(Radiographics. 2001;21:183-190.)
© RSNA, 2001
Computer-aided Surgical Planning for Implantation of Hearing Aids Based on CT Data in a VR Environment1
Florian Dammann, MD,
Andreas Bode, MD,
Erwin Schwaderer, MD,
Michael Schaich, PhD,
Martin Heuschmid, MD and
Marcus M. Maassen, MD
1 From the Departments of Diagnostic Radiology (F.D., A.B., E.S., M.S., M.H.) and Otorhinolaryngology (M.M.M.), University Hospital Tübingen, Hoppe-Seyler-Strasse 3, D-72076 Tübingen, Germany. Presented as a scientific exhibit at the 1999 RSNA scientific assembly. Received May 8, 2000; revision requested June 12 and received July 24; accepted July 27. Address correspondence to F.D. (e-mail: florian.dammann@med.uni-tuebingen.de).
 |
Abstract
|
|---|
A study was undertaken to assess the feasibility of a preoperative fitting test for an implantable hearing aid in a virtual reality (VR) environment. High-resolution spiral computed tomography (CT) of the mastoid bone was performed, and the results of a mastoidectomy were simulated with manual segmentation on a standard medical workstation. CT was also performed on a temporal bone specimen obtained at real mastoidectomy, and the bone margins were segmented automatically with threshold-based techniques. A triangulated surface representation of the bone structures including the mastoid cavity was generated. These data as well as the computer-aided design (CAD) files of the medical devices were transferred into a VR environment. The CAD components of the hearing aid were manipulated to simulate the surgical implantation procedure. Merging CAD data of an implantable hearing aid with CT data of the temporal bone in a VR environment was shown to be a feasible method of providing three-dimensional information for the presurgical determination of fit and mountability. Advances in hardware and software are expected to improve the usability of this method. Although clinical studies are needed, these results may serve as an impetus for exploring the use of low-cost, widely available VR computer equipment in a potentially broad field of clinical applications.
Index Terms: Computed tomography (CT), computer programs • Computed tomography (CT), utilization, 213.456 • Computers, simulation Ear, anatomy, 21.92 • Ear, CT, 21.1211 • Ear, prostheses, 213.456 • Hearing loss
|
Introduction
|
|---|
Inner ear hearing loss is the second most prevalent chronic disease worldwide after arthritis. Almost 15% of the adult population of Western industrialized nations suffer from sensorineural hearing loss (1). These patients usually have malfunction of cochlear outer hair cells, resulting in malfunction of the cochlear amplifier. In the past, auditory rehabilitation was possible only with conventional hearing aids, and despite major technical improvements in recent years, a significant number of patients are still not helped by this type of hearing aid because of unsolvable feedback problems, outer ear canal infections, or dynamic range problems. In addition, insufficient speech comprehension, acoustic distortion, occlusion of the outer ear canal, stigmatization, and impairments in daily life are possible disadvantages of conventional hearing devices. Recently, a hearing system known as a totally integrated cochlear amplifier (TICA) was introduced that may be totally implanted in the temporal bone (Fig 1) (2,3). Complete removal of the mastoid cells is necessary to create a cavity for the implant. However, the temporal bone is a highly complex structure, and anatomic variations may cause significant problems for the implantation procedure (4–8). Hence, a reliable preoperative prediction of the suitable fit of the medical device within the temporal bone is, along with audiometric parameters, a prerequisite for indication of this therapy.
View larger version (115K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Schematic illustrates the TICA elements (main module, transducer, microphone) implanted in the temporal bone after mastoidectomy. The microphone is facing the external auditory canal, and the rod of the transducer is connected to the ossicles.
|
|
Today, axial high-resolution computed tomography (CT) with high-quality multiplanar reformatting is generally accepted as the imaging method of choice for the evaluation of temporal bone anatomy (5,7,9,10). Standard CT postprocessing workstations are still mostly limited to planar reformatting or threshold-based three-dimensional (3D) surface generation (11–13). Recently developed volume rendering techniques are now available on a few high-end machines (14). However, these techniques are not adequate to meet the requirements of preoperative implantation planning.
