Health care, Medical research
Introduction
Advanced knee osteoarthritis is a common degenerative joint disease that seriously affects the quality of life of patients, usually causing severe pain and limitations in mobility1. Total knee arthroplasty (TKA) and unicompartmental knee arthroplasty (UKA) are common surgical interventions. Some studies have shown that MRI imaging evaluation can more accurately reflect the situation of cartilage damage, thus providing support for surgical decision-making2. In recent years, the development of artificial intelligence (AI) technology has provided new opportunities for medical image analysis. Using an AI cartilage model to evaluate knee joint images, especially to reconstruct various structures of the knee joint through MRI technology, has shown its potential in improving evaluation accuracy and reducing the influence of subjective factors3. Existing studies have shown that AI technology can help doctors make more scientific decisions before surgery, thereby improving the surgical outcomes and quality of life of patients4.
This study adopted a retrospective analysis method, using the AI model to improve the evaluation accuracy, exploring the application prospects of the AI imaging model in knee joint surgery decision-making. By studying 80 patients with advanced knee osteoarthritis and comparing the postoperative score data of the TKA group and the UKA group, as well as the comparison of postoperative functional recovery and pain control, it provides evidence-based surgical selection suggestions for clinical practice, and further optimizes the treatment plan for patients.
Materials and methods
This retrospective study was approved by the institutional board of Beijing Tsinghua Changgung Hospital(No. 24223-0-03). All participants provided written informed consent prior to enrollment. all methods were carried out in accordance with relevant guidelines and regulations.
General information
Forty patients who underwent unilateral TKA and 40 patients who underwent unilateral UKA at Beijing Tsinghua Changgung Hospital from May 2022 to February 2024 were retrospectively selected. In all UKA patients, preoperative magnetic resonance imaging (MRI) and physical examination confirmed the integrity of both the anterior and posterior cruciate ligaments. In the TKA group, there were 7 males and 33 females; in the UKA group, there were 8 males and 32 females. The average age of the TKA group was 70.03 ± 7.06 years old, and the average BMI was 27.24 ± 3.15 kg/m²; the average age of the UKA group was 66.03 ± 5.03 years old, and the average BMI was 27.48 ± 3.49 kg/m². There were no statistically significant differences in gender and average BMI between the total knee replacement group and the unicompartmental knee replacement group (P > 0.05); however, there was a statistically significant difference in average age between the two groups (P < 0.05). The JUST SKII fixed platform prosthesis was used in all patients in the TKA group, and the Zimmer Biomet Oxford UKA mobile platform prosthesis was used in all patients in the UKA group.
Inclusion and exclusion criteria
Inclusion criteria: ① All surgeries were performed by the same chief surgeon; ② All patients were diagnosed with severe knee osteoarthritis; ③ All patients were scanned using the same MRI machine. Exclusion criteria: ① Rheumatoid arthritis and other types of arthritis; ② Patients with active infection foci around the knee joint or systemic active inflammation or obvious active phase of inflammatory arthritis in the recent period and currently; ③ Patients with unicompartmental knee osteoarthritis whose symptoms could not be improved after high tibial osteotomy.
Imaging data collection
Although this study was a retrospective analysis, since 2022 our institution has incorporated MRI-based artificial intelligence (AI) cartilage model assessment into the routine preoperative evaluation for knee arthroplasty. This approach enables more precise determination of the extent of cartilage damage and the selection of surgical procedure. Consequently, all included patients underwent AI assessment preoperatively, and the results served as an important reference for surgical decision-making.All patients underwent preoperative anteroposterior and lateral X-ray films of the knee joint and thin-slice MR scans of the surgical knee joint (3.0T, Philips Ingenia CX, Netherlands). The fat-suppressed three-dimensional isotropic intermediate-weight VISTA sequence was selected to improve image identification, and the data were imported into an MRI-based intelligent reconstruction system5. Since the AI-reconstructed cartilage model can only determine whether there is cartilage defect and cannot conduct detailed grading from grade 1 to grade 3 according to the ICRS classification6, during the evaluation, if there is no cartilage coverage in a certain area of the model, this area is considered as ICRS grade 4. For the measured areas in the model, as long as there is one grade 4 cartilage injury, the entire area is judged as positive. All patients underwent full-length lower limb radiography preoperatively and at one week postoperatively to measure the hip–knee–ankle (HKA) angle, and the preoperative and postoperative values as well as their differences were recorded. The location of preoperative pain (medial, lateral, or patellofemoral joint) was documented for all patients. In addition, preoperative and 1-year postoperative knee range of motion (ROM) and their differences were recorded for all patients.
