CT data of 213 patients who received a lower extremity CT angiography examination in our hospital from December 2009 to December 2012 were collected. 200 cases met the inclusion criteria, including 137 men and 63 women. The age ranged from 50 to 85, with an average age of 69.41 ± 9.21 years. Inclusion criteria were patients (1) older than 18 years (2) who did not present with femoral head necrosis, (3) severe hip osteoarthritis or rheumatoid arthritis, (4) a hip joint or femur deformity, (5) a history of hip or femur fractures, or (6) a history of hip or femur surgery. This research project was approved by the ethics committee of Chinese PLA General Hospital. As the study was a retrospective survey of medical imaging data and the anonymity of the patients’ data was maintained, informed consent was not required from patients.
All CT data were collected from the same CT machine (Siemens AG, Erlangen, Germany) with the same scanning parameters (120 KV; 210 mA; collimation, 4 mm; table speed, 3–5 mm/sec; and number of slices, 80–100). The slice thickness of CT scans analyzed in this study was 1.2 mm. The 3D models of femur were reconstructed by the threshold segmentation and the interactive editing method in the Mimics software (version 12.0, Materialise, Leuven, Belgium), and a standardized coordinate system for each femoral model was constructed using the method described by Su et al. [14], and the coronal, sagittal, and horizontal planes were defined to avoid interference from body position during the measurement of FNTA. The reconstructed femur model was input into the 3-Matic software (Materialise N.V., Belgium) in STL format, which geometry is triangle mesh.
First, the femoral head surface was marked using the “Wave Brush Mark” method in the software, then the marked triangles of femur head was created a sphere using the “Analyze” method in the software [15]. The center of the sphere was defined as the center of the femur head, namely, point A. Second, point A as the center of the original sphere, its radius was increased by 2 mm to generate a solid ball which can fully contain the entire femoral head and just tangent to the femoral neck isthmus observed with the naked eye, according to the preliminary experiment. The generated solid ball cut the femoral neck to obtain a corresponding section. This section was treated as a fitting circle, with the center defined as point B. Finally, the line connecting point A and point B was defined as the 3D axis of the femoral neck (Figure 1a-b).
A series of continuous vertical sections was established along the axis of the femoral neck with an interval of 1 mm between adjacent sections. The software automatically generated the area of each section, and the smallest cross-section of three adjacent minimum cross-sections was defined as the FNI. The position of the anterior cross-section in which the femoral neck is connected to the greater or lesser trochanter was defined as the FNB. The cross-sectional morphology of the femoral neck was reported as oval-like shape by morphological study [16.17]. In this study, the cross section of the femoral neck was generated into a part with a thickness of 0.5mm. Two lines located on the cross-section of the three inertial axes of the part were defined as the long axes (from anterior top to the posterior bottom of the femoral neck) and short axes (from the posterior upper part to the anterior lower part the femoral neck) of the cross section of FNI and FNB. The method used to determine the long and short axes was defined as the “inertia axis” method (Figure 1c-d).
At the proximal femur, 25% and 35% of the femoral shaft length, cross-sections of the femur were created after the intersection of the femur with the transverse plane [1]. Then, the inner connecting circles of these two cross-sections were created, and the centers of these two circles were obtained. The line through the centers was defined as the axis of the proximal femur, which was distinct from the axis of the femur. The latter was not a straight line but a curve due to the anterior and lateral arch of the femur [18]. Using 3-Matic software, a plane perpendicular to the coronal plane of the femur through these two centers was defined as plane A, and then a plane perpendicular to plane A was defined as plane B, which was also named as the coronal plane of the proximal femur (Figure 2). According to the method introduced by Zhu et al. [13], the plane consisting of the long axis of the FNI cross-section and the center of the femoral head was defined as the long axial plane of the FNI, and the plane consisting of the long axis of the FNB cross-section and the femoral head center was defined as the long axial plane of the FNB (Figure 3). The FNTAs of the isthmus and basilar part were defined as the angles between the long axial planes of FNI and FNB and the coronal plane of the proximal femur, which were measured directly using 3-Matic software (Figure 4). The difference between the isthmus FNTA and the basilar FNTA was defined as the increase in the FNTA (iFNTA).
The intraclass correlation coefficient (ICC) was used to assess the reliability of the measurement method established in the present study. The sample size required in the reliability study was calculated using the formula reported by Walter and Eliasziw [19]. Subsequently, three observers and another observer made three repeated measurements of any 15 pairs of femur samples. Based on the suggestion proposed by Weir [20], a repeated-measures ANOVA was applied to avoid a significant difference in the results of the study. Two-way random and two-way fixed models were used to evaluate inter- and intraobserver reliability [21]. Fifteen paired samples were subjected to repeated FNTA measurements in a random order by one senior attending orthopedic doctor (RYZ) with a minimum of a 24-h interval between trials to evaluate the intraobserver reliability. The same measurements on the same specimens were performed in an independent manner and a random order to assess interobserver reliability by three other doctors (XYS, JXZ and JTL).
The measured data were analyzed using IBM SPSS Statistics software for Windows, Version 21.0 (IBM Corp., Armonk, NY, USA). Pearson’s correlation coefficients (normal distribution) or Spearman’s rank correlation coefficients (Non-normal distribution) were calculated to analyze potential relationships between demographic data (age, height, weight, and BMI) and the FNTA, according to whether the measured data is normally distributed. A paired T-test was used to compare the FNTA between the isthmus and the basilar part, and the FNTA in all subjects was analyzed using a two-way ANOVA. A stepwise linear regression model was applied to investigate the factors influencing the FNTA. Statistical significance was established at p < 0.05.