Outcomes of mobile bearing unicompartmental knee arthroplasty in medial osteoarthritis knee with and without preoperative genu recurvatum

IJES-OA Net: A residual neural network to classify knee osteoarthritis from radiographic images based on the edges of the intra-joint spaces

Among the musculoskeletal disorders in the world, osteoarthritis is the most common, affecting most of the body joints, especially the knee. Clinical radiographic imaging methods are commonly used to diagnose osteoarthritis thanks to their cheapness and availability. Due to the low quality and indiscernibility of these images, however, accurate osteoarthritis diagnosis has always faced inaccuracies, such as the wrong diagnosis. One of the osteoarthritis hallmarks is joint space narrowing. Thus, its degree and severity can be determined relatively by assessing the space between the bones in the joint. Therefore, in this research, a deep residual neural network, termed IJES-OA Net, is presented to automatically grade (classify) the severity of knee osteoarthritis via radiographs. This is achieved by tuning it in a way to have it focused on the distance of the edges of the bones inside the knee joint. Experimental results which are conducted on MOST (for training) and OAI (for validation and testing) datasets show that the IJES-OA Net achieves high average accuracy as well as average precision (80.23% and 0.802, respectively) while having less complexity compared to other methods. Additionally, the resulting attention maps from IJES-OA Net are accurate enough that increase experts' reliance on the provided results.

Navigated Unicompartmental Knee Arthroplasty: A Different Perspective

Backgroud

Anteromedial osteoarthritis is a recognized indication for unicompartmental knee arthroplasty (UKA). Favorable postoperative outcomes largely depend on proper patient selection, correct implant positioning, and limb alignment. Computer navigation has a proven value over conventional systems in reducing mechanical errors in total knee arthroplasty (TKA). However, the lack of strong evidence impedes the universal use of computer navigation technology in UKA. Therefore, this study was proposed to investigate the accuracy of component positioning and limb alignment in computer navigated UKA and to observe the role of navigation in proper patient selection.

Methods

A total of 50 knees (38 patients) underwent computer navigated UKA between 2016 and 2018. All operations were performed by the senior surgeon using the same navigation system and implant type. The navigation system was used as a tool to aid patient selection: knees with preoperative residual varus > 5° on valgus stress and hyperextension > 10° were switched to navigated TKA. We measured the accuracy of component placement in sagittal and coronal planes on postoperative radiographs. Functional outcomes were also evaluated at the final follow-up (a minimum of 16 months).

Results

Nine patients had tibia vara and 14 patients had preoperative hyperextension deformity. We observed coronal outliers for the tibial component in 12% knees and for the femoral component in 10% knees. We also observed sagittal outliers for the tibial component in 14% knees and for the femoral component in 6% knees. There was a significant improvement in the functional score at the final follow-up. On multiple linear regression, no difference was found in functional scores of knees with or without tibia vara (p = 0.16) and with or without hyperextension (p = 0.25).

Conclusions

Our study further validates the role of computer navigation in desirable implant positioning and limb alignment. We encourage use of computer-assisted navigation as a tool for patient selection, as it allows intraoperative dynamic goniometry and provides real-time kinematic behavior of the knee to obviate pitfalls such as significant residual varus angulation and hyperextension that predispose early failure of UKA.