Observation indicators
When the AI-reconstructed cartilage model showed only cartilage damage in the medial compartment of the knee joint, UKA was used for treatment (Fig. 1). When there was grade 4 cartilage injury in the lateral compartment and/or patellofemoral joint, TKA was used for treatment (Fig. 2). The Knee Society Score (KSS, including knee score and function score), Visual Analog Pain Score (VAS), and WOMAC Osteoarthritis Index (pain, stiffness, function dimensions) were recorded before surgery and 1, 3, 6, and 12 months after surgery in both groups.
Fig. 1.
Three-dimensional cartilage reconstruction for a patient selected for unicompartmental knee arthroplasty (UKA), generated using ITK-SNAP software (version 3.8.0; http://www.itksnap.org).
Fig. 2.
Three-dimensional cartilage reconstruction for a patient selected for total knee arthroplasty (TKA), generated using ITK-SNAP software (version 3.8.0; http://www.itksnap.org).
Statistical analysis
All statistical analyses were conducted using SPSS27 software (IBM, Armonk, NY, USA). The assessment of normality for continuous variables was conducted using the Kolmogorov–Smirnov test.
Results
Detailed patient demographics are presented in Supplementary Table S1.
Pre- and postoperative clinical outcome scores (KSS, WOMAC, and VAS) are summarized in Supplementary Table S2.
Pain location, pre- and postoperative HKA angles, and knee range of motion (ROM) data are shown in Supplementary Table S3.
Detailed patient demographics are presented in Table 1.There were no statistically significant differences in gender and average BMI between the unicompartmental knee replacement group and the total knee replacement group (P > 0.05); however, there was a statistically significant difference in average age between the two groups (P < 0.05).
Table 1.
Demographic and clinical information of patients.
| TKA | UKA | p-value | |
|---|---|---|---|
| Sex | |||
| Male | 7 | 8 | >0.05 |
| Female | 33 | 32 | |
| Body mass index(kg/m2) | 27.24 ± 3.15 | 27.48 ± 3.49 | >0.05 |
| Age | 70.03 ± 7.06 | 66.03 ± 5.03 | <0.05 |
In the total knee replacement group, the preoperative KSS knee score was 45.0 ± 4.7, the score at 1 month after surgery was 63.7 ± 5.9, the score at 3 months after surgery was 80.9 ± 7.9, the score at 6 months after surgery was 85.9 ± 6.5, and the score at 12 months after surgery was 89.3 ± 4.8. The preoperative KSS function score was 37.7 ± 6.3, the score at 1 month after surgery was 48.4 ± 8.4, the score at 3 months after surgery was 69.1 ± 7.7, the score at 6 months after surgery was 79.2 ± 6.6, and the score at 12 months after surgery was 82.3 ± 6.7. The preoperative VAS score was 7.7 ± 1.5, the score at 1 month after surgery was 4.9 ± 2.1, the score at 3 months after surgery was 2.7 ± 1.4, the score at 6 months after surgery was 1.8 ± 1.2, and the score at 12 months after surgery was 1.3 ± 0.8. The preoperative WOMAC score was 188.0 ± 29.8, the score at 1 month after surgery was 69.5 ± 21.2, the score at 3 months after surgery was 55.6 ± 16.6, the score at 6 months after surgery was 37.6 ± 15.4, and the score at 12 months after surgery was 18.2 ± 7.5.
In the unicompartmental knee replacement group, the preoperative KSS knee score was 46.9 ± 10.4, the score at 1 month after surgery was 76.3 ± 11.4, the score at 3 months after surgery was 82.3 ± 10.0, the score at 6 months after surgery was 84.1 ± 9.5, and the score at 12 months after surgery was 86.4 ± 8.7. The preoperative KSS function score was 43.5 ± 8.9, the score at 1 month after surgery was 64.7 ± 9.8, the score at 3 months after surgery was 76.3 ± 8.7, the score at 6 months after surgery was 78.6 ± 8.3, and the score at 12 months after surgery was 81.7 ± 7.5. The preoperative VAS score was 7.5 ± 1.4, the score at 1 month after surgery was 2.6 ± 1.4, the score at 3 months after surgery was 1.6 ± 0.9, the score at 6 months after surgery was 1.7 ± 1.2, and the score at 12 months after surgery was 1.3 ± 0.9. The preoperative WOMAC score was 135.3 ± 36.0, the score at 1 month after surgery was 94.8 ± 29.0, the score at 3 months after surgery was 74.4 ± 25.4, the score at 6 months after surgery was 56.8 ± 22.3, and the score at 12 months after surgery was 40.0 ± 20.3.
