Generic Implant Positioning Technology Based on Hole Projections in X-Ray Images

Abstract

Implant placement plays a key role in trauma and orthopedics. In this paper, a generic technological concept for implant positioning assistance is outlined. The system utilizes conventional radiographic devices for imaging and tracking and embeds into surgical workflows without the need for complex navigation equipment. It is based on feature extraction from cylindrical hole-projections in X-ray images for determining spatial alignment of implant and anatomy. Basic performance of a prototype system was experimentally verified in terms of tracking accuracy and robustness under varying conditions. In a second step, the system was developed into a set of application modules, each serving a pressing clinical need: Plating of the proximal humerus, cephalic nail and dynamic hip-screw placement, general anatomic plating, distal nail interlocking with adjustment of femoral anteversion and corrective osteotomies. Module prototypes were tested according to their degree of maturity from feasibility assessment in wet-labs to clinical handling tests. Orientation tracking of reference objects yielded an accuracy and precision of 0.1±0.71 deg (mean±standard deviation) with a maximum error of 4.68 deg at unfavorable conditions. This base-performance translated, e.g., into a precision of ±1.2 mm (standard deviation) screw-tip to joint distance at proximal humerus plating, or into a precision of lag screw positioning in the femoral head of ±0.6 mm in craniocaudal and ±1.6 mm in anterioposterior direction. The concept revealed strong potential to improve surgical outcomes in a broad range of orthopedic applications due to its generic and simplistic nature. Comprehensive validation activities must follow for clinical introduction.

1 Introduction

Operative fracture treatment seeks (1) restoration of anatomical relationships by (2) stable fixation of the fracture with (3) preservation to the biological environment to achieve (4) early and active mobilization of the patient [1].

A crucial step in the chain of treatment from acute injury to fracture consolidation is the anatomical reduction of the fracture followed by placement of the implant at a biomechanically and biologically favorable position in order to establish an anatomically sound and mechanically advantageous repair construct [1]. Precision of the surgery is considered a major determinant for the outcome of the operation [2]. Advancements in implant design are often negated by poor surgical execution. A significant portion of the postoperative fixation failures can be attributed to inaccurate implant positioning, particularly emphasized in porotic bone [3]. For example in hip fracture treatment, a centered screw placement in the femoral head at a specific distance to the joint is crucial in reduced bone quality [2]. Other examples with potential impairment of the skeletal function are malrotation of the limb after intramedullary nailing [4] or imprecise alignment after hip and knee arthroplasty.

Implantation of fixation hardware, a core task of orthopedic surgery, is usually performed in a freehand manner under repeated X-ray control. Implant placement is essentially a three-dimensional (3D) problem with the challenge of deriving a spatial understanding from several two-dimensional X-ray projections. For iterative control and verification, the radiologic device (e.g., a C-arm) is repeatedly repositioned to acquire multiple views of the scene. This procedure takes time and generates exposure to radiation [5].

Computer aided surgery (CAS) aims at improving the surgical precision of the average operator while reducing the expenditure of radiation [6]. However, standard navigation solutions are complex and unaffordable for many clinics and require setup-, operating-, takedown- and maintenance efforts unattractive for the daily routine [7].

We believe that the highest potential to improve patient care lies in supporting everyday orthopedic surgery with simple and effective technical aids. The aim of this paper is, hence, to introduce a generic technological concept for intra-operative surgical guidance and to demonstrate its potential for the majority of orthopedic routine interventions.

2 Xin1 Technology

The introduced technological concept (Xin1) aims (1) to improve the general precision of implant placement in routine orthopedic and trauma surgery in order (2) to reduce postoperative complications and (3) to decrease expenditure on time and radiation during surgery. It intends to overcome general drawbacks of conventional CAS systems: specialization, costs, complexity and accessibility, by utilizing standard intra-operative radiology without the need for costly additional devices. Radiology offers currently the most powerful imaging modality for investigating the human body. Currently, the use of conventional X-ray is, however, restricted to plain visualization, although X-ray images offer several opportunities to extract further information. The proposed concept involves use of this so far unused information for more efficient application of radiography in clinics.

2.1 Underlying Principle.

2D projections of a three-dimensional object allow reconstruction of the spatial orientation and position of the object. 2D-3D registration is not a novelty. Prominent examples are computed tomography or isocentric C-arm imaging, where a 3D representation of a scanned object is derived from a set of 2D X-ray projections by mathematical methods such as cone-beam algorithms [8]. When translating this principle further to conventional C-arm imaging, issues arise regarding robustness and computation time. The number of required X-ray views for reconstruction varies largely with the complexity of the target structure (e.g., implant or bone shape).

