Hip-Op: an innovative software to plan total hip ...

MED. INFORM.

(2002)

VOL.

27,

NO.

2, 71–83

Hip-Op: an innovative software to plan total hip replacement surgery RICCARDO LATTANZI{, MARCO VICECONTI* {, CINZIA ZANNONI{, PAOLO QUADRANI{ and ALDO TONI{ {Laboratorio di Tecnologia Medica, Istituti Ortopedici Rizzoli, Via di Barbiano, 1/10, 40136 Bologna, Italy {C.I.N.E.C.A., via Magnanelli 6/3, 40033 Casalecchio di Reno (BO), Italy

Abstract. This paper describes an innovative surgical simulation software environment for the pre-operative planning of total hip replacement surgery. The software is a CTbased three-dimensional planning environment, with a user-friendly graphic user interface based on the multimodal display visualization paradigm. Although it relies on a fully three dimensional internal representation, this approach represents the relevant anatomical objects by means of multiple views, each simulating a different medical imaging modality familiar to the medical professional. In the Hip-Op program the multimodal display interface integrates four different display modalities: orthogonal radiographs, Blended slices, CT slices, and arbitrary slices. A conventional surface rendering view is also available. The user ‘navigates’ the prosthetic components, which are dynamically selected from a library of available parts, within the CT volume while the implant and the patient anatomy are simultaneously rendered in each specialized view. Beside a consideration of anatomical compatibility, the surgeon may evaluate the planned implant type, size and position, also on the basis of two analysis modules that compute the achieved level of implant fitting and filling. After being evaluated in an internal clinical trial, the software is currently made available as freeware at http:// www.ior.it/hipop/. Keywords: Hip prosthesis; Computer Aided Surgery; Surgical planning; User interface

1. Introduction Every year in the world more than one million patients are treated because of traumatic or degenerative diseases affecting the hip joint. The vast majority of these patients today are treated with a surgical procedure called total hip replacement. This procedure involves the resection of the affected femoral neck and the reconstruction of the joint by means of an endo-osseous prosthetic device called the hip prosthesis. The term total refers to the fact that artificial parts replace the articular surfaces of both the ileum and the femur. The long term clinical outcome of this surgical procedure has significantly improved over the last fifty years. In the 1950s half of the prostheses had to be replaced within five years, while today 90–95% of the implants are still in place after ten years [1, 2]. Most of the early problems affecting these devices (inadequate materials, insufficient mechanical strength, etc) are now solved. However, due to the large number of operations performed, even a 5% failure Author for correspondence; e-mail: [email protected] Medical Informatics & The Internet in Medicine ISSN 1463-9238 print/ISSN 1464-5238 online # 2002 Taylor and Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/ 14639230210150346

72

R. Lattanzi et al.

rate results in the need for more than 50 000 revision procedures every year worldwide, the outcome of which is much less favourable than that of primary interventions. Currently most total hip replacements fail because of aseptic loosening [1–3], which means that the implant loses its intimate connection with the host bone and the implant becomes painful. The appearance of this complication is a complex multi-factorial process, still not fully understood [4, 5]. However, a large number of co-causes (insufficient cement thickness, lack of primary stability, presence of large bone-implant gaps, inaccurate implant positioning, etc) are directly related to the surgical accuracy [5]. The accuracy of the surgical gesture can be improved by accurately planning the intervention in advance, as well as by means of guidance instrumentation. Two approaches are currently explored: robotic surgery and intra-operative navigation. Two robotic devices and many surgical navigation systems specifically designed to perform a total hip replacement are already used in controlled clinical trials [6–12]. All these approaches share the need for an accurate pre-operative surgical planning to establish the optimal conditions that the surgeon should achieve using such specialized instrumentation. Various pre-operative planning programs have been described in the literature. As the traditional way of planning total hip replacement is the superposition of a transparent template of the implant onto a radiograph of the patient’s hip [13], some authors have simply computerized this technique [14]. Other authors suggest the use of Computed Tomography (CT) data instead of radiographs [15]. Various planning programs represent the CT data with a multiplanar reformation, which consists of a threepane window showing cross sections taken on the CT data set using three orthogonal planes selected by the user [6–8, 10, 12]. Other programs use a 3D interactive visualization alone [16, 17] or in combination with the multiplanar reformation representation [10, 12]. The penetration of these computer-aided planning tools in clinical practice is still limited. Most surgeons are familiar only with planning performed on radiographic projections using templates. The systems that replicate such 2D planning are thus user-friendly, but lack the full three-dimensional definition of the implant position, which is required to plan the operation accurately. On the contrary, systems based on CT data, which allow a fully three-dimensional planning, usually have cumbersome interfaces. In most cases, the position of the implant inside the CT data is visualized in orthogonal slices in a three-pane window or using semi-transparent surface rendering views. In a recent study, both these methods were found inadequate, especially when compared to a new visualization paradigm called multimodal display [18]. Last but not least all programs currently available are only aimed at visualizing the position and orientation of the prosthetic components, presuming that the anatomical referencing is sufficient for the surgeon to decide the correctness of the planning. However, in many cases other functional indicators must be considered to determine the adequacy of a given plan. Typical examples are the area of bone contact and the level of press-fit in cementless implants, the thickness of the cement mantle in cemented implants, the resulting skeletal range of motion, the relocation of the hip joint centre and the resulting limb lengthening or shortening.