On the other hand, several reports have already shown CT to be valuable as a basis for computer-aided construction and surgical planning in medical implantation procedures (15–24). Unfortunately, the majority of these projects require large-scale development of dedicated software or the application of expensive and complicated computer-aided design (CAD) systems (25,26). To be practical in a routine clinical setting, an implantability prediction system must have additional properties. Clear and realistic 3D depiction of the anatomic structures and the medical devices within the same environment is required. Furthermore, the manipulation of these objects must be simple and easy to perform. Another requirement is user ability to detect the collision of the manipulated objects, which may otherwise penetrate each other without being noticed.
Today, these criteria are best fulfilled by virtual reality (VR) computer systems. Until now, experience with VR applications in medicine and especially in temporal bone imaging has been limited (27–34). To our knowledge, no reports exist concerning the use of VR for preoperative implantation testing of medical devices. In this article, we demonstrate the feasibility of using a relatively cost-effective combination of commercially available systems including a VR environment to allow preoperative implantation testing.
|
Materials and Methods
|
|---|
The TICA consists of three implantable modules. The main module is implanted subcutaneously behind the auricles and includes a digitally programmed, three-channel audio processor and a rechargeable battery. A microphone is implanted in the posterior part of the auditory canal after mastoidectomy. The membrane of the microphone is covered with the intact skin of the auditory canal. The third TICA component is a piezoelectric transducer, which is completely incorporated into the mastoid cavity. The tip of its coupling rod is connected to the incus (Fig 1).
Data Acquisition
Axial high-resolution spiral CT scans (Tomoscan AV; Philips, Best, The Netherlands) of the temporal bone in a human cadaver were obtained with the following protocol: 1-mm section thickness and table feed, 0.5-mm intersection gap, 70-mm field of view, and 512 x 512 matrix with resulting 0.14-mm in-plane pixel size. A "bone" kernel was selected, and the specimen was scanned after real mastoidectomy had been performed by an experienced otosurgeon. Complete mastoidectomy included total removal of the aerated bone. A CT data set was acquired with the same protocol in a patient who was to receive an implantable hearing aid and had undergone no previous surgery of the temporal bone.
Segmentation
The CT data were transferred to a medical graphics workstation (Easy Vision 4.2, Philips). In the patient data set, the mastoidal air cells of the preoperative temporal bone were "cut out" with the use of software to simulate a mastoidectomy. The inner contour of the cortical bone was marked manually by drawing a spline curve that was subsequently adapted to the corresponding cortical bone landmarks section by section in the total data set, which consisted of 88 sections (Fig 2). The boundaries of the mastoid excavation were defined ventrally by the external auditory meatus, dorsally by the sigmoid sinus, cranially and caudally by the extent of the mastoid pneumatization, and centrally by the semicircular canals. Figure 3 demonstrates the result of this "virtual mastoidectomy" as a surface-shaded display reformatted image obtained on the medical workstation. Finally, the total volume of the data set was reduced by "cutting off" the bone structures adjacent to the mastoid cavity using a cylindric cut. The time needed for careful manual segmentation by an experienced head and neck radiologist was about 45 minutes.
View larger version (164K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2. Axial CT scan of the right temporal bone shows manual segmentation of the inner contours of the aerated mastoid bone (M). Red indicates "cut-out" area, green indicates bone structures selected for visualization, E = external auditory canal, SS = sigmoid E = external auditory canal, SS = sigmoid sinus, TMJ = temporomandibular joint.
|
|
View larger version (79K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Three-dimensional surface-shaded display reformatted image shows the mastoid cavity (M) after removal of the manually segmented volume (cf Fig 2). sinus, TMJ = temporomandibular joint.
|
|
An automated segmentation process was applied to the postoperative specimen (Fig 4) using an attenuation threshold of 150 HU after coarse manual predefinition of the outer contours of the mastoid bone. This process required less than 5 minutes.
