In this study, a total of 21 healthy asymptomatic volunteers (males: females, 10:11; mean age, 27 ± 5.86 years) and 13 patients with sprained ankle (males: females, 7:6; mean age, 30 ± 8.14 years) were initially recruited for this prospective study. All asymptomatic volunteers were supposed to undergo bilateral ankle 3D MRI examination, however, 9 volunteers had motion artifacts during scanning and 3 volunteers were reluctant to undergo bilateral ankle MRI examination due to the doubled scanning time (compared to unilateral scan). In the end, 9 volunteers with bilateral ankle 3D MRI examination (9 right ankles and 9 left ankles corresponded to each other) and 12 volunteers with unilateral ankle 3D MRI scan (5 left ankles and 7 right ankles) were enrolled, for a total of 30 ankles collected as Group 1. Five patients underwent bilateral ankle 3D MRI examination (5 right ankles and 5 left ankles corresponded to each other), and 8 patients underwent unilateral ankle 3D MRI scan (5 left and 3 right ankles), for a total of 18 ankles collected as Group 2.

For healthy volunteers, the inclusion criterion was no history of sprain to either ankle. The inclusion criteria for patients were a history of ankle sprain [Ankle sprain is defined as the stretching or tearing of lateral ligamentous complex, deltoid ligament and distal tibiofibular syndesmosis ligaments (35)] and the initial sprain must have occurred at least 3 months before study enrollment. The exclusion criteria for Group 2 were a history of lower limbs musculoskeletal surgery, fracture above the ankle, the contraindication of MRI scan (such as implant of cardiac pacemaker and/or vascular stent and other metal implantations), and individuals who are unable to cooperate with MRI scan. For healthy volunteers, the same exclusion criteria applied in addition to a history of ankle sprain.

MRI was performed using a 3.0-T magnetic resonance system (Ingenia CX, Philips Healthcare, Best, the Netherlands) with an eight-channel phased-array ankle coil. All subjects were placed in the feet first supine anatomical neutral position and scanned using a sagittal 3D isotropic proton density (PD)-weighted turbo-spin-echo (TSE) sequence (VISTA, Philips Healthcare). Sponge pads were placed around the feet to maintain a neutral ankle position and reduce the subject's movement. The magnetic resonance imaging parameters were as follows: TR, 1,000 ms; TE, 34 ms; echo-train length, 63; Field of view (FOV), 180 × 160 × 180 mm³; acquisition matrix, 400 × 356 × 400; acquisition voxel size, 0.45 × 0.45 × 0.45 mm³; reconstruction voxel size, 0.225 × 0.225 × 0.225 mm³; compressed sensing acceleration factor, 6; excitation flip angle, 90°; refocusing flip angle, 65°; number of averages, 2; scan time, 5 min and 56 s.

The FOV covered from the planta pedis to 5 cm above the ankle joint line, and all subjects were positioned and examined by the same radiographer. The sagittal plane, coronal plane, and axial plane 3D MPR images were reconstructed and stored in the standard format of DICOM 3.0.

The DICOM images were imported into the Mimics (Materialize, version 21.0, Belgium) to obtain the ATFL 3D model (Figure 1). The detailed operations of the Mimics in our study are provided in the supporting information.

The image quality of all 3D models was evaluated by two radiologist (XL and MD with 22 and 10 years of experience in foot and ankle radiology, respectively) and two surgeons (DD and LO with 22 and 8 years in foot and ankle surgery) using a 5-point subjective scoring system (36) to determine whether the image quality of 3D model affected the subsequent measurement (Table 1). All evaluators were blinded to the clinical information.

The ATFL's width and thickness in each model were measured on the fibula side, the middle side (halfway between insertion and origin), and the talus side, respectively (Figure 2). When measurements were taken using the 3D MPR data, we selected the middlemost axial layer of ATFL's footprint on the sagittal plane and the coronal plane. At this layer, the middle point of the footprint on the talus/fibula was selected as the measurement point. The distance between the two middle points was measured as the length. Similarly, the thickness was measured at three points on the same layer. Finally, the width was measured on the sagittal plane with homologous measuring sides. For the measurement on the 3D model, length was measured as the distance between the two middle points of the footprints on the talus side and the fibula side. The measurements of width and thickness were similar to the measurements on the 3D MPR images.

To determine the reproducibility of the ATFL measurements on both the 3D model and the 3D MPR images, 10 cases (5 left ankles and 5 right ankles) were randomly picked for repetitive measurement with a 2-week interval conducted by the same foot-ankle surgeon. An additional independent investigator, blinded to the first investigator's measurements, performed measurements on the same 3D models from the 10 random cases. Finally, based on the results of the two investigators, the inter-observer variability was evaluated.

The sample size was determined by using power analysis. We presumed that there was a significant difference between the ATFL measurements from the 3D model and the 3D MPR images when the mean values of the difference were larger than one standard deviation of the mean value, with an α value of.05, and a power of 0.9. With this setting, it yielded an expected paired sample size of 13. All data were statistically analyzed in SPSS (version 26.0 for Windows, IBM, Chicago, IL, USA). Image quality scores and ATFL measurement results were demonstrated as mean ± standard deviation. The data normality was evaluated with the Kolmogorov-Smirnov test. ANOVA was used to compare the ATFL length, width, and thickness between the 3D model and the 3D MPR image. When P > 0.05, the difference was considered not statistically significant.

The ICC values in the intra-observer analysis were 0.99 and 0.93 for the length and the thickness measured with 3D MPR images, respectively, and the ICC values were 0.99 and 0.93 when measured on the 3D models. The ICC values in the inter-observer analysis were 0.99 and 0.93 for the length and the thickness measured with 3D MPR images, respectively, and the ICC values were 0.99 and 0.95 when measured on the 3D models. Taken together, these ICC values indicated good intra- and inter-observer stability for the measurements taken on both the 3D MPR images and the 3D models.