Within each group, there were statistically significant differences in all scores (KSS knee score, KSS function score, VAS score, WOMAC score) at 1 month, 3 months, 6 months, and 12 months after surgery compared with those before surgery (P < 0.05).
In the inter-group comparison (comparison between unicompartmental knee replacement patients and total knee replacement patients), for the KSS knee score, P = 0.27 before surgery, P < 0.05 at 1 month after surgery; P = 0.46 at 3 months after surgery, P = 0.33 at 6 months after surgery, and P = 0.08 at 12 months after surgery. For the KSS function score, P < 0.05 before surgery, P < 0.05 at 1 month after surgery, P < 0.05 at 3 months after surgery, P = 0.73 at 6 months after surgery, and P = 0.76 at 12 months after surgery. For the VAS score, P = 0.51 before surgery, P < 0.05 at 1 month and 3 months after surgery, P = 0.82 at 6 months after surgery, and P = 0.80 at 12 months after surgery. For the WOMAC score, P < 0.05 at all time points before and after surgery.
Pre-and postoperative HKA angles and the corresponding changes are summarized in Table 2. In the UKA group, the preoperative HKA angle was 175.8° ± 1.4°, the postoperative HKA angle was 178.6° ± 1.0°, and the change was 2.9° ± 0.6°. In the TKA group, the preoperative HKA angle was 175.1° ± 2.5°, the postoperative HKA angle was 179.2° ± 0.9°, and the change was 8.1° ± 1.8°. Comparison between the two groups showed no significant difference in the preoperative HKA angle (p = 0.12), whereas the postoperative HKA angle and the change from baseline were significantly different (both p < 0.05).
Table 2.
Comparison of HKA angles between the two Groups.
| Parameter | UKA | TKA | p-value |
|---|---|---|---|
| Preoperative HKA angle (°) | 175.8 ± 1.4 | 175.1 ± 2.5 | 0.12 |
| Postoperative HKA angle (°) | 178.6 ± 1.0 | 179.2 ± 0.9 | < 0.05 |
| Change (°) | 2.9 ± 0.6 | 8.1 ± 1.8 | < 0.05 |
Regarding pain location, in the UKA group, 35 patients (87.5%) reported medial pain and 5 patients (12.5%) reported lateral pain. In the TKA group, 25 patients (62.5%) reported medial pain, 10 patients (25.0%) reported lateral pain, and 5 patients (12.5%) reported patellofemoral joint pain.
Comparisons of preoperative and postoperative knee range of motion (ROM) between the two groups are shown in Table 3. For ROM, the UKA group had a preoperative value of 105.8° ± 6.4°, a 12-month postoperative value of 130.8° ± 5.2°, and a change of 25.0° ± 1.5°. The TKA group had a preoperative value of 89.0° ± 6.8°, a 12-month postoperative value of 121.0° ± 4.5°, and a change of 32.0° ± 2.5°. Intergroup comparisons for preoperative, postoperative, and change values all showed statistically significant differences (all p < 0.05).
Table 3.
Comparison of knee range of motion (ROM) between the two Groups.
| Parameter | UKA | TKA | p-value |
|---|---|---|---|
| Preoperative ROM (°) | 105.8 ± 6.4 | 89.0 ± 6.8 | < 0.05 |
| ROM at 12 months postoperatively (°) | 130.8 ± 5.2 | 121.0 ± 4.5 | < 0.05 |
| Change (°) | 25.0 ± 1.5 | 32.0 ± 2.5 | < 0.05 |
Discussion
The background of this study is advanced knee osteoarthritis, which is a common degenerative joint disease that restricts physical mobility, brings huge physical and mental burdens to patients and their families, and also increases the social medical cost3. Although current treatment methods include conservative treatment and surgical treatment, in surgical treatment, problems such as uneven postoperative recovery and inaccurate functional assessment still exist in total knee arthroplasty (TKA) and unicompartmental knee arthroplasty (UKA)4. Therefore, it is very necessary to explore more accurate evaluation methods and surgical selections. This study used an AI cartilage model to evaluate thin-slice MRI images of the knee joint, hoping to provide more scientific surgical decision-making support and optimize the surgical outcomes and quality of life of patients.
