Ultrasound crack detection in a simulated human tooth

(005) 34, 80 85 q 005 The British Institute of Radiology http://dmfr.birjournals.org RESEARCH Ultrasound crack detection in a simulated human tooth MO Culjat 1, RS Singh 1, ER Brown, RR Neurgaonkar 3, DC Yoon 4 and SN White*,4 1 UCLA Henry Samueli School of Engineering and Applied Science, Los Angeles, California, USA; UCSB Department of Electrical and Computer Engineering, Santa Barbara, California, USA; 3 Rockwell Scientific Company LLC, Thousand Oaks, California, USA; 4 UCLA School of Dentistry, Los Angeles, California, USA Introduction Objective: Currently, diagnosis of cracked teeth generally depends upon the overall clinical assessment, or on exclusion of other clinical possibilities, not primarily on the direct identification of cracks themselves. Owing to its short wavelength in hard tissues and associated high resolution, ultrasound has the potential to allow detection of cracks within tooth structure. However, ultrasound detection of dental cracks has not previously been achieved. The purpose was to determine if an ultrasound imaging system was capable of imaging cracks in simulated tooth structure. Methods: A complete ultrasound system including a novel transducer made of PLZT-98, a novel gallium-indium alloy coupling agent, and customized electronic and digital signal processing (DSP) algorithms was developed for the specific application of optimizing crack detection within teeth. A simulated tooth with a known and uniform internal structure and acoustic properties similar to those of natural enamel and dentin was designed to model a human tooth with a crack located in dentin deep to the dentino enamel junction (DEJ). The distance between the DEJ and a crack of the simulated tooth were calculated. Results: The system unequivocally distinguished between areas with and without a simulated crack. Conclusion: A unique ultrasound dental crack detection system using a novel transducer; a novel coupling agent; and customized electronic and digital signal processing (DSP) algorithms has been validated in a simulated tooth. (005) 34, 80 85. doi: 10.159/dmfr/1901010 Keywords: tooth, ultrasonography, diagnostic imaging, cracked tooth syndrome The detection and diagnosis of fractures in teeth are vexing and difficult clinical problems. 1 Currently, dentists use patient history, visual examination, and a comprehensive endodontic examination to diagnose cracked teeth. However, clinical signs and symptoms are highly variable and are often insufficient to reach an unequivocal diagnosis. Cracks often cannot be visualized from the external surface of a tooth. Furthermore, cracks that are visible on the surface enamel often stop at the dentino enamel junction (DEJ) and are of no clinical consequence. Sometimes invasive procedures such as raising a mucoperiosteal flap or creating an endodontic access cavity, in combination with transillumination, staining, or microscopy, are used to visualize cracks, but some cracks still may not be identified. Differentiating between a shallow inconsequential crack and a through and through fracture is often *Correspondence to: Shane N White, UCLA CHS 3-087, Los Angeles CA 90095-1668, USA; E-mail: snwhite@ucla.edu Received 16 September 004; revised 6 January 005; accepted 6 January 005 extremely problematic. Although presence of a visible fracture line in enamel and a high ratio of restoration to total natural crown volume are associated with increased incidence of tooth fracture, this finding applies to a patient population, not to the diagnosis of cracks in each individual patient. Dental radiographs or microtomographs do not usually show cracks themselves, only their subsequent bony damage after a crack has eventually been colonized by bacteria and become a source of inflammation or infection. 3,4 Optical coherence tomography and electric conductance methods are poorly suited to deep crack detection. The development of methods for the assessment of cracked teeth and root fractures has been specifically identified as a top research priority by the American Association of Endodontists Foundation. 5 Owing to its short wavelength in hard tissues and associated high resolution, ultrasound has the potential to complement conventional radiography as an imaging technique in clinical dentistry. Ultrasound has the ability to penetrate hard structures and can, in principle, detect

