The primary finding of this study is that the impact of EADs on the lower limb mechanical axis is related to the angle and location of the deformity. The larger the deformity angle and the closer it is to the knee joint, the more pronounced the deviation of the mechanical axis. After performing osteotomies on 50 deformity models with visualized collateral ligament structures, the result indicate that with MA-aligned osteotomies, nine types of deformities resulted in collateral ligament injuries. However, by adjusting the osteotomy within a safe range and allowing some residual alignment deviation, theoretically, most EADs can be corrected through TKA. Only deformities with 25° varus and 25° valgus located at 90% of the position could not be corrected. Adjustments to the prosthesis, including controlled varus and valgus, were shown to mitigate ligament damage, though extreme deformities still posed a risk. This AI JOINT system allowed for real-time observation of the spatial relationship between osteotomy surfaces and collateral ligaments, aiding in safer surgical planning. Regarding the impact of EAD on the lower limb axis, Wolff et al. [14] constructed deformities at different locations along the full length of the femoral mechanical axis in a two-dimensional plane and found that for the same 20° angular deformity, for every 10% closer the deformity is to the joint axis, the lower limb eccentricity increases by 2° Considering that the deformity exists on the anatomical axis of the femur. In this study, the anatomical axis was taken as the axis line, and deformities were created from proximal to distal at 10%, 30%, 50%, 70%, and 90% of the length. For the same 20° angular deformity, for every 20% closer the deformity is to the joint on the anatomical axis, there is a corresponding increase of 3.8° in the lower limb eccentricity. This deviates from Wolff’s conclusion because the length of the femur’s anatomical axis is shorter than the mechanical axis, and the presence of the femur anterior tilt causes the two axes not to be in the same plane.

EAD affects the hip, knee, and ankle joints simultaneousl. Deformities closer to the knee joint have higher impact on it, while being further away from the knee joint have greater impact on the hip or ankle joint [14, 27]. In cases of femoral EAD, although knee alignment abnormalities can be corrected by intra-articular osteotomy to correct the lower limb axis, hip alignment abnormalities persist. In deformities away from the knee joint, although knee alignment abnormalities can be easily corrected by osteotomy, hip alignment abnormalities are more severe. Patients can typically correct hip alignment abnormalities by hip abduction or adduction. However, unlike femoral EAD, ankle alignment abnormalities caused by tibial EAD are difficult to correct through joint movements. Therefore, surgeons should consider the alignment of the hip, knee, and ankle joints comprehensively before clinical surgery, rather than solely focusing on knee alignment.

For knee osteoarthritis with combined EAD, the purpose of surgery is not only to alleviate pain but, more importantly, to restore the alignment of the lower limb. Currently, it is mainly divided into intra-articular compensatory osteotomy TKA and TKA with preoperative or simultaneous osteotomy. Some authors believe that preoperative or simultaneous osteotomy is the main standard for treating EAD because it allows correction in all three planes without requiring extensive soft tissue dissection [28]. While osteotomy can correct the mechanical axis in all three planes, performing it in two stages may lead to increased surgery time and blood loss, and patients may face a prolonged recovery process with potential complications such as osteotomy nonunion or ligamentous laxity [29]. Through meticulous preoperative planning, precise intraoperative alignment, secure fixation (such as long-stem cemented fixation for osteotomy), and close postoperative follow-up, complications can be prevented [30,31,32,33]. It is also crucial to address serious concerns such as ligamentous laxity and peroneal nerve injury during TKA. Additionally, implementing strategies to manage less severe but preventable issues like pin-site infections and osteotomy nonunion is essential. Achieving these outcomes requires a high level of surgical expertise. In elderly individuals with lower hemodynamic reserves and limited consolidation potential, intra-articular compensatory osteotomy TKA has advantages [34]. Wang and Wang proposed a simple method to determine whether coronal plane deformities of the femur can be corrected through intra-articular osteotomy. In cases of femoral EAD, a line is drawn from the center of the femoral head to the center of the knee joint, followed by a perpendicular line passing through the highest point of the femoral condyle. If this line extends beyond the attachment of the collateral ligament (approximately 25 mm around the knee joint line), extra-articular osteotomy is recommended. If the line falls within the attachment of the collateral ligament, intra-articular osteotomy is preferred [35]. They suggested that femoral coronal plane EADs with 20° or when the Center of Rotation of Angulation (CORA) is located outside the metaphyseal area, and tibial deformities with 30°, can be corrected through intra-articular osteotomy [35]. However, the exact location of the collateral ligament attachment cannot be visualized on X-rays and can only be estimated, which limits the clinical application of this method.