The proposed concept uses a simple, highly explicit and easy to incorporate reference structure—a cylindrical hole [9]. A single 2D image of a minimum set of two cylindrical holes with known heights and diameters in a radiopaque object allows rapid and robust tracking of the object position in six degrees-of-freedom from distinct landmarks in the characteristic lens-shaped hole projections (Fig. 1). Several objects may be tracked in relation to each other. Besides robust and fast calculation (<1 s per image with a standard computer), the major advantage of the approach lies is the fact that holes can be easily integrated into implants or instruments. Holes can be included in dens metallic objects. In contrast, alternative markers, like bits or spheres always require a lower surrounding density for visibility on X-rays. Opposed to standard CAS systems, a conventional C-arm or comparable radiographic device can be used as imaging means. No additional tracking equipment is required. Surgeons are used to intra-operative C-arm imaging. Standard surgical workflows are hence maintained. Use of radiography allows additional identification of the anatomy such as the location of the femoral/humeral head in the same work-step. No additional patient registration is hence required. Hole-markers and corresponding aiming devices in the sterile field are purely mechanical and can be treated analog to other instruments in terms of handling, cleaning and sterilization. No additional connections (e.g., cables) bridge sterile and nonsterile fields.

Fig. 1.

Illustration of Xin1 principle. From the characteristic lens-shaped perspective projection of a hole in an X-ray image (left) the orientation of an object can be back-calculated (right). If a set of two or more holes is used (not shown here), the registration is explicit in six degrees-of-freedom from a single image. Advantage over other 2D–3D registration methods is, besides robust and fast computing, a simple integration of cylindrical holes in metal objects, such as implants or instruments.

2.2 Clinical workflow.

The general clinical procedure is sketched as follows:

1. An implant and/or an instrument, equipped with hole-markers is prepositioned with respect to the anatomy of the patient.

2. The scene will be statically captured by taking a minimum of one X-ray image (number depending on the application). The C-arm can be arbitrarily oriented under the condition that all relevant hole-markers and the relevant anatomy are visible in the image.

3. The image(s) will be processed by a custom software algorithm to derive the spatial orientation of all relevant elements (instruments, implants, anatomy). The algorithm can either be implemented on the radiographic device or on an external computer connected to the C-arm through available image interface standards.

4. A desired implant position will be planned based on the actual scene. Planning can be performed manually or automatically by image processing algorithms.

5. The required reorientation of the elements relative to each other will be calculated. Correction values or guiding landmarks will be displayed on a screen.

6. Reorientation of the implant/instrument is performed manually, either by an iterative approach or by adjusting a mechanical aiming device.

7. A control image is taken to confirm the desired alignment of instruments, implants and/or anatomy.

In the following this general workflow is illustrated by way of example of various developed prototype application modules.

2.3 Basic System Performance.

A generalized prototype software library was developed in matlab programming language (Mathworks Inc., Natick, MA) capable of performing the following core functions: (1) segmenting X-ray images and automatically finding hole markers, (2) calculating spatial hole-marker orientation in six degrees-of-freedom (three rotations and three translations) in respect to a C-arm based coordinate system (Fig. 7(a)) by means of least-squares error minimization. Various laboratory tests were carried out to verify performance and robustness of the developed algorithm. Selected tests on the most important performance characteristics are described and summarized in the following:

Fig. 7.

Xin1 proximal femur module with modified PFNA insertion handle (hole markers were drilled into the handle): (a) two X-rays are taken at an angle of approximately 30 deg in a–p view, no lateral view is required, (b) image pair is processed by software module and correction values for center-center blade position and blade length are depicted, and (c) optional: mechanical position indicator attached to the insertion handle to execute the suggested correction

2.3.1 Segmentation.

Performance of the custom-made image segmentation algorithm was investigated on a dataset of 212 cadaveric and clinical X-ray images with hole-markers included. 207 images revealed successful marker detection (97% detection rate). Markers in 5 images were not detected because of extensive hole tilt or motion blur (Fig. 2).

Fig. 2.

Exemplary test images from segmentation robustness testing (212 images, 97% detection rate). Left: correctly detected marker-set. Middle: Detection failed due to motion blur. Right: Failed detection due to extensive hole tilt.

2.3.2 Orientation Finding.

Performance of the custom-made orientation finding algorithm was investigated in terms of influence of the following parameters on accuracy and robustness: marker size and geometry, marker position in the X-ray beam (location in X-ray, distance to Image intensifier), image intensity, optical distortion (flat panel versus image intensifier) and image blur.

12 dummy objects with varying hole marker size and configuration were produced from stainless steel. A multifactorial parametric experiment was designed and conducted by taking defined X-ray images of the dummy objects with two different C-arms with conventional image intensifier (Siemens Arcadis Varic) and digital flat-panel detector (Ziehm VisionFDVario3D, Ziehm Imaging GmbH, Nürnberg, Germany).

