ABSTRACT
Craniofacial reconstruction surgery requires precise measurement of human faces and skulls. In this research, the developed procedure combines data from various non-contact precise imaging sensors, i.e. laser scanning via VIVID910 (together with stereo photogrammetric) and CT scan, for imaging the human faces and skulls respectively. This paper focuses on the data processing for generating 3D computer models (of the faces and skulls) and precise measurement of the models. The collected data are processed using the following software: RAPIDFORM, DVP and 3DSLICER for laser, photogrammetric and CT data respectively. The results show that the procedure is applicable for craniofacial reconstruction.

1.0 Introduction
Digital era has changed the technique and procedure in health to help patient lives in a better way. The three dimensional (3D) acquisition of data opens new perspectives for medical research and surgery. The rapid development in information processing technology allows quick and more accurate determination of spatial points (Heinrich & Falk, 2000). To date, many research projects worldwide deal with medical science applications.

Medical applications (such as craniofacial reconstruction surgery) require precise measurement (and modeling) of human faces (or soft tissue) and skulls (or hard tissue). Our preliminary research indicated that most surgeons (in Malaysia) are still relying on laborious traditional contact method (for example, calipers) for measuring anthropometric landmarks on human face (Halim et al, 2004)

Currently, various computer packages (such as CASSOS™ and DENTO-FACIAL PLANNER™) have become available that have partially replaced the manual method of simulating orthognathic and maxillofacial operations (Balvinder et. al, 2002).

Obtaining a 3D image of the underlying skeletal hard tissue is routinely carried out using computerized tomography (CT) or magnetic resonance imaging (MRI) and for 2D images using lateral X-ray. Many techniques for 3D soft tissue capture are available: biostereometrics, morphanalysis, laser scanning, direct digitization, Moire scanning, stereolithography, ultrasonography and stereo video techniques. The most promising method of soft tissue capture is stereo photogrammetry (Balvinder et. al, 2002). This involves the use of a pair of stereo video cameras to capture a stereo image pair of each side of the face, and the software allows the construction of a photo-realistic 3D facial model. The model can be rotated, translated and dilated on the computer screen

This multi-disciplinary research focuses on the development of surgical planning system for craniofacial reconstruction, and comprises of three main parts: imaging, database and planner. The developed imaging procedure combines data from various non-contact precise imaging sensors, i.e. laser scanning via VIVID910 (together with stereo photogrammetric) and CT scan, for imaging the human faces and skulls respectively.

2.0 Research Contribution
Part of the research output is a craniofacial database for Malaysian population. The database is useful for applications such as craniofacial reconstruction, designs of glasses, helmets, etc. It is anticipated that this research will transform the medical approach and method in handling patient's health data management for Malaysia.

3.0 Data acquisition and processing
In this research (Figure 1), the data acquisition stage captures soft tissue (via laser scanning and stereo photogrammetry), hard tissue (from CT scan) and dental cast (by laser scanning). Other additional data such as X-ray and MRI can be included too. The combination provides advantages in terms of rapid 3D modeling (via laser scanning) and precise measurement (via photogrammetric) of the soft tissue (Halim & Mohd Sharuddin, 2004). A total of 102 landmarks (44 landmarks for soft tissue, 25 for hard tissue, and 33 for dental cast) were identified in this research.

This paper focuses on the data processing for generating 3D computer models (of the faces and skulls) and precise measurement of the models. The collected data are processed using the following software: RAPIDFORM, DVP and 3DSLICER for laser, photogrammetric and CT data respectively. The relevant results from each software are stored in our in-house database system (Figure 2) called "Malaysia CranioFacial Morphology Database Client Application" (Deni Suwardhi, 2005).