Surface Generation
The pixel depth of the segmented images of the temporal bone was reduced to a 1-bit mask by setting all attenuation values exceeding the threshold level to 1. The resulting "black-and-white" images (Fig 5) were then transferred as DICOM (Digital Imaging and Communications in Medicine) data files to an Indigo 2 workstation (SGI, Mountain View, Calif) running software designed for the generation of anatomic models (MIMICS; Materialise, Leuven, Belgium). This software package was used to generate a 3D triangulated surface representation of the segmented CT scans (Fig 6). The number of triangles that are generated by the software can be influenced by predetermining a maximum allowable deviation of the resulting triangulated surface from the original CT data set. To adapt the complexity of the data set to the graphics system used, we chose a deviation tolerance of about 1 mm with a maximum deviation of 15° between the surface facets and the original surface. As a result, approximately 10,000 triangles were generated to depict the surface of the temporal bone. Besides being affected by the parameters of the triangulation process, the number of triangles is mostly dependent on the overall size of the area of interest and the surface structure of the object. The triangulated surface of the temporal bone including the mastoid cavity was exported in a stereolithographic (STL) data format (3D Systems, Valencia, Calif), which was introduced as a CAD file format and is now widely used for STL modeling in medical and other applications.
View larger version (91K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6. Three-dimensional triangulated surface representation generated with the modeling software from the data set of 1-bit mask images (cf Fig 5) demonstrates a lateral view of the temporal bone.
|
|
VR Environment
The STL data were transferred into a VR software package (REALAX; Realax, Ettlingen, Germany) that had been installed on the Indigo 2 workstation. The resulting surface display was of excellent quality (Fig 7) and closely resembled the intraoperative anatomy. The VR software is designed as an open system with a variety of applications already available (eg, architecture, automobile crash test simulation, computer games, medicine). No modification of any part of the software was necessary. Original construction data of the microphone (represented by 144 triangles) and the transducer (570 triangles), the intramastoidal implantable parts of the hearing aid, were available in the initial graphics exchange specification CAD file format and could be imported into the VR software. The surface data of the temporal bone and the CAD data of the medical devices were loaded simultaneously in the real-time environment of the VR software package (Fig 8). A conventional two-dimensional (2D) cathode ray tube monitor was used for visualization. All three objects could be rotated and translated independently with a conventional 2D mouse or 3D mouse ("space mouse"). Freedom of movement was restricted only by a collision detection mechanism, which was a standard feature of the VR software package. The contact of two objects at any location was indicated by red highlighting, and the movement of the object was stopped. First, the transducer was inserted into the mastoid cavity (Fig 9a). The contact between the tip of the coupling rod and the ossicle chain was confirmed with a close-up view through the external auditory meatus (Fig 9b). Finally, the microphone was placed on the posterior wall of the external meatus (Fig 9c).
View larger version (82K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7. Triangulated surface representation generated with the VR software demonstrates a lateral, slightly dorsoventrally angulated view of the temporal bone. The mastoid cavity (M) (especially the distance between the sigmoid sinus [SS] and the dorsal wall of the external auditory canal [E]) is quite large in this case. This figure represents an uncomplicated situation in which clear demonstration of the procedure is facilitated and no impairment of the implantation procedure is anticipated. Figure 8 Triangulated surface representation generated with the VR software demonstrates both the temporal bone and the elements of the implantable hearing device: the transducer (red) and the microphone (blue).
|
|
View larger version (82K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8. Triangulated surface representation generated with the VR software demonstrates both the temporal bone and the elements of the implantable hearing device: the transducer (red) and the microphone (blue).
|
|
View larger version (86K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9a. Triangulated surface representations generated with the VR software demonstrate virtual implantation. (a) First, the transducer (red) is placed in the definitive position in the mastoid cavity. Microphone is shown in blue. (b) Close-up view through the external meatus shows the tip of the coupling rod (red) in contact with the ossicles (arrow) and allows the position of the transmitter to be verified. (c) Finally, the transducer (red) and microphone (blue) are placed in their permanent positions.
|
|
View larger version (92K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9b. Triangulated surface representations generated with the VR software demonstrate virtual implantation. (a) First, the transducer (red) is placed in the definitive position in the mastoid cavity. Microphone is shown in blue. (b) Close-up view through the external meatus shows the tip of the coupling rod (red) in contact with the ossicles (arrow) and allows the position of the transmitter to be verified. (c) Finally, the transducer (red) and microphone (blue) are placed in their permanent positions.