Our study accurately reconstructed the collateral ligament structures and visualized them in the preoperative planning system, allowing for individualized prosthesis positioning without damaging the collateral ligaments. Additionally, whether EADscan be corrected through intra-articular osteotomy should be comprehensively analyzed based on both the angle and location of the deformity. For example, in our study, although there was a 25° deformity in the proximal femur, its impact on mechanical axis deviation was minimal, and osteotomy did not damage the collateral ligaments. Hip alignment abnormalities could be corrected by hip adduction or abduction. Moreover, compensatory intra-articular osteotomy requires uneven resection of the distal femur, which may lead to iatrogenic flexion-extension gap inequality and ligamentous instability [36]. Ligament release on the side with the smaller gap can balance the joint in extension. In cases of severe deformity, more aggressive procedures such as lateral femoral sliding osteotomy or sliding medial epicondyle osteotomy can be performed [37,38,39]. A retrospective study by Santiago P. Vedoya and Hernán Del Sel demonstrated that intra-articular bone resections during TKA can effectively correct extra-articular deformities, achieving favorable outcomes over a 10-year follow-up period in patients with angular deformities up to 20° in the femur and 30° in the tibia, with successful results observed in the majority of cases [40].

AI is rapidly transforming the field of orthopedics by enhancing diagnostic accuracy and surgical planning precision, significantly improving patient outcomes [41]. The integration of AI in knee arthroplasty enables further precise, personalized care. A systematic review by Hinterwimmer et al. emphasizes the importance of specific, high-quality data in developing effective AI-Based models for knee arthroplasty, as these models can significantly enhance preoperative planning, intraoperative decision-making, and postoperative outcomes [42]. Supervised learning is useful in predicting outcomes and optimizing surgical techniques, while unsupervised learning can identify patterns in large datasets that may lead to new insights in patient care [17, 43]. The practical deployment of these models was given by Oeding et al. in their two-part guide, requires orthopedic surgeons to understand both the technical aspects of AI and the nuances of orthopedic-specific applications to maximize the potential benefits for patients [18, 19]. Ko et al. developed AI models tailored to orthopedic imaging [41]. In total TKA, particularly in cases involving extra-articular deformities, computer-assisted navigation not only reconstructs complex deformities and precisely aligns the lower limb mechanical axis but also plays a crucial role in personalized prosthesis implantation planning and accurate osteotomies tailored to different deformities [22]. TKA aimed at neutral MA can alter the native knee anatomy, especially in the alignment of deformed lower limbs. To avoid replicating extreme anatomical structures during TKA aimed at neutral MA, P.A.V. proposed a restrictive kinematic alignment (KA) protocol in 2011 [44], providing a hybrid approach between MA and true KA. This method involves adjusting osteotomies within a defined ‘safe zone,’ ensuring tibial and femoral cuts remain within ± 5° of their respective mechanical axes, and maintaining HKA angle within ± 3° of the neutral position. Cortina et al. found that restricted kinematic alignment (rKA) aims to restore natural knee kinematics and provides similar or better clinical outcomes compared to MA, without increasing the risk of implant failure [45]. In this study, the femoral and tibial osteotomies were initially adjusted to ensure that lower limb alignment remained unchanged, with the osteotomy angle of the femur and tibia within a range of 3°. If collateral ligament damage persists after adjusting the femoral and tibial osteotomies, residual varus or valgus of the HKA within 3° postoperatively can be considered. Using MA alignment osteotomy, nine types of deformities resulted in collateral ligament damage. However, after adjusting the osteotomy within a safe range and allowing some residual alignment deviation, theoretically, most EADscan be corrected with TKA. Only varus deformity of 25° and valgus deformity of 25% located at the 90% position cannot be corrected. For patients with severe extra-articular deformities, reducing intra-articular compensatory osteotomy within a safe range and allowing residual alignment deviation can reduce the risk of collateral ligament damage, minimize soft tissue release, and avoid knee instability caused by over-reconstructing the knee joint anatomy.

This study has several limitations. Firstly, during model manufacturing, the cases where unilateral cartilage wear due to EADs could exacerbate the deformity were not accounted for. Additionally, simulating osteotomy through AI does not replicate the soft tissue balance and ligament laxity encountered during surgery. Despite these limitations, the experiment still contributes valuable insights into the effects of different femoral coronal plane EADs on lower limb alignment and ligament damage during TKA, providing a theoretical basis for clinical practice. However, this study’s use of a single healthy volunteer for model generation may limit the generalizability of the findings to a broader population with varying degrees of deformity and different demographic characteristics. Overcoming these challenges will likely require clinical trials with large datasets or intraoperative evaluations that allow for real-time adjustments to correction plans during surgery. While this study primarily focuses on femoral coronal plane deformities and their impact on the knee’s collateral ligaments, the model’s potential application could be extended to include deformities in the sagittal, axial, and rotational planes of both the femur and tibia. Future studies could explore these dimensions to develop a more comprehensive multi-planar correction approach, which may help evaluate the model’s limitations and further refine surgical planning for complex deformities.