Hip-Op: software for planning total hip replacement surgery

73

The Hip-Op research project was aimed at the development of a complete surgical simulation software environment for the pre-operative planning of total hip replacement surgery. The software had to fulfil the following basic requirements: . . . .

a CT-based three-dimensional planning environment; a user-friendly graphic user interface based on the multimodal display approach; the possibility of integrating analysis modules, with the aim of providing the surgeon with additional functional data; and complete independence from the type of hip prosthesis or from the intraoperative instrumentation.

The present paper describes such software from a functional and implementation point of view. This paper has to be considered a preliminary report of a novel technique, the advantages of which have not yet been clinically demonstrated. 2. Description 2.1. Overview From observation of orthopaedic surgeons from our institution planning total hip replacement operations it was evident that the current method, based on the superposition of transparent templates of the various prosthetic components onto a radiograph of the hip region, was seriously limited by the two-dimensional nature of the diagnostic image. Certain anatomical features, such as intra-extra rotation are virtually impossible to control in these conditions. This is true also for computer aided planning (CAP) software based on radiographs. This problem is solved using orthopaedics CAP programs based on CT data, commonly associated to intra-operative navigation systems or surgical robots. Unfortunately, as mentioned in the Introduction, the user interface of most of these programs is ineffective. Furthermore, all the currently available CAP programs limit their support to anatomical visualization. Other functional data, such as fit and fill, would be useful while planning a total hip replacement. Consistent with these observations the Hip-Op software was conceived around three main concepts: . .

.

A fully 3D internal representation of the patient anatomy, of the prosthetic components and of their position within the anatomical space. An effective user interface, in which the diagnostic imaging data are represented not as computer scientists expect them, but rather how medical professionals are used to seeing them. This concept is called multimodal display [18] Availability of analysis modules, fully integrated with the planning environment, able to provide not only a visualization of the implant position and size with respect to the patient anatomy, but also functional information which is considered useful by the surgeon to evaluate the correctness of the surgical planning.