Ultrasound crack detection MO Culjat et al 81 hard tissue discontinuities, or pathoses, under existing radiopaque restorations, a task that is difficult for conventional radiography. In addition, ultrasound is highly effective in detecting physical discontinuities such as fractures or cracks, even if such discontinuities are smaller than the acoustic wavelength. Therefore, ultrasound may provide a significant benefit to patients by allowing early detection of dental pathology, especially cracks. Additionally, ultrasound lacks the hazards of ionizing radiation. Ultrasound imaging has been investigated as far back as the 1960s as a possible diagnostic dental tool. Enamel and dentin thickness were measured on extracted teeth using lead zirconate titanate (PZT) piezoelectric transducers with oil coupling. 6,7 Eventually, similar results were achieved by coupling PZT transducers to extracted teeth using water, glycerine, and even mercury. 7 10 A, B, and rough C scan images of extracted teeth or parts of extracted teeth have also been made. 8,10 All of these studies provided useful information about the interaction of ultrasound with teeth and the thickness of dentin and enamel along specific trajectories. However the studies did not accurately scan the DEJ over large sections of a tooth and were unable to locate cracks or caries within teeth. Furthermore, the measured trajectories could not be considered representative of an entire tooth since the thickness of both enamel and dentin vary substantially within teeth. Recently, a 10 MHz PZT hydrophone transducer was used to perform an accurate circumferential scan around an extracted human molar in water. 11 That study demonstrated the feasibility of imaging the DEJ with ultrasound; but, like prior studies, it underscored the need for improvements in transducer technology, coupling efficiency, and signal processing to image deeper features like cracks or the pulp. The purpose of this paper was to determine if an improved ultrasound imaging system, including a novel method of acoustic coupling, a superior transducer, and improved electronic signal processing, was capable of imaging microcracks in a simulated tooth. Materials and methods System description A complete ultrasound system including a novel transducer; a novel coupling agent; and customized electronic and digital signal processing (DSP) algorithms was developed for the specific application of optimizing crack detection within teeth. The system was optimized by using a customized Mason Model program in MATLAB (Mathworks, Natick, MA) and PZFlex (Weidlinger Associates, Los Altos, CA) to simulate one and two dimensional acoustic wave patterns, and by using Advanced Design System software (ADS, Agilent Technologies, San Jose, CA) to simulate the transmit and receive electronics. The resulting system had a frequency of operation centred at 19 MHz and an instantaneous bandwidth of 7 MHz. The high frequency allows for the detection of fine features due to its associated short acoustic wavelength, and the modest bandwidth allows for narrow acoustic pulses in the time domain, thus both improve the axial resolution. PLZT-98 transducer The ultrasound piezoelectric transducer was specifically designed to acoustically impedence-match to human enamel, thus allowing for maximum ultrasound transfer into the tooth. Piezoelectric materials have the special property of being able to convert electrical energy into mechanical energy in the form of acoustic waves. The mechanical energy is then used to produce acoustical waves. Lead zirconate titanate (PZT) transducers have been widely used owing to their high efficiency in converting energy between the two domains. Recently, a new dense ceramic piezoelectric material with further improved energy conversion efficiency, PLZT-98, was developed by the Rockwell Scientific Company (Thousand Oaks, CA) by the addition of lanthanum to PZT. 1 The PLZT-98 material has a piezoelectric strain coefficient, d 33, of 850 pc/n, a strain percentage of 0.7, an electromechanical coupling coefficient, k 33, of,0.8, and a relative dielectric constant of 3500. These properties are markedly superior to those of commercially available PZT materials. The transducer was made by thinning down a piece of PLZT to 130 mm, dicing it into 175 mm squares and then evaporating platinum and gold electrodes on to the top and bottom. The dimensions of these piezoelectric capacitor elements were dictated by a trade-off between acoustic power generation, beam width and associated electrical properties. For example, a larger element is capable of more acoustic power output, but also results in a larger beam and in decreased lateral resolution. The resonance of the transducer element is primarily determined by the thickness of the device, such that the resonant frequency increases with decreasing thickness. Our transducer s 130 mm thickness resulted in a 19 MHz resonant frequency. For each transducer, a single 175 mm square active element was attached to the top of a thick glass substrate and electrically grounded to an aluminium casing with conductive silver paint. The back of the transducer material was electrically connected to an SMA connector with a 1 mm diameter gold wire. The thick glass substrate was chosen to provide an acoustic delay line that allows sufficient time for acoustic ringing from the applied electrical pulse to subside. The glass also serves as a stable base for the active element and provides a close acoustic impedance match to enamel, thus allowing for good power transmission from the piezoelectric transducer to the tooth. Gallium-indium alloy coupling agent A novel coupling agent, gallium-indium alloy, was used to couple the transducer to the tooth. This eutectic metal alloy is liquid at room temperature, and has near ideal acoustic properties, including low acoustic loss and appropriate impedance match, for this application, providing an excellent acoustic medium to transfer acoustic energy between the transducer and enamel (Table 1). Galliumindium alloys have been proposed for use as alternatives to mercury in dental amalgam owing to their superior biocompatibility, but have not gained acceptance because of their poor stability. 13 In this experiment, a drop of alloy