This study compared the postoperative effects of 80 patients with advanced knee osteoarthritis who received TKA and UKA through retrospective analysis. Intra-group comparison: In both groups, all scores at 1 month, 3 months, 6 months, and 12 months after surgery were significantly improved compared with those before surgery (P < 0.05). This indicates that there are extremely significant statistical differences between the scores at different time points after surgery and those before surgery for both types of surgeries. We can speculate that both types of surgeries have very obvious effects on improving the relevant symptoms of patients (such as knee joint function, pain degree, etc., specifically depending on the meaning represented by the scores).
Inter-group comparison: Previous research results have shown that there are significant differences between the two surgical methods in improving knee joint function and pain control7. This study shows that for the KSS knee score, the UKA group was significantly better than the TKA group at 1 month after surgery (76.3 ± 11.4 vs. 63.7 ± 5.9, P < 0.05), which may be related to the smaller trauma of UKA and the preservation of more normal structures. For the VAS pain score, the pain relief in the UKA group was more obvious at 1 and 3 months after surgery (2.6 ± 1.4 vs. 4.9 ± 2.1, 1.6 ± 0.9 vs. 2.7 ± 1.4), suggesting that minimally invasive procedures are better at early pain control. At 6 and 12 months after surgery, there were no statistically significant differences in the KSS knee score, function score, and VAS between the two groups, indicating that the final curative effects of the two surgical methods were similar. The UKA group had a lower baseline WOMAC score compared with the TKA group (135.3 ± 36.0 vs. 188.0 ± 29.8), and demonstrated significantly better outcomes at all postoperative time points (p < 0.05). These findings indicate marked improvements in pain, stiffness, and functional limitation, which may be attributed to the more severe preoperative condition and limited recovery potential in the TKA group.
There was no significant difference in preoperative lower limb alignment between the two groups (p = 0.127), indicating a comparable degree of preoperative deformity. Postoperatively, the TKA group achieved significantly better alignment than the UKA group (p < 0.05), although both groups reached clinically satisfactory alignment. The degree of correction was significantly greater in the TKA group than in the UKA group (p < 0.05), consistent with the characteristic requirement of greater correction in cases with more severe deformity undergoing total knee arthroplasty.
Unicompartmental knee arthroplasty preserves normal joint structures, resulting in limited correction of joint deformity and minimal change in alignment (mean 2.9°). In contrast, total knee arthroplasty is suitable for patients with severe deformity, allowing comprehensive correction of overall joint malalignment with a larger change in alignment (mean 8.1°).
In the UKA group, the majority of patients presented with medial knee pain (87.5%), which is consistent with the indication of UKA for medial compartment osteoarthritis. In the TKA group, pain location was more widely distributed but predominantly medial (62.5%), reflecting the broader indications of TKA for multicomartment disease.
The application of AI imaging models can enhance the precision of evaluations and reduce the influence of subjective factors, providing more reliable decision-making support for clinical practice8. These findings highlight the potential application value of AI technology in knee joint surgeries and offer important references for subsequent research9. The innovation of this study lies in the first systematic evaluation of the effectiveness of surgical methods (total knee arthroplasty (TKA) and unicompartmental knee arthroplasty (UKA)) for patients with advanced knee osteoarthritis using an artificial intelligence (AI) cartilage model. Compared with previous studies10–12, our research provides an AI-based image analysis method in clinical applications. This method can effectively improve the precision of preoperative evaluations, thereby offering a more scientific basis for surgical decision-making. Previous studies mostly employed traditional imaging evaluation methods and lacked accurate comparisons of the functional recovery and pain control of different surgical methods. However, through a retrospective analysis of 80 patients, this study clarifies the significant differences in postoperative outcomes between TKA and UKA, providing important references for clinical practice3,4,13.