|
|
View larger version (87K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9c. Triangulated surface representations generated with the VR software demonstrate virtual implantation. (a) First, the transducer (red) is placed in the definitive position in the mastoid cavity. Microphone is shown in blue. (b) Close-up view through the external meatus shows the tip of the coupling rod (red) in contact with the ossicles (arrow) and allows the position of the transmitter to be verified. (c) Finally, the transducer (red) and microphone (blue) are placed in their permanent positions.
|
|
|
Results and Discussion
|
|---|
The purpose of our study was to test the feasibility of using a combination of 3D imaging–based computer systems for preoperative evaluation of the fit and mountability of an implantable hearing aid by simulating the surgical procedure. Such a system must be fast, user-friendly, and highly precise for successful clinical application. At the same time, it must be adaptable to different types of medical and technical data formats, universally available, and entail only moderate hardware and software costs. The successful application in otologic implantation described in this article, with its high demands owing to anatomic complexity and the need for great precision, suggests the universal usability of such a system.
To meet these requirements, we chose to combine two software packages that are already commercially available and can be applied without any specific adaptation. The software used for surface triangulation (MIMICS) is one of the standard medical graphics systems already in use at many institutions for a variety of computer-aided radiology and CAD applications. MIMICS and the VR software package (REALAX) do not require high-cost hardware but can be run on medium-performance workstations or personal computers (PCs). For our purposes, the advantages of REALAX as a basic VR software module included relatively moderate cost and ease of handling. REALAX also provides a high number of data import filters, including STL and the most commonly used CAD formats, and supports a variety of input and output hardware devices.
Typical VR features (eg, merging of different data sets in one environment and fitting and packaging [mounting] tests with collision detection) have been prepared for the initial applications of REALAX in architecture, the automotive industry, and interactive computer games. They also proved feasible in the medical application described in this article.
VR simulation of hearing aid implantation allowed us to evaluate the final fitting of the medical device and to determine the feasibility of the implantation procedure. For example, the distance between the posterior wall of the external acoustic canal and the sigmoid sinus is frequently very short (Fig 10) (35), which may complicate or even prevent entry of the implant. In such a case, sufficient space is available for the device at the definitive site of implantation, but the surgical procedure itself is impossible.
View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 10. Clinical photograph (lateral view) of a cadaveral specimen including the mastoid cavity demonstrates typical anatomy following real total mastoidectomy. The frequently narrow space between the anterior wall of the sigmoid sinus (SS) and the posterior wall of the external auditory canal (E) (arrows) may hamper implantation of the hearing device.
|
|
Medical data sets have much greater size and complexity than those in other technical applications. The irregular contours of the anatomic structures require a large number of triangles for precise surface representation. Because of the limited power of the computer hardware used, we had to reduce the number of triangles and thus the overall accuracy of the surface representation. However, the resulting surface deviation tolerance of approximately 1 mm appeared to be acceptable in this initial experience. Because of the resulting maximum frame rate of 4 images per second with 10,000 triangles used for surface generation, the interactive movement of objects was markedly hampered. State-of-the-art PC-based hardware and software allows depiction with more than 10 times the number of triangles with a frame rate that should easily meet the requirements of high-resolution imaging with real-time interaction.
Another drawback of the system configuration we used was the difficulty in recognizing the exact position of the implantable devices within the 3D space of the mastoid cavity owing to the 2D video screen. Moderately expensive 3D rendering techniques that have recently become available may help overcome this problem.
A major drawback of the approach involves segmentation of the prospective mastoid cavity. Manual segmentation is a very time-consuming task. Furthermore, the limitations of a 2D section-by-section procedure result in unavoidable "stair-step" effects along the z axis of the image stack and may markedly impair the clarity of 3D images (Fig 11). To our knowledge, however, no currently available automated segmentation tool yields acceptable results for this purpose. Further research is required for the development of automated segmentation algorithms and interactive manipulation of the medical data sets. In addition, ongoing refinements in PC software and hardware will help improve VR visualization and interactivity.