In summary, accuracy and precision of marker orientation in a first test series (standard image intensifier) was 0.1±0.71 deg (mean±standard deviation) within a corridor of ±30 deg tilt about α and β axes (total 170 test images, Fig. 3). In this test series, the maximum tilt error was 2.03 deg. A three-hole marker appeared as a good compromise of performance and marker size. Accuracy improved with increasing hole-size. Accuracy was neither significantly influenced by the distance of the object to the image intensifier nor by image blur. Accuracy dropped with rising image intensity. A maximum tilt error of 4.68 deg was measured with an overexposed image. Even though accuracy was not different between flat-panel and image intensifier, the maximum tilt error was lower for the flat panel (2.24 deg versus 3.45 deg). Higher errors were seen at the image boundaries compared to the image center when using the image intensifier.

Fig. 3.

Basic performance testing. (a) Laboratory setup to test accuracy and precision of tilt assessment about alpha and beta angles of an exemplary hole marker configuration. (b) Exemplary positioning of hole markers within the X-ray beam of a C-arm. Z dimension defines the distance to the X-ray source. (c) and (d) Exemplary X-rays taken of a hole marker with different orientation and position in the X-ray image (flat panel detector). Circles represent automatically detected hole boundaries.

3 Application Modules

The abstracted principle offers potential to be translated into a variety of applications in trauma and orthopedics. Several prototype modules were developed to demonstrate the generic nature of the concept. The modules have different degree of maturity from laboratory tested to clinically investigated and answer different clinical problems. Various individual validation tests were performed for the modules (Table 1). For sake of conciseness only, the most relevant testing outcomes are presented.

Table 1.

Xin1 application modules with the addressed clinical problem and the so far conducted validation activities

	Application module	Clinical problem	Validation activities	Tracking strategy
3.1	Proximal humeral plating	Fixation failure in porotic bone	Performance cadaver testing, clinical handling testing	Marker attachment
3.2	Proximal femoral nailing	Fixation failure in porotic bone	Performance cadaver testing	Marker attachment/marker integration
3.3	Femoral anteversion/distal nail interlocking	Mal-union/X-ray exposure	Handling cadaver testing	Marker attachment
3.4	Freehand distal interlocking [9]	X-ray exposure	Performance cadaver testing against gold-standard control	Marker integration
3.5	Controlling corrective ostotomies [16]	Mal-union	Performance bench testing, handling cadaver testing, clinical testing	Marker attachment
3.6	Dynamic hip screw insertion	Fixation failure in porotic bone	Handling cadaver testing	Marker attachment
3.7	Locking plate positioning	Fixation failure in porotic bone	Handling cadaver testing	Marker attachment

Generally, two basic tracking strategies were followed:

1. Marker attachment: External attachment of holes to gain maximum independence from implant designs and manufacturers. A marker-clip was developed to attach to cylindrical structures like drill guides (Fig. 4(b)). This is seen as a short-term strategy toward clinical application.

2. Marker integration: Integration of holes into implants or instruments. To fully utilize the inherent advantage concept, holes must be incorporated into metal instruments and implants. This is the long-term vision of the project, since in many cases modification of existing devices is required.

Fig. 4.

Xin1 proximal humerus module in cadaveric trial. (a) OR setup: tablet computer connected to C-arm via PACS. (b) Prepositioned plate with guide block, drill sleeve and marker clip attached. (c) mevis software prototype. Screw trajectories (green) are projected into the image. From an image pair all screw lengths to a desired tip-joint-distance are displayed on the screen. (d) confirmation image after screw insertion.

Table 1 relates the tracking strategies to the application modules. Developed modules are introduced in the following.

3.1 Proximal Humeral Plating.

Proximal humeral fractures are frequently stabilized by anatomical plates with four to nine angular stable locking screws extending into the humeral head segment. Despite plate position [3], the challenge is to choose the screw lengths such that screw tips anchor in the subchondral bone for superior purchase. Too short screws may lead to fixation failure, particularly in the elderly osteoporotic patients [10]. Too long screws penetrate the joint and may trigger reoperation. Incidence of postoperative mechanical complications at the proximal humerus is high [11]. The spherical shape of the humeral head in combination with oblique screw trajectories and plain X-ray projections render the screw insertion process particularly difficult, resulting in extensive use of radiation and surgery time. An efficient guiding system could improve the surgical outcome of this operation markedly.

Based on the described concept, we developed an Xin1 proximal humerus prototype module, tested its feasibility in a human cadaveric setting and conducted a clinical handling test. A marker-clip comprising a metal body with three holes and polymer clamps was designed to attach to commonly used drill sleeves (Fig. 4(b)). The clip was designed for multiple use and repeated autoclaving cycles.