The research results have important implications for clinical practice and policy-making, especially in terms of providing a new perspective when selecting surgical plans. Our findings indicate that there are significant differences in the improvement of knee joint function and pain control among different surgical methods. This provides a scientific basis for surgeons when choosing surgical methods, helping them make better decisions according to the specific conditions of patients. For example, UKA may be a more suitable choice for younger patients and those with higher functional requirements, while TKA may have more advantages for elderly patients or those with specific functional needs14,15. In addition, this also reflects that when formulating medical policies, more attention should be paid to evidence-based personalized medicine to improve the overall quality of life of patients and the postoperative recovery effect16–18.
Despite the important clinical significance of the study results, there are still some limitations. As there were differences in age and baseline function between the two groups preoperatively, future studies will employ propensity score matching or multivariate regression analysis to adjust for potential confounding factors.Firstly, the sample size of this study is relatively small, including only 80 patients, which may affect the wide applicability of the results. Secondly, the retrospective analysis design adopted in this study may have selection bias and fails to comprehensively consider all potential factors affecting postoperative outcomes. In addition, although the two groups were matched in terms of gender and BMI, there was a significant difference in age (P < 0.05), which may affect the results. For example, if the UKA group is younger, their bone quality and recovery ability may be better than those of the TKA group, leading to differences in early scores. In the future, multivariate analysis or propensity score matching should be used to control age confounding factors, the sample size should be expanded, and a prospective randomized controlled trial design should be adopted to verify the findings of this study and explore the impact of different patient characteristics on surgical outcomes, thereby optimizing the clinical decision-making process8,19. As the AI model can only identify the presence or absence of cartilage defects and cannot perform refined grading of ICRS grades 1–3, some moderate cartilage lesions may be classified as “normal,” potentially reducing the sensitivity of surgical procedure selection. Future developments should focus on AI algorithms capable of accurate grading to enhance the precision of clinical decision-making.
In conclusion, the use of an AI cartilage model to evaluate the selection of surgical methods for patients with advanced knee osteoarthritis is of significant importance, especially in improving knee joint function and pain control in the early postoperative period. The AI-based three-dimensional cartilage model derived from MRI not only accurately identifies areas of severe cartilage damage preoperatively, thereby aiding in the differentiation of indications for TKA and UKA, but also helps to avoid inappropriate procedure selection and reduce unnecessary total knee arthroplasties. Without this tool, some patients with localized lesions might undergo the more invasive TKA, which could result in slower early functional recovery.However, considering the limitations of the study, future research should continue to explore the impact of different patient characteristics on surgical outcomes and further optimize postoperative management strategies to enhance the scientific nature of clinical decision-making and the quality of life of patients.
Supplementary Information
Below is the link to the electronic supplementary material.
Author contributions
Conceptualization: Wei Zhang, Hong Gao, Xu Cai. Data curation: Hong Gao, Baojing Zhao, QI Ma. Formal analysis: Wei Zhang, Hong Gao, Investigation: Hong Gao. Supervision: Xu Cai. Validation: Hong Gao. Visualization: Xu Cai. Writing – original draft: Wei Zhang.
Funding
National Key R&D Program of China (2023YFB4606704).
Data availability
The data that support the findings of this study are available from the first author, Wei Zhang, upon reasonable request. Requests for access to the data should be submitted through the journal’s submission system and addressed to Wei Zhang.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Baojing Zhao, Email: zbjA01794@btch.edu.cn.
Xu Cai, Email: caixu_changgeng@163.com.