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 11. On a 3D rendering of the temporal bone with VR software following virtual mastoidectomy, the clarity of the anatomic details is markedly inferior compared with that seen in the postoperative specimen (cf Fig 10) or the VR-generated image based on its automatically segmented CT data set (cf Fig 7). E = external auditory canal, SS = sigmoid sinus.
|
|
|
Conclusions
|
|---|
The merging of radiologic images of individual anatomic structures with CAD data in a VR computer environment is feasible for the simulation of surgical procedures such as the implantation of medical devices. Ongoing research in semiautomated and automated segmentation techniques and advances in 3D rendering hardware will improve the usability of the method presented in this article. Its practical value has yet to be assessed in a clinical study. Nevertheless, our study may serve as an impetus for exploring the use of low-cost, widely available VR computer equipment in a potentially broad field of clinical applications.
|
Footnotes
|
|---|
Abbreviations: CAD = computer-aided design,
PC = personal computer,
STL = stereolithography,
TICA = totally integrated cochlear amplifier,
VR = virtual reality
See the commentary by Dennis
.
|
References
|
|---|
-
Maniglia AJ. State of the art on the development of the implantable hearing device for partial hearing loss. Otolaryngol Clin North Am 1996; 29:225-244.[Medline]
-
Zenner HP, Leysieffer H. Active electronic cochlear implants for middle and inner ear hearing loss: a new era in ear surgery. I. Basic principles and recommendations on nomenclature. HNO 1997; 45:749-757[German].
-
Zenner HP, Leysieffer H. Totally implantable hearing device for sensorineural hearing loss (letter). Lancet 1998; 352:1751.[Medline]
-
Englmeier KH, Haubner M, Losch A, et al. Hybrid rendering of multidimensional image data. Methods Inf Med 1997; 36:1-10.[Medline]
-
Maher N, Becker H, Laszig R. Quantification of relevant measurements of the petrous bone with computerized tomography before cochlear implant operation. Laryngorhinootologie 1995; 74:337-342.[Medline]
-
Mukherji SK, Mancuso AA, Kotzur IM, et al. CT of the temporal bone: findings after mastoidectomy, ossicular reconstruction, and cochlear implantation. AJR Am J Roentgenol 1994; 163:1467-1471.[Abstract/Free Full Text]
-
Mafee MF, Valvassori GE, Kumar A, et al. Tumors and tumor-like conditions of the middle ear and mastoid: role of CT and MRI—an analysis of 100 cases. Otolaryngol Clin North Am 1988; 21:349-375.[Medline]
-
Tomura N, Sashi R, Kobayashi M, Hirano H, Hashimoto M, Watarai J. Normal variations of the temporal bone on high-resolution CT: their incidence and clinical significance. Clin Radiol 1995; 50:144-148.[Medline]
-
Arriaga M, Curtin H, Takahashi H, Hirsch BE, Kamerer DB. Staging proposal for external auditory meatus carcinoma based on preoperative clinical examination and computed tomography findings. Ann Otol Rhinol Laryngol 1990; 99:714-721.[Medline]
-
Gunlock MG, Gentry LR. Anatomy of the temporal bone. Neuroimaging Clin N Am 1998; 8:195-209.[Medline]
-
Howard JD, Elster AD, May JS. Temporal bone: three-dimensional CT. I. Normal anatomy, techniques, and limitations. Radiology 1990; 177:421-425.
