Deep-learning based 3D reconstruction of lower limb bones from biplanar radiographs for preoperative osteotomy planning.

PURPOSE: Three-dimensional (3D) preoperative planning has become the gold standard for orthopedic surgeries, primarily relying on CT-reconstructed 3D models. However, in contrast to standing radiographs, a CT scan is not part of the standard protocol but is usually acquired for preoperative planning purposes only. Additionally, it is costly, exposes the patients to high doses of radiation and is acquired in a non-weight-bearing position. METHODS: In this study, we develop a deep-learning based pipeline to facilitate 3D preoperative planning for high tibial osteotomies, based on 3D models reconstructed from low-dose biplanar standing EOS radiographs. Using digitally reconstructed radiographs, we train networks to localize the clinically required landmarks, separate the two legs in the sagittal radiograph and finally reconstruct the 3D bone model. Finally, we evaluate the accuracy of the reconstructed 3D models for the particular application case of preoperative planning, with the aim of eliminating the need for a CT scan in specific cases, such as high tibial osteotomies. RESULTS: The mean Dice coefficients for the tibial reconstructions were 0.92 and 0.89 for the right and left tibia, respectively. The reconstructed models were successfully used for clinical-grade preoperative planning in a real patient series of 52 cases. The mean differences to ground truth values for mechanical axis and tibial slope were 0.52° and 4.33°, respectively. CONCLUSIONS: We contribute a novel framework for the 2D-3D reconstruction of bone models from biplanar standing EOS radiographs and successfully use them in automated clinical-grade preoperative planning of high tibial osteotomies. However, achieving precise reconstruction and automated measurement of tibial slope remains a significant challenge.

The influence of the weight-bearing state on three-dimensional (3D) planning in lower extremity realignment - analysis of novel vs. state-of-the-art planning approaches.

BACKGROUND: The use of 3D planning to guide corrective osteotomies of the lower extremity is increasing in clinical practice. The use of computer-tomography (CT) data acquired in supine position neglects the weight-bearing (WB) state and the gold standard in 3D planning involves the manual adaption of the surgical plan after considering the WB state in long-leg radiographs (LLR). However, this process is subjective and dependent on the surgeons experience. A more standardized and automated method could reduce variability and decrease costs. PURPOSE: The aim of the study was (1) to compare three different three-dimensional (3D) planning modalities for medial open-wedge high tibial osteotomy (MOWHTO) and (2) to describe the current practice of adapting NWB CT data after considering the WB state in LLR. The purpose of this study is to validate a new, standardized approach to include the WB state into the 3D planning and to compare this method against the current gold standard of 3D planning. Our hypothesis is that the correction is comparable to the gold standard, but shows less variability due compared to the more subjective hybrid approach. METHODS: Three surgical planning modalities were retrospectively analyzed in 43 legs scheduled for MOWHTO between 2015 and 2019. The planning modalities included: (1) 3D hybrid (3D non-weight-bearing (NWB) CT models after manual adaption of the opening angle considering the WB state in LLR, (2) 3D NWB (3D NWB CT models) and (3) 3D WB (2D/3D registration of 3D NWB CT models onto LLR to simulate the WB state). The pre- and postoperative hip-knee-ankle angle (HKA) and the planned opening angle (°) were assessed and differences among modalities reported. The relationship between the reported differences and BMI, preoperative HKA (LLR), medial meniscus extrusion, Outerbridge osteoarthritis grade and joint line convergence angle (JLCA) was analyzed. RESULTS: The mean (std) planned opening angle of 3D hybrid did not differ between 3D hybrid and 3D WB (0.4 ± 2.1°) (n.s.) but was higher in 3D hybrid compared to 3D NWB (1.1° ± 1.1°) (p = 0.039). 3D WB demonstrated increased preoperative varus deformity compared to 3D NWB: 6.7 ± 3.8° vs. 5.6 ± 2.7° (p = 0.029). Patients with an increased varus deformity in 3D WB compared to 3D NWB (> 2 °) demonstrated more extensive varus alignment in LLR (p = 0.009) and a higher JLCA (p = 0.013). CONCLUSION: Small intermodal differences between the current practice of the reported 3D hybrid planning modality and a 3D WB approach using a 2D/3D registration algorithm were reported. In contrast, neglecting the WB state underestimates preoperative varus deformity and results in a smaller planned opening angle. This leads to potential under correction in MOWHTO, especially in patients with extensive varus deformities or JLCA. CLINICAL RELEVANCE: Incorporating the WB state in 3D planning modalities has the potential to increase accuracy and lead to a more consistent and reliable planning in MOWHTO. The inclusion of the WB state in automatized surgical planning algorithms has the potential to reduce costs and time in the future.