8 Ultrasound crack detection MO Culjat et al Table 1 Acoustic properties of natural tooth, simulated tooth, coupling agent, air and water. 19 1 Acoustic velocities and acoustic impedances determined in our laboratory are indicated by an asterisk Material Density (kg m 3 ) Acoustic velocity (m s 1 ) Acoustic impedance MRayls Air 1. 330 0.0004 Water 1000 1480 1.48 Soft tissue, average 1060 1540 1.6 Enamel 900 5700p 16.5p Glass, soda-lime 40 5600p 13.4p Dentin 100 3800p 8.0p Dental composite 090 3600p 8.0p Gallium-indium alloy 6500 740p 17.8p was placed between the transducer and the tooth. For clinical application, it is suggested that the alloy be isolated within an acoustically thin impermeable parylene membrane. Electronics and digital signal processing The electronics in the system can be viewed as having three main components: the transmit electronics; the receive electronics; and the digital signal processing (DSP) (Figure 1). Known techniques from the fields of radar, sonar, and communications were applied to maximize the sensitivity of the system to the weak acoustic echo returns from cracks in the strongly reverberative environment of a tooth. 14 16 The transmitted waveform was a time-gated 19 MHz sinusoidal carrier. The gating was accomplished with a solid-state switch driven by a pulse generator. The electrical waveform was applied to the transducer via a transmit/receive (T/R) switch that isolated the transmit electronics and the receive electronics. The amplitude of the carrier was kept low enough to maintain the transducer in its linear transfer regimen. The transducer then produced an acoustic waveform nearly identical to the electrical waveform, and transmitted acoustical power into the sample through the gallium-indium alloy. Discontinuities within the simulated tooth, i.e. cracks and the DEJ, produced reflected waves, which propagated back through the alloy and into the transducer. During propagation the T/R switch was activated to connect the transducer to the receive electronics. A lownoise amplifier and superheterodyne circuit performed the first stage of the signal processing in the analogue domain, to frequency-down-convert the echo to baseband. In-phase (I) and quadrature (Q) mixers were in the heart of the superheterodyne receiver. The I and the Q mixer output signals were fed into a digital oscilloscope that was used both for pulse-echo visualization and as an analogue-to-digital (A/D) converter. The digitized Figure 1 Electronics block diagram of the ultrasound crack detection system

Ultrasound crack detection MO Culjat et al 83 waveform was post-processed using MATLABe running in a Windowse 000 (Microsoft, Redmond, WA) environment. In the DSP portion of system, the I signal and the Q signal were squared and added together to obtain the envelope of the received echoes. All white Gaussian noise (AWGN) from the electronics and false echoes from clutter and reverberation within the simulated tooth was included in this signal. The waveform of the signal was then crosscorrelated against a previously saved echo envelope from a glass surface. This last operation emphasized the desired components of the received signal, the echo envelope, and de-emphasized the noise and clutter. 17,18 The output of the system was a cross-correlation waveform with respect to the delay time, t, in which the peaks correspond to reflections of the transmitted ultrasound pulse off discontinuities encountered along the acoustic path. This was identical in functionality to a traditional A-scan ultrasound transceiver. Tooth simulation To validate the crack detection capabilities of the system, a simulated tooth with a known and uniform internal structure was designed to model a human tooth with a water-filled crack located in dentin deep to the DEJ. The simulated tooth was designed to have acoustic properties similar to those of natural enamel and dentin. The most critical acoustic properties are density, compressional wave 19 1 velocity, acoustic impedance, and loss (Table 1). Dental composite has slightly more acoustic loss than dentin, thus making the model slightly more demanding to image than natural tooth structure. Acoustic loss is significant in high frequency ultrasound because waves get attenuated much more rapidly as frequency increases. It is important to note that attenuation also increases exponentially with distance. Acoustic impedance, which is approximately the product of the density and compressive wave velocity, is the acoustic analogue to the index of refraction in optics. Thus the difference in acoustic impedance of two adjacent materials gives an idea of how much energy will be transmitted through the interface and how much will be reflected by the interface. Thus, the gallium-indium alloy serves as a better