Image segmentation method based on gray level threshold of surface rendering technology, as a result, the given default gray threshold range of soft tissue (245−490 GV) based on proper manual adjustment, eventually we set the gray-level threshold range between 130 and 350 GV, made the large difference of adjacent area of gray threshold, with well efficiency and accuracy in the reconstruction.

The acquisition of thin-slice MRI images is crucial for accurate 3D modeling. In prior studies, the commonly used thickness of clinical ankle MRI is 3–4 mm (14, 40–43). As such, structural loss of the ATFL and the partial volume effect could hamper the establishment of 3D models. With the recent development of fast imaging techniques (compressed sensing in our study), we achieved a layer thickness of 0.45 mm. Our protocol added a 3D high-resolution PD-Weighted image (5 min and 56 s) which reduced the acquisition time by using compressed sensing technology as compared to conventional 2D imaging in all three orientations (acquisition time of about 12 min). This acquisition time was indeed acceptable to the subjects. Cushion pads and sandbags were used to help patients stay in a fixed neutral ankle position and to reduce the field inhomogeneities that may contaminate the resulting images.

The image quality score was 4.57 ± 0.32, indicating that the slight distortion in 3D modeling is negligible and would not affect the observation and measurement (36). According to our study, in contrast to two-dimensional graphics, a 3D model can provide the surgeon with more information about the ATFL injury location and type, as well as the presence or absence of avulsion fractures (Figure 3).

One of the major defects of conventional measurements is the dependance on specific axial and sagittal sections, making it difficult to accurately define the size and injured location of the ATFL due to the oblique ligaments course. To present a complete view of the ligament, Kim et al. (44) captured the ATFL using the regular axial plane by changing ankle position during the scan, leading to complicated scans and patient suffering. Park et al. (45) obtained the CFL views on the sagittal T2WI in the plane at 45° angled projections based on the plantar surface at mid-calcaneus level, making it difficult to read the other structures on coronal plane. 3D MRI has been widely used in recent years (46, 47), which allows image reconstruction in flexible orientations. Moreover, it has been reported that 3D MRI can be used to measure the meniscus position and its size (48, 49) and intervertebral disc (50). Teramoto A et al. (51) suggested that 3D MRI might be a useful modality to visualize both the ATFL and the CFL. However, their approach was still planar image and not intuitive. An advantage of 3D modeling based on 3D MRI is the potential to find the injuries of location, types, and degrees of the ATFL more accurately and intuitively. Such modeling allows for large sample origination/insertion and footprint research of the ATFL and the CFL in living subjects due to its non-invasive nature.

Patients with ankle sprains may undergo 3D MRI scan to diagnose and determine the appropriate course of treatments. Operative therapy is required if instability persists after a comprehensive conservative treatment (52). Ankle arthroscopy, which simultaneously provides good intraarticular vision and lesion treatment, has been widely used recently. The current operative methods for ATFL injuries are anatomic ligament repairment and anatomic reconstruction with tendon grafts. For patients with origination or insertion of the ATFL avulsed, repair was required. If the ATFL was completely torn, anatomic ligament reconstruction was required (33). Whether it is anatomical repairment or reconstruction, the precise location of ATFL's insert point is necessary (31). Technology based on 3D MRI could accurately determine the area and location of the footprint of the ATFL and select the appropriate rivet or bore bit for bone tunnel before the operation begins [as the bone tunnel should not be drilled by the diameter of the graft, but by the size of the footprint (33)]. On the other hand, anatomic reconstruction with tendon grafts includes autografts and allograft tendon grafts. When tendon allografts were chosen, the thick tendons were often trimmed to reach the appropriate diameter, which causes a certain amount of waste. In summary, we can use this technology to perform preoperative simulations in order to determine the location and diameter of the bone tunnel. These simulations would enable the choice of the appropriate drill and allograft tendon grafts to avoid waste in the future. And we could also determine the drill point directly as intraoperative navigation after completely preoperative planning. One drawback in the practice would be the slightly increased post-processing time for the 3D reconstruction.

The measurements in this study went as far as 0.01 mm. However, in the modern practice, manual surgery usually operates in the scale of millimeter precision. Increased precision is necessary in many aspects of orthopedics, and thus modern orthopedics surgery suffers from insufficient accuracy. In fact, the increased use of orthopedic robots has the potential to dramatically improve the operation accuracy (53). Won-Joon et al. (54) reported that robot-assisted anterior cruciate ligament reconstruction using an MRI-based navigation system had improved the accuracy of the femoral tunnel position. Zhang Q et al. (55) had stated that robotic navigation during spine surgery showed good results and had potential for further research. As a result, we would expect that with the aid of orthopedic robots and with proper preoperative planning and intraoperative navigation, it is possible to achieve precision in the formation of bone tunnels to enhance operative ATFL repair and/or reconstruction surgery. In this sense, higher accuracy measurements will lead to a better and earlier understanding of the pathological emergence and progression. Such insights which will ultimately lead to benefits in the surgical repair of the ligament, given sufficient development in surgery technologies.

Our study encountered some limitations. Firstly, MRI produces a static image and can only be measured in a neutral position (one cannot study different joint positions). Secondly, there were no weightbearing during the MRI scans. Thirdly, the sample size was relatively small. Fourthly, the cost of MRI scan may be relatively expensive and not suitable for everyone else. Finally, we did not evaluate the CFL of the lateral ankle complex. In future investigations, a large population prospective study is needed to evaluate, optimize and further establish the ATFL and CFL 3D model, using MRI.