The software core-library (2.3) was translated into C++ code for performance reasons and foreseen product development with stringent regularity requirements. A graphical user interface was first sketched in matlab and then implemented in mevislab Software in collaboration with Mevis Breastcare GmbH (Mevis Medical Solutions AG, Bremen, Germany). Software was installed on a medical grade tablet computer (Intel i5 processor) with touch sensitive screen and Windows 7 operating system. The software can handle analog video input signals and digital imaging and communications in medicine (DICOM) image transfer. Here, an implemented picture archiving and communication system (PACS) server communicated with the C-arm (Siemens Arcadis Varic with image intensifier, Siemens Healthineers, Erlangen, Germany) via cable connection (Fig. 4(a)). The software comprises an implant database, which was loaded with geometrical information of a standard proximal humerus plate (Philos, DepuySynthes Inc., Oberdorf, Switzerland).

Feasibility of the prototype was tested in a fresh frozen human cadaveric specimen. A trained trauma surgeon performed the operation. After prepositioning the plate on the intact proximal humerus with soft tissue mantle, a standard drill-sleeve was inserted, and the marker-clip was attached. The nose of the clip locks into the Philos guide block to constrain clip rotation (Fig. 4(b)). Two anteroposterior X-ray images of the shoulder were taken at an angle of approximately 30 deg and transferred to the external computer. The boundaries of the humeral head were identified in both images by semi-automatic image segmentation. Potential solutions (circles) are suggested to the operator for final selection. The accuracy of the anatomical segmentation approach, hence, depends on the input of the user. However, a check on distinct quality parameters defining geometrical feasibility and precision of the selected solution is performed by the software as a safety measure to detect false user selection. From a single X-ray projection trajectories of all screws can be computed (Fig. 4(c)). A second X-ray additionally provides 3D information on the position of the humeral head estimated as a perfect sphere. The proximal screws are automatically truncated to the required length to a given tip-joint distance. All screw lengths can be read from the display and immediately inserted without manual length measurement (Fig. 4(d)). Predrilling may be performed but can be restricted to opening the lateral cortex, thereby minimizing the risk of perforating the joint space while drilling.

An accuracy test on Philos implantation was conducted. 34 X-rays were taken from different view angles without moving the instrumented plate relative to the bone. Pairwise combinations of these images were retrospectively processed to compute the required screw-lengths for all nine proximal screws in each image pair. Inclusion criterion for a valid pair of images was a view angle between the images of at least 35 deg resulting in a total of 195 valid pairs and therefore 195 × 9 = 1755 single measurements. The precision of the achieved distance of the screw-tips to the joint surface was ±1.2 mm (standard deviation).

A first clinical handling test was conducted on 10 patients undergoing proximal humeral plating in a single medical center (UZ Leuven, Belgium). The study was approved by the local ethical committee (AFMPS/SE/80M0661, clinical trial NCT03427112). All patients gave informed consent to participate. Image processing was done retrospectively; no system feedback was provided to the surgeon during operation. The study revealed good handling of the marker clips in the operating room (OR) and confirmed the respective study hypothesis claiming that the system performs reliably in a clinical context in the presence of real-life bone fractures (Fig. 5).

Fig. 5.

Proximal humerus case during clinical study (UZ Leuven, Belgium) at prepositioning of the plate. Anteroposterior image processed retrospectively by Xin1 prototype software (graphical user interface by Mevis Breastcare GmbH). Projected screw trajectories suggest that the calcar screws, having particular importance for stability, may not fit into the head at current plate position. All suggested screw lengths to a tip-joint-distance of 5 mm are calculated from one image pair and are displayed at the output panel (left). Circle section: Semi-automatically segmented humeral head.

3.2 Proximal Femoral Nailing.

Proximal femur fractures are among the most frequent of all fractures. They often have devastating consequences especially in the elderly population, if the patient's mobility is not quickly regained. Standard of care for proximal femur fractures with vital and intact joint is cephalic intramedullary nailing, where a lag screw is positioned through the nail along the femoral neck into the center of the head at the defined distance to the joint (tip-apex distance) [2]. Similar to the proximal humerus, accurate implant positioning in compromised bone stock is of utmost importance to avoid mechanical complications [12].

Two prototype modules were developed and tested in a cadaveric setting. In a first approach, a marker-clip was used rendering modifications to existing implants and instruments un-necessary. The clip was attached to the tip of the standard drill-sleeve of the PFNA (Proximal Femoral Nail Antirotation, DepuySynthes Inc., Oberdorf, Switzerland) (Fig. 6(a)).

Fig. 6.

