A computer aided measurement method for unstable pelvic fractures based on standardized radiographs

The most severe injury observed in an orthopaedic trauma centre are disruptions of the pelvic ring, especially unstable pelvic ring fractures, which are characterized as a posterior pelvic ring fracture with the partial or total displacement of unilateral or bilateral pelvis. In cases of increased displacement and the incidence of complications from fractured pelvis, it is much more difficult to manage these types of complicated fractures. Traditionally, the fracture displacement is assessed using radiological tools, including roentgenography and CT scans. As radiographic assessment remains the standard for preoperative assessment, the first step of management is rapid and precise measurement and the determination of the displacement of the injured pelvis. For a long time, CT based measurements method of pelvic displacements was considered as the “gold standard” [1, 2]. However, regarding the higher cost and patients’ radiation exposure of CT scan, the imaging examinations based on the radiographs were still the most available and convenient tools in most clinical settings.

A review of the literature revealed that displacement assessments of the pelvis relied on three-direction roentgenography, including inlet, outlet and anterior-posterior (AP) plain radiographs [3‐9]. As with other outcome measurement tools, reliability and validity of radiographic measurement methods are of utmost importance if there is an intention or attempt to use them to conduct pre- or post-operative assessment and to correlate them with an operation’s outcome. However, the reliability and validity of current measurement techniques are not strong [8, 10], because of patient factors (the presence of overlapping osseous structures, bowel gas and intestinal contrast materials) and technical factors (unstandardized technique, deficiency of validation) [2].

Nystrom et al. used the Sawbones model to simulate unstable pelvic ring fracture patterns and reconstructed the 3-dimensional (3D) model based on CT scan data [11]. After measuring the pelvic displacement using 3 previous established methods, they found that the measurement of vertical translation and sacroiliac (SI) joint separation as described by Henderson et al. was reliable [3], and the measurement of SI joint displacement as described by Matta et al. was difficult to make and unreliable [7]. Previous measurement techniques mainly focused on translational displacement of the pelvis, whereas unstable pelvic ring fractures often lead to translational or rotational displacement of the pelvis in different planes in 3D space. Solely measuring translational displacement without considering rotational displacements not only increased the measurement error but also neglected important anatomic information, which underlied the following treatment plan.

We simulated standardized X-ray images using Armira software and hypothesized that some formulas can be developed to precisely measure and indicate the translational and rotational displacements of the injured hemipelvis in each plane based on the 2D coordinate information of those representative anatomic landmarks. Ideally, this study could be used to standardize the operative plan and establish a theoretical basis for robot-assisted surgery based on 2D radiographs.

The 25 patients included in this study were of a mean age of 37.4 years at the time of injury, with 17 men and 8 women. The case series included AO/OTA type C (vertical shear) and type B (AP or lateral compression) with unilateral pelvis injured and dislocated.

The evaluation of the translational measurement outcome of the standardized X-rays showed an overall very good agreement between observers with an ICC of 0.9 and 95 % confidential interval greater than 0.75, as seen in Table 3. Table 4 showed that the agreement of rotational measurement outcomes of the standardized X-rays with an ICC of 0.9 (almost perfect). The interobserver reliability of the reader’s impression of the rotational results based on the 3D-CT model was fair overall with a kappa statistic between 0.265 and 0.426 (Table 5), among which the agreement of full-time orthopaedic trauma surgeons was higher than that of orthopaedic trauma fellows, with average kappa statistics of 0.428 and 0.162, respectively (not shown in table).

As the Majeed scale defines 0–2 mm displacement as the anatomic reduction of a fractured hemipelvis [15], we defined the absolute value of rotational parameter between 0 and 2 calculated in 3D model by 3-matic as neutral position. Based on this precondition, the assessment of the rotational parameters between the impression outcome of the 3D model and the displacement measurement calculated by 3-matic showed that the interobserver reliability was in a wide range between 0.634 and 0.169 (Table 6). It can be seen that the agreement of full-time orthopaedic trauma surgeons (number 1–3) was higher than that of orthopaedic trauma fellows (number 4, 5) (Table 6).

As the rotational direction assessed in the standardized X-rays was the same as that of 3D reconstruction group, the interobserver agreement between both groups was evaluated using rotational displacement parameters calculated by formula. Table 7 and 8 showed the interobserver agreement of rotational and translational displacement parameters with both the ICCs being higher than 0.9 (almost perfect).

Currently, radiographic assessment remains the standard for preoperative assessment and the most frequently used outcome measure during studies of the pelvis. As with other outcome measurement tools, reproducibility, reliability and validity of radiographic measurement techniques are of utmost importance if there is an intention to use these to assess outcome. However, many authors argue that the current measurement methods for fractured pelvises lack standardization, well-accepted reliability, and validity [8, 10].