74

R. Lattanzi et al.

2.2. Hip-Op data and internal representation The Hip-Op program uses two types of data: the implant components, which are described as polygonal surfaces, and the CT data, which are represented as a scalar field sampled over a regular (but not evenly spaced) 3D lattice. Every prosthetic model has a transformation matrix that controls its location and scaling in the global space. The model location is specified using orientation, position, and scale factors along the co-ordinate axes. All the rotations take place around the origin of the prosthetic model. This origin is initially set at the centre of the implant bounding box, but can be relocated to any other point by the user. 2.3. Multimodal display The graphical interface of Hip-Op is based on an innovative visualization paradigm, which is called multimodal display. That approach is founded on the realization that biomedical professional tends to reject any tool presenting an unfamiliar representation of the information. Indeed skills built up over years of experience are not readily discarded for the unproven and questionable benefits provided by a newly introduced technology. Adopting the multimodal display paradigm, Hip-Op represents the relevant anatomical objects by means of multiple views, each of which simulates a different medical imaging modality familiar to the medical professional. In the Hip-Op program the multimodal display interface integrates four different display modalities into one consistent user interface. Each modality is displayed in a different window. Two windows are opened automatically at the launch of Hip-Op and, together with the command bar, they fill all the space on the screen. The RX-window represents the anatomy as two orthogonal radiographic images, synthetically generated from the CT data, plus a merged view from two CT slices, one proximal and the other distal, to visualize the femoral neck anteversion (figure 1). The CT-window simultaneously displays six computed tomography images selected by the user from those available by using an intuitive colour-coded lines grid superimposed on the radiography views (figure 2). The user can open two additional display windows. The AS-window displays an arbitrary slice of the data set and of the prosthesis (visualized as a coloured contour) (figure 3). The position of the slicing plane, interactively defined by the user, is visualized in the radiographic views. The 3D-window shows both the prosthesis and the bones with a surface rendering (figure 4). Bones surfaces can be generated directly using a marching cube algorithm, or alternatively imported following creation via an external segmentation procedure. This window is static in the sense that the user cannot move the prosthesis with respect to the bones, but can only change the rendering point of view (pan, zoom and rotate). 2.4. Analysis modules Two analysis modules are currently integrated in Hip-Op to provide clinically relevant three-dimensional indicators of the implant fit and fill in the host femur [19]. The density map computes the contact areas of stem with a heterogeneous density environment (figure 5). The user can define a low and a high threshold (LT and HT), expressed in Hounsfield number and the contact area is evaluated for the three corresponding ranges of density (less than LT, between LT and HT, and higher than HT). In order to give significance to the ranges of density,


75

Figure 1. The RX window. On this window the CT data set is represented as two orthogonal projection (bottom) plus a merged view (top) of two CT slices selected to make evident the femoral anteversion.

a colour masking tool allows the user to select for LT and HT two value of density that correspond to the lower density value of spongy bone and to the lower density value of cortical bone respectively. The distance map module calculates the fractions of stem area whose distance from a surface at a constant density is respectively more than 1 mm (gap), less than 71 mm (penetration), or ranges from 71 to 1 mm (contact). The results are provided both in values and percentages, and visualized with different colours (figure 6). The results of both maps are also reported in a numerical form in order to allow quantitative evaluations. To improve the communication with the surgeons, these are reported for the whole implant surface and also for twelve anatomically defined sub-regions of interest, called Gruen’s regions [20], which are commonly used in the orthopaedic literature to report bone-implant interface observations. Hip-Op defines automatically these sub-regions but allow the user to adjust such partition (figure 7). Other modules are currently under development and will be included in future releases. Among them an automatic fitting module for cementless stems, which will also estimate the implant primary stability, a range of motion analysis module and a few specialized analysis modules for revision surgery and for cemented implants.

76

Figure 2.

R. Lattanzi et al.

The CT window. On this window six CT slices are represented selected by positioning coloured lines in the RX window.

2.5. Standard planning procedure Any patient to be planned with Hip-Op must first undergo a CT scanning of the hip region. CT data are exported in standard DICOM3 format. Hip-Op reads the DICOM files and displays the CT slices in a rendering window. The user can see all images of the CT stack, and using a cropping tool is able to select the portion of data containing the affected hip. The program generates a volume from the CT data, called CT dataset, and automatically sets up the HipOp planning environment. DICOM files contain the full bit-depth of a CT image (4096 values). As only 256 grey levels can be visualized with standard graphic boards, some data windowing is required. The user is able to adjust grey scale to obtain the best visualization, before starting the planning session. The implant database contains geometric models of the available prosthesis components, sorted by prosthesis manufacturers and device’s name. The prosthetic models are imported inside the 3D space of the CT dataset and an optimized algorithm allows the user to change type and size of the components dynamically, in order to select the best assembly. The prosthetic models can be translated and rotated toward the desired position either by mouse or by more precise interactors. The planning can be checked visually using the 3D-window and the ASwindow. An additional feedback, both quantitative and qualitative, is obtainable


77

Figure 3. AS window. On this window is displayed an arbitrary slice of the data set.