Xin1 proximal femur module in a cadaveric test: (a) marker clip attached to the PFNA drill-sleeve and skin incision, (b) mevis software prototype with displayed correction values derived from one pair of X-ray images to achieve center-center blade position and suggested blade length to a desired distance to the joint, and (c) final implant position in lateral view

Analog to the proximal humerus application processing of two oblique anteroposterior images allows estimation of the screw/blade position with respect to the femoral head. In advantage to current practice, no lateral view is required, which markedly facilities the surgical procedure. Correction values to achieve center-center position of the blade implant in both planes as well as the required implant length to a desired distance to the joint are depicted on the screen (Fig. 6(b)). The position can be iteratively adjusted and controlled until a satisfying result has been achieved (Fig. 6(c)).

Accuracy was investigated on a fresh-frozen human cadaveric lower body specimen. 133 image pairs were generated from 30 individual X-rays (view angle >35 deg). Precision of the measured blade position with respect to the femoral head center was ±0.6 mm in craniocaudal and ±1.6 mm (standard deviation) in anterioposterior direction.

The clip approach has the advantage that it can be directly used with existing implants and instruments without modifications. However, clips can move or detach and interfere with soft tissues. Integration of hole-markers into implants or instruments is therefore the ultimate goal. For proximal femoral nailing an alternative prototype module was developed (Fig. 7). A standard PFNA insertion handle was modified with a set of holes at the proximity of the docking point to the nail (Fig. 7(b)). The procedure is performed analog to the previously described process. Optionally, to avoid iterative repositioning of the implant, a position-indicator was designed that can be clicked into the insertion handle. This mechanical device allows reading depth and rotation of the nail with respect to the bone (Fig. 7(c)). The position-indicator is temporarily pinned to the proximal femur by means of a Kirschner-wire.

3.3 Femoral Anteversion/Distal Nail Interlocking.

In intramedullary nailing bone segments can rotate around the nail before locked in place with interlocking bolts. For example, in femoral nailing the knee is often mal-rotated after surgery [4], which can lead to orthopedic issues in the long term. Installing an anatomically favorable rotation (anteversion) at the femur is, hence, key to a good outcome [4].

Another demanding step in the operation is the actual insertion of the distal interlocking bolts. Unlike proximal interlocking, a rigid aiming arm for drill and screw guidance cannot be used due to unpredictable nail deflection inside the intramedullary canal [13]. The alignment of aiming arm and interlocking hole is, hence, lost with the bending of long nails.

Following the described principles, a universal aiming arm was developed to aid in both closely related tasks, anteversion correction and distal interlocking. The arm is proximally attached to the nail's insertion handle (Fig. 8(a)). The distal section of the aiming arm comprises a set of hole-markers for tracking, several guiding holes to receive drill sleeves, and one translational degree-of-freedom to compensate for nail bending. The arm is compatible with several nail families (here all DepuySynthes Femur- and Expert Tibia nails) despite different shapes through a flexible and lockable connector.

Fig. 8.

Distal interlocking concept with universal aiming arm and femoral anteversion correction in cadaveric trial. (a) Universal aiming arm attached to a PFNA insertion handle. The flexible arm allows precalibration to various femur and tibia nails. Distally the arm enables compensation of nail bending. (b) Software module depicting current femoral anteversion and correction value for distal locking from a single X-ray. (c) Verification after correction.

Prior to operation, the aiming arm is calibrated to the specific nail in undeformed state. After nail insertion into the intramedullary canal and attaching the precalibrated aiming arm, a single X-ray of the distal femur region in roughly mediolateral orientation is taken and transferred to the external computer. The custom software routine processes the image by relating the hole markers to the interlocking hole of the nail and delivers a correction value to compensate nail bending (Fig. 8(b)). The compensation set screw is then adjusted accordingly to achieve alignment between drill-axes and interlocking holes.

Before inserting the interlocking bolts, anteversion of the femur is checked. The posterior aspects of the medial and lateral femoral condyles are identified in the recent radiograph by semi-automatic segmentation. The current anteversion of the femur is estimated by relating the spatial condyle position to the orientation of the femoral neck, known from proximal locking (see Sec. 3.2). Anteversion can be iteratively adjusted to a desired value. Finally, a control image is taken to verify correct alignment before the nail is locked distally into place (Fig. 8(c)).

3.4 Freehand Distal Interlocking.

Common surgical practice for distal targeting of intramedullary nails is a freehand process under fluoroscopic control [13–15]. After aligning the C-arm with the hole axis indicated by circular appearance of the holes in the X-ray, the drill tip is manually positioned into the center of the hole projection under X-ray control followed by aligning the drill axis with the C-arm axis to drill the hole. This process is demanding particularly for the young surgeons and involves considerable use of radiation at each step of the procedure [13–15].