A literature review by Mataliotakis et al. on the radiographic measurement of pelvic fractures showed various measurement systems introduced by different authors [16]. Radiological evaluation on vertical pelvic displacement was mainly dependent on AP view of the pelvis [3, 7‐9]. Matta et al. and Dickson et al. measured the vertical displacement as the difference in the height of the femoral heads in AP view [4, 7]. Henderson et al. and Griffin et al. also used the AP view to measure the perpendicular difference between the most superior points of the bilateral iliac wings as the vertical displacement [3, 17]. Sagi et al. modified the method of Matta and Tornetta and used the superior edge of the femoral head in the outlet radiograph to determine the vertical displacement of the displaced hemipelvis [18]. Lefaivre et al. used the top of the iliac crests in the outlet film to determine the vertical displacement of the displaced hemipelvis relative to its counterpart [8].

Almost all methods for the vertical displacement measurement used by authors were dependent on the pelvic AP X-ray. Actually, both the AP and the outlet radiographs could be used to measure the vertical displacement of the pelvis. The main difference is that which reference system is selected beforehand. With regard to the movements in 3D space, the best method to depict and determine it is to establish a Cartesian coordinate system which includes three mutually orthogonal axes, such as the X, Y and Z axes. If the vertical displacement of the pelvis is measured on the AP radiograph, it means that the vertical axis is considered as an axis of the rotational or translational movements, and another two axes have to be determined using the transverse and the body’s anterior-posterior axes. However, the types of open and close book of the pelvic injuries are the most frequently mentioned type of pelvic injuries which compose almost all the Young-Burgess lateral compression type of pelvic injuries. The open and close book of the pelvic injuries are mainly determined and better demonstrated on the inlet radiographs, which could be considered as the external or internal rotation of the displaced hemipelvis around the axis across the centre of the sacrum and perpendicular to the inlet plane. Based on this precondition, another two axes could be established perpendicular to this axis at the same time, which are the body’s transverse axis and the axis perpendicular to the outlet plane.

Another reason why the physiologic coordination is not chosen is that the oblique degrees of the pelvis are inconsistent among people and vary with a range between 40° and 60° [19], but the pelvis specific coordination could be considered as a relative objective reference system to depict and measure the movement type of the pelvis itself. Thus, although almost all methods for the vertical displacement measurement used by authors were dependent on the pelvic AP X-ray, the measurement results based on the AP view might not correspond to the true one. Boontanapibul et al. compared hemipelvic vertical displacement measurements results based on the pelvic outlet and AP radiographs, and concluded that pelvic outlet radiograph could provide efficient measurements of hemipelvic vertical displacement, but the measurements results based on the AP radiographs were inconsistent [20].

In addition, as severe complications and multiple injuries are common in high-energy pelvic fractures, compared to the hip joint, which is easy to implicate and fracture, the iliac crest seems to be a more consistent anatomic landmark to use as the reference to measure the displacement. Consequently, it is more reliable and precise to reflect the vertical displacement of hemipelvis by measuring the positions of bilateral iliac crests’ superior points in outlet view rather than measuring the hip joints’ positions in the AP view.

For the measurement of AP displacement of pelvis, Henderson et al. used both ischial spine positions in the inlet view as the bony landmark [3]. Sagi et al. also used the inlet radiograph for the AP displacement [18]. Lindahl et al. used pelvic CT scans’ axial sections for the measurement of the AP displacement, even though no details of measurement methods were available [21].

Regarding the rotational displacement, current inconsistent evaluation methods mainly focused on the rotation in the unilateral plane. Lefaivre et al. evaluated the rotational displacement of the hemipelvis in the horizontal plane by measuring the pubic symphysis diastasis in AP X-rays. Dickson et al. measured the difference between the width of bilateral ischiums as the reference reflecting the rotational direction of hemipelvis, and measured the angle between the midline and the quadrilateral plate of the injured side in the CT scan axial section to represent the rotational angle of the hemipelvis [4]. Sagi et al. measured the perpendicular distances between the acetabulum and sacrum in inlet and outlet views, separately, and evaluated the asymmetry of the pelvis in inlet and outlet views by calculating the ratio between the bilateral distances in the respective plane, which was also named as the inlet/outlet ratio method [18]. Keshishyan et al. evaluated the pelvic asymmetry by calculating the ratio between the distances of the inferior aspect of the sacroiliac joint to the inferior aspect of the contralateral teardrop on both sides of the pelvis. As the pelvis represents a three-dimensional structure, any rotational measurements applied should be as representative as possible to its three dimensions. Although the present different methods for rotation displacement were complicated, there has not been a well-accepted validated method thus far. Nystrom et al. developed a rotational assessment method based on the computer-reconstructed radiographs [1]. However, in their study, the pelvic rotational displacements were only measured around two axes perpendicular to the inlet and sagittal planes, respectively.