-
Andrews JC, Anzai Y, Mankovich NJ, Favilli M, Lufkin RB, Jabour B. Three-dimensional CT scan reconstruction for the assessment of congenital aural atresia. Am J Otol 1992; 13:236-240.[Medline]
-
Kirchgeorg MA, Prokop M. Increasing spiral CT benefits with postprocessing applications. Eur J Radiol 1998; 28:39-54.[Medline]
-
Reisser C, Schubert O, Forsting M, Sartor K. Anatomy of the temporal bone: detailed three-dimensional display based on image data from high-resolution helical CT—a preliminary report. Am J Otol 1996; 17:473-479.[Medline]
-
Besimo C, Lambrecht JT, Nidecker A. Dental implant treatment planning with reformatted computed tomography. Dentomaxillofac Radiol 1995; 24:264-267.[Abstract]
-
Ellis DS, Toth BA, Stewart WB. Three dimensional imaging and computer-designed prostheses in the evaluation and management of orbitocranial deformities. Adv Ophthalmic Plast Reconstr Surg 1992; 9:261-272.[Medline]
-
Ekestubbe A, Grondahl K, Grondahl HG. The use of tomography for dental implant planning. Dentomaxillofac Radiol 1997; 26:206-213.[Abstract]
-
Eufinger H, Wehmoller M, Machtens E, Heuser L, Harders A, Kruse D. Reconstruction of craniofacial bone defects with individual alloplastic implants based on CAD/CAM-manipulated CT-data. J Craniomaxillofac Surg 1995; 23:175-181.[Medline]
-
Esselman GH, Coticchia JM, Wippold FJ, Fredrickson JM, Vannier MW, Neely JG. Computer-simulated test fitting of an implantable hearing aid using three-dimensional CT scans of the temporal bone: preliminary study. Am J Otol 1994; 15:702-709.[Medline]
-
Mankovich NJ, Samson D, Pratt W, Lew D, Beumer J. Surgical planning using three-dimensional imaging and computer modeling. Otolaryngol Clin North Am 1994; 27:875-889.[Medline]
-
Takeshita F, Suetsugu T. Accurate presurgical determination for implant placement by using computerized tomography scan. J Prosthet Dent 1996; 76:590-591.[Medline]
-
Xenakis TA, Gelalis ID, Koukoubis TD, et al. Neglected congenital dislocation of the hip: role of computed tomography and computer-aided design for total hip arthroplasty. J Arthroplasty 1996; 11:893-898.[Medline]
-
Amet EM. Computerized tomography with CT models for contemporary ramus frame implant planning and construction. J Oral Implantol 1998; 24:152-158.[Medline]
-
Casselman JW, Ars B, Van de Heyning P, Koekelkoren E. Preoperative computed tomography in patients requiring a bone-anchored hearing aid. Eur Arch Otorhinolaryngol 1995; 252:401-404.[Medline]
-
Moharir VM, Fried MP, Vernick DM, et al. Computer-assisted three-dimensional reconstruction of head and neck tumors. Laryngoscope 1998; 108:1592-1598.[Medline]
-
Carney AS, Patel N, Baldwin DL, Coakham HB, Sandeman DR. Intra-operative image guidance in otolaryngology: the use of the ISG viewing wand. J Laryngol Otol 1996; 110:322-327.[Medline]
-
Carney AS, Baldwin D, Patel N, Sandeman D. Virtual reality surgery in otorhinolaryngology (letter). BMJ 1994; 309:1233.[Free Full Text]
-
Coleman J, Nduka CC, Darzi A. Virtual reality and laparoscopic surgery. Br J Surg 1994; 81:1709-1711.[Medline]
-
Ecke U, Klimek L, Muller W, Ziegler R, Mann W. Virtual reality: preparation and execution of sinus surgery. Comput Aided Surg 1998; 3:45-50.[Medline]
-
Kuhlen T, Dohle C. Virtual reality for physically disabled people. Comput Biol Med 1995; 25:205-211.[Medline]
-
Krummel TM. Surgical simulation and virtual reality: the coming revolution. Ann Surg 1998; 228:635-637.[Medline]
-
Seipel S, Wagner IV, Koch S, Schneider W. Oral implant treatment planning in a virtual reality environment. Comput Methods Programs Biomed 1998; 57:95-103.[Medline]
-
Seemann MD, Seemann O, Englmeier KH, Allen CM, Haubner M, Reiser MF. Hybrid rendering and virtual endoscopy of the auditory and vestibular system. Eur J Med Res 1998; 3:515-522.[Medline]
-
Frankenthaler RP, Moharir V, Kikinis R, et al. Virtual otoscopy. Otolaryngol Clin North Am 1998; 31:383-392.[Medline]
-
Atilla S, Akpek S, Uslu S, Ilgit ET, Isik S. Computed tomographic evaluation of surgically significant vascular variations related with the temporal bone. Eur J Radiol 1995; 20:52-56.[Medline]
Related Article
-
Invited Commentary • Author's Response
- Michael J. Dennis and Florian Dammann
RadioGraphics 2001 21: 190-191.
[Full Text]
[PDF]
Top
Abstract
Introduction