For cases where a freehand process is still preferred over an aiming arm approach (3.3), an alternative freehand variant of the system was developed. This was published previously in Ref. [9]. From a single X-ray, the interlocking holes of the nail are detected to calculate the hole axes with respect to a C-arm set of coordinates. Guiding circles are projected into the image to guide a specifically designed drill sleeve with radiopaque structures into place. The procedure controls drill-bit positioning and orientation by iterative repositioning. In contrast to the gold standard freehand procedure, alignment of C-arm and hole axis is not required. A cadaveric study for distal locking of tibia nails was performed on ten fresh-frozen human lower leg specimens. Comparing the newly developed method to the gold standard showed a 58% reduction of X-ray images and a 22% reduction of procedure time when using the system as compared to the gold-standard [9].

3.5 Controlling Corrective Osteotomies.

When long bone rotation was not installed properly during intramedullary fixation of a fracture, patient inconvenience and orthopedic problems such as arthrosis can be the consequence [4]. Reoperation involving a corrective osteotomy procedure is indicated in severe cases. Still, after reoperation, considerable deviations from the planned value can persist [4]. We believe that this is mainly due to the lack of intra-operative feedback in current practice.

An Xin1 module for controlling corrective osteotomies was developed and published previously in Ref. [16], involving two reference flags with hole markers intra-operatively attachable to Schanz-pins, as routinely positioned during surgery at both sides of the osteotomy. The custom software routine calculates the spatial relation between both reference flags from X-rays before and after the osteotomy to provide intra-operative feedback on the degree of performed anatomical correction. This module was extensively tested in vitro and was used in a first clinical trial on 12 patients to test its feasibility and clinical value under real-life conditions (BGU Tübingen, Germany). The study was approved by the local ethical committee (309/2016BO1). All patients gave informed consent to participate. Clinical handling of the marker flags appeared appropriate. The system measured rotational correction intra-operatively with an error of 1.61 deg ± 0.86 deg (unsigned mean±standard deviation) as compared to postop computer tomography (CT) measurements as baseline [16] (Fig. 9). In comparison, according to the literature, 33% of the patients suffer from residual deformities greater than 10 deg when operated with current practice [4].

Fig. 9.

Xin1 corrective osteotomy module in clinical trial. (a) Patient from BGU Tübingen, Germany, equipped with hole-marker flags attached to the Schanz-pins undergoing rotational correction osteotomy at the femur. (b) X-ray after correction with hole markers.

3.6 Dynamic Hip Screw Insertion.

Analog to previously described Xin1 modules, the marker-clip was further attached to the insertion handle of the dynamic hip screw (DHS, DepuySynthes Inc., Oberdorf, Switzerland) to facilitate the aiming process during femoral neck fracture fixation. The workflow is in accordance with the cephalic nailing procedure as described in Sec. 3.2. A prototype module for DHS application was developed, and its feasibility was proven in a cadaveric setting (Fig. 10).

Fig. 10.

Xin1 prototype module for DHS positioning in cadaveric trial. (a) A marker clip is attached to the DHS insertion handle. (b) Software module suggesting correction values for center-center lag screw positioning and screw length to a desired tip-joint distance derived from an X-ray image pair. (c) and (d) Final implant position in a-p and lateral views.

3.7 Locking Plate Positioning.

The Xin1 principle was further applied to a variety of anatomical plates (locking compression plates (LCP), DepuySynthes Inc., Oberdorf, Switzerland) with fixed- or variable angle locking options. The marker-clip attaches to standard drill-sleeves of different diameters and is designed for engaging with the combi-holes of 3.5 and 5 mm LCP plates (Fig. 11). Testing was performed for two exemplary plates, namely, the LCP for distal- and proximal femur (both DepuySynthes Inc., Oberdorf, Switzerland).

Fig. 11.

Xin1 use with anatomical locking plates. (a) Marker-clip attached to drill-sleeve and LCP distal femur plate (DepuySynthes Inc.). (b) Xin1 software module for the proximal femur locking plate in cadaveric test with projected virtual screw trajectories.

4 Discussion

Orientation in anatomical terrain is a key task in surgery. With the first attempts in the field of stereotaxis at the beginning of the 20th century, computer assisted surgery made its way into the operation theater [6,17,18]. Standard surgical navigation systems mainly rely on pre-operative 3D medical imaging such as CT, which is afflicted with a considerable radiation dose [19], and confront hospitals with a major investment. Tracking devices, mainly based on optical motion tracking, are required in the OR in order to interlink the patient with the operational plan [20]. This step called patient registration is currently the most challenging issue in the field of computer aided surgery and impacts accuracy and procedure time [17,21]. Consequently, currently CAS solutions are mainly used for specific and complex cases where the additional efforts are outweighed by the benefits.