Lefaivre et al. analyzed almost all the radiographic measurement methods for the rotational displacement of the pelvic fracture and concluded three most reliable methods [22], including the Sagi, the Keshishyan or Smith and the Lefaivre methods [8, 18, 23, 24]. Hereafter, they performed a comparative study to test the interobserver reliability of three commonly used radiographic measurement methods [10]. The cross measurement technique of Keshishyan showed that overall there is very good agreement between observers, with an ICC coefficient above 0.9 [24]. The inlet and outlet ratio method of Sagi showed less interobserver reliability with a coefficient between 0.5 and 0.8 [18]. The absolute displacement methods described by Lefaivre et al. showed the poorest interobserver agreement [8]. The authors analyzed and noted that the more reference lines or marks and the more steps used, the poorer the agreement would be concluded and the greater the error would be magnified.

The aim of this study was to set up a measurement system, which could allow observers to complete the measurement of displacement in a more convenient manner within as few steps as possible. Ideally, it could minimize the disagreement and errors brought by multiple steps and precisely indicate the rotational direction in different planes, which could be used as the basis for preoperative assessment and treatment plan.

As unstable pelvic fractures are classified as a pelvis with the complete disruption of the posterior osseous ligamentous structure, the posterior pelvic ring was the keystone for preoperative assessment and treatment plans. Thus, the measurement technique of the pelvic translational displacement should focus on the area around the SI joint or posterior pelvic ring. Furthermore, we studied the relationship between the positions of bilateral corresponding anatomic landmarks in 3D coordinate systems and the motion of the hemipelvis in different directions, set up a series of formulas to indicate the rotational displacement around three mutually perpendicular axes, and calculated the associated displacement parameters. Regarding the three-dimensional structure of pelvis, to minimize the measurement error, we selected three landmarks with different 3D coordinates dispersed far away from each other, such as ASIS, the iliac aspect of the SI joint, and ischial tubercle. By comparing with the coordinates of the corresponding anatomic landmarks in bilateral hemipelvis, we concluded a series of formulas that could indicate the rotational direction of the injured hemipelvis, based on the variation of the difference between the distance between landmarks of the injured side in one plane and the distance between the same corresponding landmarks of the intact side in the same plane, following the rotation of injured hemipelvis in the same specified plane. For example, if the injured hemipelvis flexed in the sagittal plane, the distance between the ASIS and ischial tubercle would be reduced compared with the distance between corresponding points on the opposite side, and vice versa. The same situation happened in other types of motions in the corresponding plane.

After the previous step where the formula is established, we also need the ratio between the actual distance and the measurement on the X-ray images, by acquiring and setting each femoral head diameter through measuring its 3D model using 3-matic software, and the relationship between the measurement on the X-ray images and the distance between pixels on the 2D coordination system. In the end, the process of all relevant cumbersome calculations could be simplified by the acquisition of special anatomic landmarks’ 2D coordinates in the corresponding planes and carried out in a few mouse clicks. Ideally, this method could reflect the actual spatial rotational displacements of injured hemipelvis more precisely, and direct the manual reduction manipulations in the clinical practices. However, as was the case in most previous measurement methods, the premise of the current measurement system is that the injured hemipelvis was relatively intact. The proposed measurement method could be used by all the researchers, medical students, radiologists and orthopedic surgeons who are familiar with the pelvic anatomic morphology and landmarks and the injury mechanisms, including AO/OTA and Young-Burgess classifications, and displacements characteristics of the unstable pelvic fractures.

Viegas et al. demonstrated that the three-dimensional CT scan reconstruction technique could provide the bone measurement a volumetric accuracy of 94 % and a linear accuracy of 97 % [25]. Nystrom et al. also tested and verified the validity of CT based measurements of fracture displacement [11]. Thus, the measurement on the 3D CT model was regarded as the standard to test the measurements on the standardized X-ray images. According to the results, interobserver reliability of either translational or rotational measurement was high as both ICCs were higher than 0.9, which means the agreement was nearly perfect and would allow subsequent authors to repeat it.

The new method for measuring translational and rotational displacement based on the established formula is reliable and can minimize the errors in pelvic fracture measurements. Furthermore, this study was possible for the standardization of the preoperative plan and used as the theoretical basis for initiating a process of robot-assisted pelvic fracture surgery based on two-dimensional radiographs.