coupling material than water between the glass front of the transducer and enamel. In the simulated tooth, soda-lime glass was used to represent enamel and a resin-based dental restorative composite (Build-It; Pentron Corporation, Wallingford, CT) was used to mimic dentin (Figure ). Both are similar in their acoustic properties to their natural counterparts (Table 1). A 5 mm wide crack was introduced into the composite during the curing process using a hardened steel feeler gauge. Dental radiographs cannot visualize such a crack unless it would happen to fall in a plane perfectly parallel to the X-ray beam, which is a most unlikely scenario. The 1 mm thick soda-lime glass (enamel) layer was sandblasted gently before it was attached to the composite (dentin) using a thin layer of resin composite cement of similar properties to the bulk resin composite (Variolink II, Ivoclar Vivadent, Schaan, Liechtenstein). Figure Placement of ultrasound transducer to scan areas with and without cracks in a simulated tooth. Not to scale Imaging procedure To ensure accurate positioning for imaging, the simulated tooth was mounted on goniometer attached to an XYZ translation stage, which allowed for two axes of rotation and three axes of translation. The ultrasound transducer was mounted on a two-axis gimbal. When imaging, the simulated tooth was brought into contact with the galliumindium alloy couplant and 19 MHz pulses were sent from the transducer. The translation and rotation micro-positioners were used to maximize the echo return from each of the interfaces in the simulated tooth. The amplitude of each reflected echo depends on the acoustic impedance difference at the interface of interest as well as the scattering of the waves at that point in the simulated tooth. Once the signal was maximized, the data were recorded and saved for processing as described above. The range, or distance between each interface and the transducer, was calculated by the relation: R ¼ cdt ð1þ where R is the range, c is the speed of sound (compressional waves) in a given material, Dt/ is the single pass acoustic time of flight, Dt being the acoustic round trip time. In the correlation output of the receiver, Dt is measured between the feature of interest and a reference, such as the DEJ. This measurement is equivalent to Dt with respect to the reference, and substitution of Dt into the above range equation gives the relative distance between the feature of interest and the reference point. The roundtrip time was measured on an oscilloscope. The speeds of compressive waves were measured acoustically on a block of each material of known dimension.

84 Ultrasound crack detection MO Culjat et al Results Numerous repeated measurements were made by moving the simulated tooth laterally in relation to the transducer, scanning both its crack and crack-free sections. When the transducer was facing the surface of the phantom where there was no crack, the only echoes came from the couplant/glass (tooth surface) and the glass/composite (DEJ) interfaces. In each case the DEJ echo arrived, 360 ns after that of the alloy/glass echo, corresponding to the 1 mm thickness of the glass. The calculated range difference, Dx, with Dt corresponding to the time delay between each of the two echoes is: Dx ¼ cdt ¼ 5600 m s1 360 ns < 1mm ðþ When the sections of the simulated tooth with the crack were scanned laterally, another echo was visible following the DEJ echo by, 670 ns. Assuming the speed of compressional waves in dental composite to be 3600 m s 1, and performing the same calculation as in (), then: Dx ¼ cdt ¼ 3600 m s1 670 ns < 1: mm ð3þ Typical matched-filtered output data plots are shown in Figure 3. The upper plot shows the recording point over the crack and the bottom shows the point over the phantom with no crack present. In both cases, the DEJ echo is, 360 ns from that of the couplant/glass interface echo, and in the upper plot the crack echo follows the DEJ by,670 ns. The presence of a crack echo is clear in the first image, while there is no sign of a crack in the second image. The signal-to-noise ratio (SNR) in regards to all white Gaussian noise (AWGN) of the system was 38 db. Discussion The detection of the crack in a simulated tooth shows that ultrasound has the potential to provide clinically relevant information, complementing current diagnostic procedures. Cracks cannot usually be directly identified using diagnostic evaluations or X-rays. The A-scans (Figure 3) very clearly differentiate between the cracked and un-cracked portions of the same simulated tooth. Additionally, the high SNR makes interpretation unequivocal. The calculated depth of the crack in simulated dentin was, 1. mm, not the 1 mm expected. Several possibilities may account for this discrepancy. In this simulated tooth, the DEJ was only within a few degrees of being parallel to the tooth surface, thus the echo occurred at an angle, taking a longer return path and therefore increasing the round-trip time. Additionally, the speed of compressive waves in dental composite varies upon curing conditions, which can result in an inhomogeneous specimen. Another source of error is that the cement layer, while thin and very similar to the bulk dental composite, was not accounted for acoustically. Also, the depth of the simulated crack may not have been uniform. Nevertheless, crack detection was undeniable and repeatable on this and other simulated teeth. It is also important to note that when scanning teeth with ultrasound, the relative locations of landmarks within teeth are as clinically useful as the exact locations of each Figure 3 Processed ultrasound images of water-filled crack and un-cracked simulated tooth