from the analysis modules. The user also relies on numerical indicators providing information about the joint centre relocation. Once the implant position is considered appropriate the user may simulate the reaming of the femur canal or the neck resection using the Boolean subtraction tools. A specific measurements panel allows users to take linear or angular measures on any window and save those values properly labelled in a text file, in order to have numerical information of the planning reproducible during surgery. A complete HTML report of the planning session can be saved for later use, which contains the surgeon ID, patient’s personal data as derived from the DICOM files, CT exam data, planning results (type, size and position of each implant component), analysis results, information on the Boolean operations performed, image snapshots of the RX window with the implant in the final position, any saved measurement, etc. CT datasets and implant positions can also be save in separated files for later re-loading into Hip-Op. In principle, implant position data eventually could be exported in any format compatible with intra-operative navigation systems or surgical robots, although such integration has not yet been tested. 3. Implementation 3.1. Implementation tools Hip-Op was developed using Visual TCL (http://vtcl.sourceforge.net/), a freely available application development environment that generates pure Tcl/Tk

78

R. Lattanzi et al.

Figure 4. 3D window. On this window the bones are shown with a surface rendering. Bones surfaces can be generated using a marching cube algorithm or alternatively imported.

code. Tcl stands for Tool Command Language and it is a scripting language, similar to other UNIX shell languages [21]. Tk is its associated GUI toolkit. All the visualization functions were based on the Visualization Tool Kit (VTK), which is an open source, freely available software system for 3D computer graphics, image processing, and visualization [22, 23]. VTK has an object-oriented structure, and its class libraries are written in C++. TCL interpreter can be easily integrated with the VTK classes, allowing the creation of fully functional user interfaces with advanced visualization and interaction capabilities with a moderate programming effort. VTK provided also the main data structures used in Hip-Op. Implant models were represented with VTK polydata and the CT data with VTK rectilinear grids. All the implants data are stored in the Hip-Op implant database. Commercial data are distributed only in encrypted form through a password-based encryptor using DES-EDE2 (128-bits key) and HMAC/SHA-1 (160-bits key). The encryptor was developed using Crypto++, which is a free C++ class library of cryptographic schemes (http://www.eskimo.com/~weidai/cryptlib.html). 3.2. Main algorithms The main algorithms are implemented using Tcl/Tk procedures, which encase commands of the VTK library. The algorithm used to compute the density maps uses mainly two functions: vtkProbeFilter and vtkClipPolyData. vtkProbeFilter has two inputs: a geometric structure (i.e. polydata of the prosthetic stem) and a source


Figure 5.

79

Density map. The density map displays with different colours the areas of the stem surface in contact with the heterogeneous density environment.

data volume (i.e. CT dataset). The filter estimates the source data values at each specified geometry locations by trilinear interpolation. vtkClipPolyData clips the polygonal data which means that it actually ‘cuts’ through the cells of the datasets, returning everything greater (or lower and equal, at choice) than an input scalar value, including ‘pieces’ of a cell. The algorithm used to generate the distance maps is based on vtkDistanceFilter that computes distances from iso-surfaces. The filter has two inputs: the geometric structure (i.e. polydata of the prosthetic stem) and the source data volume (i.e. CT dataset). The vector distances are estimated from iso-surfaces of defined value obtained by interpolating into the source data along the normal at the input point locations. In some cases it has been necessary to develop new specific filters to extend the VTK library. For instance, vtkMaskPolyDataFilter has been developed to create the procedure for the Boolean subtraction. This procedure executes a Boolean subtraction between a geometric structure (i.e. polydata of the prosthetic stem) and a source data volume (i.e. the CT dataset). The filter takes the CT dataset as main input and the geometric structure as input mask. Other important parameters are the greater distance from the polydata to be considered during filter computation and the distance from the geometric structure surface to be considered for subtraction. The filter establishes for each point of the CT dataset the closest point of the polydata. Then, if the distance between those two

80

Figure 6.