This paper introduces a comprehensive and generic implant and anatomy positioning technology targeting orthopedic routine interventions. The system does not require extensive navigation equipment and embeds into established surgical procedures. It relies on 2D-3D registration from intra-operative fluoroscopy rendering the technology simple, inexpensive and efficient. The distinguishing feature from other known 2D–3D registration approaches is the use of cylindrical holes as tracking markers, which are, if not already present in implant or instruments, easy to integrate and explicit to track in X-rays. This underlying principle was translated and developed into a variety of system modules serving clinically relevant surgical needs.

All modules were built on a custom developed software core-library providing common functions. Robust performance of the core algorithm became apparent from laboratory testing under varying circumstances. Initially, the test was used to tune any systematic tracking error (accuracy) toward zero. Precision (measure for random error) of orientation tracking, as most important outcome measure, was <1 deg (standard deviation) under conditions deemed to reflect normal operation. Under extreme conditions (e.g., overexposed images), the maximum detected error rose to ∼5 deg (see Sec. 2.3). Consequently, the algorithm was programmed to exclude mal-composed images automatically. Automatic detection of holes in the X-ray functioned reliably due to the characteristic silhouette, which is considered an important advantage of the proposed approach. The basic performance of the core algorithm is translated into application specific performances of the respective modules. Accuracy and precision depend on several factors such as hole marker size and distance to the region of interest and may, thus, vary with the application. For proximal humerus plating, for instance, the precision of screw length estimation was found to be 1.2 mm (standard deviation). For proximal femoral nailing, precision was lowest in orthogonal direction to the imaging plane (anteroposterior, 1.6 mm). For corrective osteotomies, the precision found in clinical investigation (0.86 deg) [16] verified the results of the core library test. Overall, the findings suggest sufficient performance for the respective clinical applications. However, additional testing with a dedicated control group, such as a freehand gold-standard procedure, would allow appropriate weighting of the data and should be performed in the future.

The technical implementations are in a prototype stage for proving the value of the guiding concept in a practical context. Presented modules have different degrees of maturity ranging from idea phase, over bench tested to clinically investigated (two clinical studies finished). One selected pilot application is underway to be commercially available as a medical device. Generally, future activities will be directed toward medical device development according to ISO13485. Each module development must run according to a specific development plan with a risk-based verification and validation strategy.

We believe that the general concept of X-ray based simplified navigation carries strong potential to sustainably transform orthopedic and other surgery. An already available navigation device following similar principles is the Adapt for Gamma3 System (Stryker Leibinger GmbH & Co. KG, Freiburg, Germany). As opposed to holes, this technology uses radiopaque bits as tracking means in plain X-rays. Currently, the system supports implantation of a single implant (Gamma3 proximal femur nail, Stryker Trauma GmbH, Schoenkirchen, Germany) in terms of lag screw positioning and distal interlocking [22–24]. Even though well designed, it is still considered a niche product for a confined use. Such stepwise approach into clinical translation is suggested by considerable regulatory requirements and efforts in the medical device domain. The key for a wide-spread use of such technology is, however, genericity and applicability for the vast majority of applications, throughout all fields of orthopedics, trauma, spine and other disciplines. This paper intended to draw such comprehensive picture of the concept.

The incidence of suffering a bone fracture is reported to be greater than 10 in 1000 individuals per year [25]. It is, thus, estimated that worldwide about 50 × 10⁶ traumatic fractures occur annually whereof approximately 8 × 10⁶ receive operative treatment with conventional trauma implants. There is, hence, a considerable number of patients, who could benefit from faster surgery, reduced exposure to radiation and reduced complication rates through higher surgical precision. For hip fixation, severe mechanical complications range between 3.6 and 12% [12,26]. For proximal humerus fractures, complication rates are markedly higher [11]. Reasons are manifold. Implant positioning is believed to play a major role [3]. A properly placed shoulder or hip implant could prevent mechanical issues and could encourage the surgeon to mobilize the patient earlier, which is crucial in particular for the elderly.

Radiation is a prevailing problem in the OR. For instance, tibia and femoral nailing require fluoroscopy times between 0.56 and 6.95 min of radiation [14,27], which in the long-term can become a health-threatening issue in particular for the OR personnel [5,27–30] with a relative risk for cancer of 5.37 with respect to the general population [5]. Radiography as a tool of daily use should be carefully applied, as radiation exposure accumulates over the life of the health care professionals. The Xin1 freehand module for distal interlocking (Sec. 3.4) led to a 50% reduction of X-ray images [9]. For distal interlocking with a mechanical aiming arm (Sec. 3.3), the reduction rate is believed to be even higher, since two images are effectively needed for placement of two screws. This, however, needs to be verified in a clinical context.

Advantages became evident to a group of surgeons. In 2012, we demonstrated at the time available prototype modules (femoral nailing Secs. 3.2 and 3.3) in a hands-on fashion to 41 expert surgeons and collected their feedback (Fig. 12). From our interpretation of the survey results physicians start realizing the potential value of such simplified digital aids embedding into the known surgical routine. However, benefits need to be proven in practice on an individual basis.