Fig 1: Data acquisition and processing procedure

Fig 2: Craniofacial Database (Deni Suwardhi, 2005)
3.1 CT Scan image acquisition
CT scan data of human skull was obtained from the Radiology Department of Universiti Sains Malaysia. Figure 3 shows the CT scan of real patient in 2D and processing procedure via 3DSLICER. Each scan typically consume several megabytes, using current spatial resolution and image depth, and could potentially consume over hundreds of megabytes with increasing resolution and depth (Deni Suwardhi et. al, 2004). The average thickness of CT scan data are 1.25mm, depending on the medical requirements and patient conditions.


Fig 3: CT scan data (2D) and processing procedure via 3DSLICER
In one CT scan, a total of 170 slices were captured and stored as DICOM images on a CD-ROM or hard drive. The images were then imported into 3DSLICER Software to build 3D skull (Figure 4 to 8).


Fig 4: Front view of 3D SLICER

Fig 5: Import CT scan data (Dicom format) into 3DSLICER

Fig 6: Thresholding the CT scan image to extract skull

Fig 7: Final result using 3DSLICER software

Fig 8: 3D modeling result in STL binary format
3.2 Laser scanning
VIVID910 laser scanner system (Figure 9) uses a laser beam to measure an object, and has the capability to rapidly record the whole measurement (about 0.3 sec (fast mode), 2.5 sec (fine mode), and 0.5 sec (color mode)). There are 3 main advantages of VIVID910, i.e. speed, precision, and simplicity. The accuracy (Z, typically) of laser scanner are within 0.008 mm using fine mode. VIVID910 employs 3 types of mounted lens, depending on the object sizes and measurement distances. VIVID910 comes with Polygon Editing Tool (PET) software for real time scanning and data processing (Halim et al, 2005). The process comprises of four steps: scanning of human face, three dimensional image registration, three dimensional modeling of human face, and automatic measurement of anthropometric landmark (Zulkepli, 2004). The data are then cleaned and smoothed using RAPIDFORM. After that, the data are converted into STL binary format (contain only point and wireframe exclude texture). Figure 10 shows the procedure of processing laser scanning data with RAPIDFORM.


Fig 9: Konica Minolta VIVID910 laser scanner and processing procedure

Fig 10: Processing procedure for laser scanning
The procedure consists of imaging using 2 laser scanners, registration of 2 shell, merge, clean data and smoothing (Figure 11). The end result from Rapid Form is a 3D model of human faces in STL binary format.


Fig 11: Rapid form processing procedure
3.3 Three-dimensional stereo photogrammetry image
A stereo photogrammetry platform is used to provide precise 3D craniofacial landmarks. The outcome is landmarks from 3 pairs of stereo images of a face (with 44 landmarks). The data are processed using Digital Video Plotter (DVP) software, a stereo based photogrammetric system. This system utilizes computer with 2 monitor for data processing and stereoscopic glasses for viewing stereo images in 3 dimensional (Halim & Mohd Sharuddin, 2004).

The developed image capturing system combines the laser scanning using Minolta Vivid 910 (L1, L2) and stereo photogrammetric (C1, C2, C3) techniques for acquiring high-resolution 3D models of craniofacial soft tissue (Zulkepli et al, 2004) (Figure 12). Figure 13 shows the procedure of data processing with DVP. The result is 3D landmarks that will be used for precise landmark measurement (Figure 14 to 16).


Fig 12: Configuration of set-up system (laser and camera)

Fig 13: DVP processing procedure

Fig 14: Image from front view (C3)

Fig 15: Landmark processing

Fig 16: Measure distance
3.4 Dental Cast study model
For a complete craniofacial database (i.e. hard and soft tissue) additional data such as dental cast are used. Several landmarks have been identified and measured to define dental characteristic of human skull. Dental casts (Figure 17) are prepared by School of Dental, Universiti Sains Malaysia. Figure 18 shows the procedure of processing dental cast using RAPIDFORM. Dental casts are scanned in pairs (upper and lower parts) using rotating table in order to get full coverage (360 degree) of dental model using PET and RAPIDFORM (Figure 19). Measurement data are then converted into STL binary format.