R. Lattanzi et al.

Distance map. The distance map computes the distance of the stem surface from an iso-density environment and visualizes the results with different colours.

points is minor than the value set as filter parameter, the scalar value corresponding to the lowest possible density is assigned to that dataset point. 3.3. Accessory services As mentioned above, Hip-Op imports CT data directly in standard DICOM format. The DICOM reader, implemented as a VTK input filter, based on vtkImageUnpacker, can read any of the nearly 3000 tags defined in the standard DICOM dictionary. However, only those tags that are relevant for specific application are processed. Technical data on the CT dataset, such as pixel spacing or attenuation data calibration, are used within the Hip-Op program. Details on the patient and on the CT exam are included in the working session HTML report. 4. Distribution Hip-Op version 1.0, 1.1 and 1.2 were intended for internal research purposes only. Version 1.2 in particular was used to perform a large part of the validation studies and the clinical trial at the Istituti Ortopedici Rizzoli. The first public release was Hip-Op version 1.3, which was released in September 2001. The current release, which was used for this article, is Hip-Op 1.4 and it has been available as freeware at http://www.ior.it/hipop/ since March


Figure 7.

81

Gruen’s regions. Hip-Op allow users to define the Gruen’s sub-regions for the stem by adjusting the position and orientation of coloured planes.

2002. Hip-Op copyright is property of Istituti Ortopedici Rizzoli and C.I.N.E.C.A. The original Tcl source code is not available. Users can download Hip-Op in a compiled form, together with the graphical libraries and the implant database, after accepting the terms of a license agreement. The Hip-Op technology is also currently used to develop a specialized CAE software environment for the design of hip prostheses called JPD, aimed to evaluate the anatomical and functional compatibility of a new design on a digital database of anatomies. 5. Future developments Hip-Op version 1.5, planned for release around June 2002, will be a multilingual version, supporting commercial prosthesis in the implant database. Version 1.6, scheduled for release in summer 2002, should contain optimized graphic performance, especially for the 3D view. A range of motion analysis modules, currently under development, should also be included. The last planned version, Hip-Op v1.7, is currently planned for release in 2003. It will contain more analysis modules and possibly an integrated finite element analysis solver. After this release, Hip-Op 1.x will go in maintenance mode, while the new HipOp 2.0 is under development. This new program will be based on a radically

82

R. Lattanzi et al.

different software infrastructure, the Multimod Application Framework that will be developed in the next three years by the MULTIMOD European consortium, in the framework of a project partially supported by the European commission Fifth Framework Program. This new program should allow remote execution of analysis modules on high-performance computers, support for immersive visualization, haptic and vocal interfaces, and seamless integration with multiple medical imaging data and movement analysis data. Acknowledgements The authors would like to thank Luigi Lena for the illustrations and Mauro Ansaloni for the support during the development. This work was partially supported by European Commission through the JPD project (contract IST1999-20343) and by the Spinner consortium. References