Fig. 12.

Extract from a surgeon survey conducted in 2012 after hands-on demonstration of an Xin1 femoral nailing prototype. 41 participating surgeons from world-wide origin. All senior physicians.

4.1 Limitations

4.1.1 Real-Time Ability.

Conventional radiography, as available in hospitals around the globe, is utilized. Due to its static nature and the inert response of image intensifiers, the proposed concept involves placement of an implant to a predefined position via distinct static stages as opposed to dynamic position tracking. This drawback becomes more apparent in freehand tasks like pedicle screw insertion in spine, which would benefit from dynamic positioning feedback. Consequently, the current scope of the system mainly targets applications without strong real-time requirements.

4.1.2 Accuracy.

The system accuracy depends on a variety of factors. Main determinants are the actual application, the size of the hole-markers, and the used X-ray device. Current image intensifiers provide areal resolutions of 4–5 pixels/mm. Older C-arms offer significantly lower resolutions. The more pixels constitute the hole boundaries in the image, the more accurate the calculation becomes. There is, hence, a tradeoff between marker-size and accuracy. E.g., for the distal interlocking problem (Sec. 3.3), submillimeter accuracy is needed to guide the drill-bit through the nail-hole. Thus, the holes for tracking are dimensioned larger. Optical distortions of conventional image intensifiers due to electromagnetic- or the earth's magnetic field (pincushion or S-distortion) further impact the system accuracy. These effects could be eliminated by linearization with a calibration grid prior to use of the system [24]. However, this means additional setup efforts and is desired to be avoided. Sketched applications in this paper perform sufficiently without linearization. However, distortions are C-arm specific and difficult to predict. Flat-panel detectors eliminating these effects, are currently on the uprise and steadily replace old image intensifier technology on the market.

Besides distortions, performance of the system relies on the general image quality. Several factors can compromise the function such as overlays of elements in the projection, motion blur or under/over exposure of the images. In particular segmentation of anatomical structures and edges such as the silhouette of the femoral head can be challenging in noisy images and may depending on the application, directly impact accuracy. To prevent calculation errors and, hence, potential harm to the patient, the image processing algorithm involves detailed quality checks on various test parameters (e.g., shape and symmetry of the lens-shaped hole projections or local image exposure), which need to be passed before any result is displayed to the surgeon.

4.1.3 System Integration.

In its current implementation, the system requires interfacing with the imaging device. Cross-platform compatibility can be challenging due to a variety of image transfer standards on the market (DICOM/PACS, video graphics array, high definition multimedia interface, S-Video, Composite Video, etc.). Some C-arm manufacturer offer image interfaces only as additional upgrade. Furthermore, in many cases important parameters, such as image-magnification, flip or rotation, are not transmitted together with the image raw data. Desirably, third party software tools enhancing the value and use of plain X-rays shall be deployable directly to the C-arm to avoid interface issues. This, however, faces certain regulatory hurdles, which will hopefully be tackled by C-arm manufacturers in the near future.

4.1.4 Radiation.

As mentioned earlier, a key advantage of the method is believed to be the use of standard radiography. On the other hand, radiation is a health hazard and should be minimized. Even though the net requirement of the method (1–2 X-rays for calculation) is low, one must consider the gross amount of radiation, involving initial images to adjust the field of view to all required features and elements, as well as control images to confirm the success of the procedure. A conventional laser pointer on the C-arm might already significantly reduce the amount of unnecessary X-ray shots. The effective potential to reduce radiation by means of the Xin1 system and similar devices needs to be proven in a clinical setting.

4.1.5 Testing.

Most of the performed verification and validation activities represent preliminary tests with limited sample size and potential operator bias. However, this paper intends to provide a brief system overview only. Upon completion of the various application modules and comprehensive testing, detailed individual reporting must follow.

5 Conclusion

Orientation in anatomical terrain, a key task in surgery, requires experience of the operator, takes time and is often accompanied by intense radiation. Particularly for routine surgeries, use of computer assistance is often inefficient due to high costs, high specialization and handling complexity. A new concept is proposed for simplified implant positioning utilizing conventional radiographic devices as imaging and tracking means. The concept embeds into current surgical workflows without the need for complex navigation equipment. As opposed to other approaches, the method uses lens-shaped hole projections from intra-operative X-ray images as tracking markers to guide the implant into place. As proven by prototype applications with different degrees of maturity in preclinical and clinical studies, the concept carries strong potential in terms of improved surgical precision and reduced radiation exposure in a broad range of applications.

Funding Data

• AO Foundation via the AOTRAUMA Network (Grant No. AR2009_04; Funder ID: 10.13039/501100001702).