Fig 17: Dental cast and scanning result

Fig 18: Dental cast processing procedure

Fig 19: Dental cast processing procedure using Rapid Form
4.0 Final Processing and database (modeling and measurement)
The precision of measurement from all sensors (CT scan, laser scanning, stereo photogrammetric and dental cast) are within mm level. All results from these sensors are stored directly into the database system (Figure 20). The system has the capability to perform 3 tasks, i.e. database, generating 3D model and precise measurement (Deni Suwardhi, 2005).


Fig 20: Measuring in database
5.0 Conclusion
This paper focuses on the precise measurement of human faces and skulls for craniofacial reconstruction using specialized imaging system (i.e. laser scanning, photogrammetry, CT scan) and software (RAPIDFORM, DVD, 3DSLICER). The results show that the procedure is applicable for craniofacial reconstruction.

6.0 Acknowledgement
This research is part of a prioritized research IRPA vote 74537 sponsored by Ministry of Science, Technology & Innovation Environment (MOSTI) Malaysia.

7.0 References

• Balvinder Khambaya, Jean-Christophe Nebelb, Janet Bowmanb, Ashraf Ayouba, Frasier Walker & Donald Hadley (2002). A pilot study: 3D stereo photogrammetric image superimposition on to 3D CT scan images - the future of orthognathic surgery Appeared in: The International Journal of Adult Orthodontics & Orthognathic Surgery, Winter 2002, Volume 17, Issue 4, p331-341,Publisher: Ablex Publishing Corporation, New Jersey, U.S.A.
• Deni Suwardhi, Halim Setan, Albert Chong, Zulkepli Majid, Anuar Ahmad, Intan Sajidah & Mohd. Farid (2005). Coordinate systems integration for craniofacial spatial database from multimodal devices. Malaysian Science & Technology Congress (MSTC) 2005 16-18 April, Cititel Mid-Valley, Kuala Lumpur.
• Deni Suwardhi (2005). Development of Data Structure to Enhance Craniofacial Morphometric Database Management System. On going PhD research. Universiti Teknologi Malaysia.
• Deni Suwardhi, Intan Sajidah Abd. Aziz, Halim Setan, Zulkepli Majid, Anuar Ahmad & Albert Chong (2004). An innovative information system to support a nation-wide database for normal and malformation craniofacials. Presented at the 3D Modelling 2004, Paris, 28-29 April, 2004.
• Halim Setan & Mohd Sharuddin Ibrahim (2004). Close range measurement and 3D modeling. Presented at the 1st International Symposium on Engineering Surveys for Construction Works and Structural Engineering, Nottingham (United Kingdom), 28 June-1 July 2004.
• Halim Setan, Mohd Sharuddin Ibrahim & Zulkepli Majid (2005). Precise Measurement and 3D Modeling for Medical and Industrial Applications: Verification Tests. Presented at the FIG Working Week 2005 and GSDI-8, Cairo, Egypt April 16-21, 2005.
• Halim Setan, Zulkepli Majid & Deni Suwardi (2004). The Development of Image Capturing System and Information System for Craniofacial Reconstruction. Presented at the 3rd FIG Regional Conference Jakarta, Indonesia, October 3-7, 2004.
• Heinrich Schewe & Falk Ifert (2000). Soft Tissue Analysis and Cast Measurement in Orthodontics Using Digital Photogrammetry, IAPRS, Vol. XXXIII, Amsterdam, 2000.
• Zulkepli Majid, Halim Setan & Albert Chong (2004). 3D modeling of human face with non-contact three dimensional digitizer. Presented at the International Symposium and Exhibition on Geoinformation 2004, Kuala Lumpur, 21-23 September, 2004.
• Zulkepli Majid, Halim Setan, Albert Chong & Anuar Ahmad (2004). Modeling human faces with non-contact three dimensional digitizer. Presented at the 3D Modelling 2004, Paris, 28-29 April, 2004.