1. HERBERTS, P. and MALCHAU, H., 2000, Long-term registration has improved the quality of hip replacement: a review of the Swedish THR Register comparing 160,000 cases. Acta Orthopaedica Scandinavica, 2, 111–121. 2. HAVELIN, L. I., ENGESAETER, L. B., ESPEHAUG, B., FUNES, O., LIE, S. A. and VOLLSET, S.E., 2000, The Norwegian Arthroplasty Register: 11 years and 73,000 arthroplasties. Acta Orthopaedica Scandinavica, 4, 337–353. 3. LUCHT, U., 2000, The Danish Hip Arthroplasty Register. Acta Orthopaedica Scandinavica, 5, 433– 439. 4. PIZZOFERRATO, A., CIAPETTI, G., STEA, S. and TONI, A., 1991, Cellular events in the mechanisms of prosthesis loosening. Clinical Materials, 1, 51–81. 5. CHAMBERS, I. R., FENDER, D., MCCASKIE, A. W., R EEVES, B. C. and GREGG, P. J., 2001, Radiological features predictive of aseptic loosening in cemented Charnley femoral stems. The Journal of Bone and Joint Surgery, British Volume, 6, 838–842. 6. TAYLOR, R. H., JOSKOWICZ , L., WILLIAMSON, B., GUEZIEC, A., KALVIN, A., KAZANZIDES, P., VAN VORHIS, R. and YAO, J., 1999, Computer-integrated revision total hip replacement surgery: concept and preliminary results. Medical Image Analysis, 3, 301–319. 7. GRUENEIS, C., RICHTER, R. and HENNIG, F. F., 1999, Clinical introduction of the CASPAR system: problems and initial results. Proceedings of 4th Int. Symp. Computer Assisted Orthopaedic Surgery, Davos, Switzerland. Computer Aided Surgery, 4, 160. 8. BARGAR, W. L., BAUER, A. and BORNER, M, 1998, Primary and revision total hip replacement using the Robodoc system. Clinical Orthopaedics and Related Research, 354, 82–91. 9. BRANDT, G., ZIMOLONGM, A., CARRAT, L., MERLOZ, P., STAUDTE, H. W., LAVALLEE, S., RADERMACHER, K. and RAU, G., 1999, CRIGOS: a compact robot for image-guided orthopedic surgery. IEEE Transactions on Information Technology in Biomedicine, 4, 252–260. 10. O’TOOLE , III, R. V. JARAMAZ, B., DI GIOIA, A. M., VISNIC, C. D. and REID, R. H., 1995, Biomechanics for preoperative planning and surgical simulations in orthopaedics. Computers in Biology and Medicine, 2, 183–191. 11. SUGANO, N., SATO, Y., SASAMA, K., NAKAHODO, S. and YODEN , T., 1999, Control of limb length and femoral anteversion in total hip arthroplasty using Optotrak. Computer Aided Surgery, 6, 343. 12. DIGIOIA, A. M., JARAMAZ, B., BLACKWELL, M., SIMON , D. A. and MORGAN, F., 1998, The Otto Aufranc Award. Image guided navigation system to measure intraoperatively acetabular implant alignment. Clinical Orthopaedics and Related Research, 355, 8–22. 13. CAPELLO, W. N., 1986, Preoperative planning of total hip arthroplasty. Instructional Course Lectures, 35, 249–257. 14. BONGINI , D., CARFAGNI, M. and GOVERNI, L., 2000, Hippin: a semiautomatic computer program for selecting hip prosthesis femoral components. Computer Methods and Programs in Biomedicine, 2, 105–115. 15. SUGANO, N., OHZONO, K., NISHI, T., HARAGUCHI, K., S AKAI, T. and OCHI, T., 1998, Computedtomography-based computer preoperative planning for total hip arthroplasty. Comput Aided Surgery, 6, 320–324.


83

16. HANDELS, H., EHRHARDT, J., PLOTZ, W. and POPPL, S. J., 1999, Computer-assisted planning and simulation of hip operations using virtual three-dimensional models. Studies in Health Technology and Informatics, 686–689. 17. HANDELS, H., EHRHARDT, J., PLOTZ , W. and POPPL, S. J., 2000, Virtual planning of hip operations and individual adaption of endoprostheses in orthopaedic surgery [In Process Citation]. International Journal of Medical Informatics, 58, 21–28. 18. VICECONTI, M., LATTANZI, R., ZAUNONI, C. and CAPPELO, A., 2002, Effect of the display modality on the spatial accuracy of orthopaedic surgery pre-operative planning applications. Medical Informatics and the Internet Medicine, in press. 19. TESTI, D., S IMEONI , M., ZANNONI, C. and VICECONTI, M., 2001, Validation of two algorithms to evaluate th interface between bone and orthopaedic implants. Computer Methods and Programs in Biomedicine, submitted. 20. GRUEN, T. A., MCNEICE, G. M., and AMSTUTZ, H. C., 1979, ‘‘Modes of failure’’ of cemented stemtype femoral components: a radiographic analysis of loosening. Clinical Orthopaedics and Related Research, 141, 17–27. 21. WELCH, B. B., 1997, Practical programming in Tcl and Tk, 2nd edition (Upper Saddle River, New Jersey: Prentice Hall). 22. SCHROEDER, W., MARTIN, K., and LORENSEN, B., 2000, The visualization toolkit user’s guide, (Clifton Park, New York: Kitware Inc). 23. SCHROEDER, W., MARTIN, K., and LORENSEN, B., 1998, The visualization toolkit, 2nd edition (Upper Saddle River, New Jersey: Prantice Hall).