References
- 1.Abaydi, A., Radi, J., Tbatou, A., Lahrach, K. & Boutayb, F. Tibio-Talo-Calcaneal arthrodesis: evaluation of the effectiveness of a specific surgical technique in 17 cases. Cureus16 (6), e62014 (2024). [DOI] [PMC free article] [PubMed]
- 2.Alyami, J. Identification of severe grading in knee osteoarthritis from MRI using ensemble deep learning. Curr. Med. Imaging. 20, e15734056289963 (2024). [DOI] [PubMed] [Google Scholar]
- 3.Han, S. H., Lee, M. S. & Kong, S. H. Clinical benefit of Robotic-Assisted total knee arthroplasty over conventional total knee arthroplasty when using Mobile-Bearing implants. Med. (Kaunas Lithuania). 60 (7), 1103 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Wang, W. et al. Subchondral bone expansion in advanced knee osteoarthritis: relation with radiographic severity and role in surgical decision-making. Osteoarthr. Cartil. open.6 (2), 100461 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Shen, L. et al. A feasibility study of knee joint semantic segmentation on 3D MR images[J]. CT Theory Applications. 31 (6), 1–12 (2022). [Google Scholar]
- 6.England, J. R. & Cheng, P. M. Artificial intelligence for medical image analysis: a guide for authors and reviewers. AJR Am. J. Roentgenol.212 (3), 513–519 (2019). [DOI] [PubMed] [Google Scholar]
- 7.Zhang, Z. et al. Can robotic Arm-assisted total knee arthroplasty remain Cost-effective in Volume-based procurement system in china? A Markov Model-based study. Orthop. Surg.16 (6), 1434–1444 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Alturki, A. A. et al. Regional anesthesia for high-risk patients undergoing total knee arthroplasty: A case report. Cureus16 (2), e55269 (2024). [DOI] [PMC free article] [PubMed]
- 9.Dagher, T., Dwyer, E. P., Baker, H. P., Kalidoss, S. & Strelzow, J. A. Dr. AI will see you now: how do ChatGPT-4 treatment recommendations align with orthopaedic clinical practice guidelines? Clin. Orthop. Relat. Res.482 (12), 2098–2106 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Twinprai, P. et al. AI classification of knee prostheses from plain radiographs and real-world applications. Eur. J. Orthop. Surg. Traumatol.35.0 (1.0), 107 (2025). Published 2025 Mar 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Yu, Y., Cho, Y. J., Park, S., Kim, Y. H. & Goh, T. S. Development of an artificial intelligence model for predicting implant size in total knee arthroplasty using simple X-ray images. J. Orthop. Surg. Res.19 (1), 516 (2024). Published 2024 Aug 27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Velasquez Garcia, A. et al. Artificial intelligence-based three-dimensional templating for total joint arthroplasty planning: a scoping review. Int. Orthop.48 (4.0), 997–1010 (2024). [DOI] [PubMed] [Google Scholar]
- 13.Güzel, İ., Gül, D., Akpancar, S. & Lyftogt, J. Effectiveness of perineural injections combined with standard postoperative total knee arthroplasty protocols in the management of chronic postsurgical pain after total knee arthroplasty. Med. Sci. Monitor: Int. Med. J. Experimental Clin. Res.27, e928759 (2021). [DOI] [PMC free article] [PubMed]
- 14.Shon, O. J., Kim, G. B., Jo, S. H., Medicina & Kaunas Sex Differences in Sarcopenia in Patients Undergoing Total Knee Arthroplasty for Advanced Knee Osteoarthritis. (Lithuania), 60(2), 226. (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Latifpoor, M., Sarzaeem, M. M., Amouzadeh Omrani, F. & Raissi Dehkordi, S. Comparison of functional outcomes of two knee arthroplasty techniques (Total knee arthroplasty and unicompartmental knee arthroplasty) for the treatment of Osteoarthritis, simultaneously done in the same patients. Arch. Bone Jt. Surg.12 (10), 695–700 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Klasan, A., Tay, M. L., Frampton, C. & Young, S. W. High usage of medial unicompartmental knee arthroplasty negatively influences total knee arthroplasty revision rate. Knee Surg. Sports Traumatol. Arthrosc.30 (9), 3199–3207 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Pritchett, J. W. Disease progression after unicompartmental arthroplasty: add a compartment or revise to total knee arthroplasty? J. Arthroplasty. 37 (10), 2004–2008 (2022). [DOI] [PubMed] [Google Scholar]
- 18.Lee, J., Tay, M. L., Frampton, C. M. & Young, S. W. Clinical and functional outcomes of TKA after HTO or UKA: a new Zealand joint registry study. Arch. Orthop. Trauma. Surg.144 (9), 4095–4100 (2024). [DOI] [PubMed] [Google Scholar]
- 19.Sun, C. et al. Evaluation of the consistency of the MRI- based AI segmentation cartilage model using the natural tibial plateau cartilage. J. Orthop. Surg, Res.19 (1), 247 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the first author, Wei Zhang, upon reasonable request. Requests for access to the data should be submitted through the journal’s submission system and addressed to Wei Zhang.