Ultrasound crack detection MO Culjat et al 85 interface. It is proposed that this technology will be used to verify the presence of suspected cracks or caries in teeth, so the primary requirement is that an ultrasound system gives a positive or negative diagnosis. Therefore, clinical crack diagnosis could be by inclusion, rather than by exclusion of other clinical causes, as is current practice. This ultrasound dental crack detection system using a novel transducer, a novel coupling agent, and customized electronic and digital signal processing (DSP) algorithms has been validated in a simulated tooth. The high signalto-noise may permit the imaging of more complex situations and future automation. It is important to note that natural teeth usually have curved outer surfaces, DEJs, pulpal walls, and oblique or curved micro-anatomical features. Although many cracks are planar, they too can be curved. However, crack reflections from a single transducer are extremely angle dependent, and the findings of this study are limited to planar cracks and interfaces oriented perpendicularly to the transducer. Therefore, in order to achieve clinical application, future goals are to demonstrate validity in more geometrically complex situations, in narrower cracks, in real human teeth, in locating caries, in locating pulp chambers and in locating caries and interfaces hidden behind dental restorations. Our long-term strategy to develop phased arrays will give more positional data, decreased dependence on sensor angulation, and increased power on target to achieve broader clinical application. Our unequivocal data and high SNR suggest that these goals may eventually be achieved. Thus, a novel dental imaging system, capable of imaging deep within tooth structure, and without ionizing radiation, may be created. Acknowledgment The authors most gratefully appreciate funding provided for this work by the National Science Foundation under award number ECS-9980875 and by the National Institutes for Health / National Institute for Dental and Craniofacial Research under award number DE 14189. References 1. Cohen S, Blanco L, Berman L. Vertical root fractures: clinical and radiographic diagnosis. J Am Dent Assoc 003; 134: 434 441.. Bader JD, Shugars DA, Martin JA. Risk indicators for posterior tooth fracture. J Am Dent Assoc 004; 135: 883 89. 3. Dowker S, Davis G, Elliott J. Microtomography: nondestructive three-dimensional imaging for in vitro endodontic studies. Oral Surg Oral Med Oral Path Oral Radiol Endod 1997; 83: 510 516. 4. White SC, Pharoah MJ. Oral radiology, principles and interpretation (5th edn). St. Louis, MO: Mosby, 004: p 63. 5. American Association of Endodontists Foundation. Endodontic research guidelines. Chicago, IL: American Association of Endodontists, 004. 6. Lees S, Barber FE. Looking into teeth with ultrasound. Science 1968; 161: 477 478. 7. Barber FE, Lees S, Lobene RR. Ultrasonic pulse-echo measurements in teeth. Arch Oral Biol 1969; 14: 745 760. 8. Ng S, Payne P, Ferguson M. Ultrasonic imaging of experimentally induced tooth decay. International Conference on Acoustic Sensing and Imaging. London: Institute of Electrical and Electronics Engineers, 1993: pp 8 86. 9. Huysmans M, Thijssen J. Ultrasonic measurement of enamel thickness: a tool for monitoring dental erosion? J Dent 000; 8: 187 191. 10. Ghorayeb S, Valle T. Experimental evaluation of human teeth using noninvasive ultrasound: echodentography. IEEE Trans Ultrason Ferroelectr Freq Control 00; 49: 1437 1443. 11. Culjat M, Singh R, Yoon D, Brown E. Imaging of human tooth enamel using ultrasound. IEEE Trans Med Imaging 003; : 56 59. 1. Rockwell Scientific Company. Piezoelectric motors and actuators. Thousand Oaks, CA: Rockwell Scientific Company, 004. 13. McComb D. Gallium restorative materials. J Canad Dent Assoc 1998; 64: 645 647. 14. Skolnik MI. Introduction to radar systems (3rd edn). New York, NY: McGraw-Hill, 00. 15. Burdic WS. Underwater acoustic system analysis. New Jersey, NJ: Prentice-Hall, 1984. 16. Gibson JD. Principles of digital and analog communications. New York, NY: Macmillan Publishing Company, 1989. 17. Di Franco JW, Rubin WL. Radar detection. Norwood, MA: Artech House Incorporated, 1980. 18. McDonough RN, Whalen AD. Detection of signals in noise. New York, NY: Academic Press, 1995. 19. Hedrick W, Hykes D, Starchman D. Ultrasound physics and instrumentation (3rd edn). St Louis, MO: Mosby, 1995: p 7. 0. O Brien W. Dental materials and their selection (3rd edn). Chicago, IL: Quintessence, 00. 1. Onda Corporation, Acoustic material table. Sunnyvale: Onda Corporation, 004.