Imaging Capability at Close Range of High Frequency Ultrasound Transducer on the Perforation of Bony Walls Simulating Pedicle Screw Placement


 Background: Ultrasound has been proved to be a promising alternative spine navigation method. High frequency ultrasound transducer has the advantage of high resolution on surface structure, but imaging at close range is difficult, especially in narrow space of the pilot-hole in pedicle.Methods: Twenty cortical bone chips were made and different size of hole with diameter of 1mm, 2mm, 3mm or 5mm was randomly carved in each bone chip. A tailored 30MHz high frequency transducer scanned bone samples at the distance of 4mm, 3mm, 2mm and 1mm. Successive transmission ringingeffect elimination, Hilbert transform and Gray-scale mapping method were utilized to process and optimize attained original images.Results: At the distance of 4mm, 3mm, 2mm and 1mm, the holes with diameter of 5mm, 3mm and 2mm could be discerned. At the distance of 1mm, only the holes with 5mm and 3mm could be clearly distinguished and the 2mm hole appeared obscure. The holes with diameter of 1mm could not be detected at any distance. The holes with diameter of 2mm were able to be detected at the distance of 1mm.Conclusions: This study indicated that the high frequency transducer had limited imaging capability at close range on the bony surface. These results lay a foundation for further developing a novel ultrasound-based spinal pedicle interior imaging and navigation system.


Background
Pedicle screw has been routinely used in posterior spinal fixation for its advantage of rigid 3-column stability [1]. Traditional free-hand pedicle screw placement technique mainly relies on clear exposure of anatomical landmarks and clinical experience. The accuracy of screw placement cannot be guaranteed by using fluoroscopy assistance, especially in spinal deformity, cervical and thoracic spine and mini-invasive spinal surgery. Pedicle screws misplacement could lead to severe iatrogenic neurological, vascular and visceral complications [2].
To improve the accuracy of pedicle screw placement, various image-based navigation (including fluoro-based and computed tomography based navigation) and robotic assistance navigation have been developed [1], [3]. Although such auxiliary equipments have been reported to reduce the rate of screw misplacement, there are some obvious disadvantages: intra-operative ionizing radiation exposure, large space occupied, difficulties in registration and tracking, and high cost [4][5][6]. Additionally， some studies showed such technologies did not demonstrate significant superiority over conventional technique [7,8].
Due to the advantages of real-time imaging, free of ionizing radiation and portability, some attempts in developing ultrasound-assisted pedicle screw placement have been reported [5], [9][10][11][12][13] and demonstrated promising potential. Mujagi´ et al [14] investigated the acoustic properties of human vertebral cancellous bone specimens using low frequency transducers (1MHz and 3.5MHz). Based on this result, Aly et al [12] obtained B-mode images by rotating 3.2MHz transducer from within the boreholes in the pedicles, and the trabecular and cortical bone could be detected, which suggested developing an ultrasonic device to guide pedicle screw placement was feasible. Penetration depth and resolution are two important parameters in ultrasound imaging. Although the above-mentioned low frequency ultrasound beam can penetrate into the trabecular bone and imaging the surrounding cortical wall, the inner surface of the bore holes cannot be visualized. In view of the high resolution in the imaging of intravascular lesions, Kantelhardt et al [9,10] firstly reported the intrapedicular imaging using 20MHz intravascular ultrasound (IVUS) transducer. At the site of experimental pedicle perforation, ultrasonography showed a focal loss of the circular shape of the highly echoic bone interface, which suggested high frequency transducer had the potential to detect the breach and corresponding site in the walls of bore hole during pedicle screw placement procedure.
To accommodate the imaging needs of smaller pedicles (cervical and thoracic spine, deformity, children, dysplasia, et al), we have persisted in pursuing smaller size of higher frequency transducer, better image quality (higher resolution) and shorter near-field than above-mentioned 20MHz counterpart. We have fabricated a novel 30MHz transducer that meets the above requirements. In this study, we utilized the novel transducer to detect different sizes of holes within cortical bone samples from different distances, aiming to investigate the limit close-range imaging capability.
This study simulated a scenario in which high frequency transducers detected the perforations in outer cortex walls from within pilot holes in pedicles.

Results
Twenty cortical bone chips were cut from the same part of the ilium of cattles, which was smooth and large enough. In the center of each bone chip, one round hole with diameter of 1mm, 2mm, 3mm or 5mm was randomly selected and carved (five chips for every same diameter size) (Fig. 1b, c, d, e).
All the experiments were performed in the 3D scanning system Ultrasound measurements system III (UMS III, Precision Acoustics Ltd., Dorchester, UK) ( Fig.   2). Using ultraviolet (UV) adhesive, the bone chips were fixed on, Acrylonitrile-butadiene-styrene (ABS) rod, which is a weak ultrasonic reflector. This fixation method can supply sufficient stability and minimize artifacts (Fig. 3). In the process of scanning, the specimens were placed parallel to the 30MHz transducer to get the maximum amplitude of the signals. The distance between the transducer and each bone specimen was successively set at 4mm, 3mm, 2mm and 1mm. Every scanning covered the entire surface of each bone specimen to measure the size and detect the relative location of the hole within.UMS III acquired the original sound field images by the square of the average Voltage Peak-Peak (VPP) of each point.
At the distance of 4mm, 3mm and 2mm from the transducer to the bone surface, the holes with diameter of 5mm, 3mm and 2mm could be discerned. The holes with diameter of 5mm and 3mm demonstrated circular shape at each testing distance.
Although the hole with diameter of 2mm at each testing distance could be distinguished, they exhibited distorted images, which were elongated in one direction.
At the distance of 1mm, only the holes with diameter of 5mm and 3mm could be clearly distinguished, while the hole with diameter of 2mm appeared obscure in the image. The hole with diameter of 1mm could not be detected at any distance (Fig. 4a).

The elimination method of transmission ringingeffect signal
In normal condition, the echo signal is far from the transmission ringingeffect signal ( Fig. 5b), which demonstrates clear images. When the distance between the transducer and bone surface is gradually reduced, the echo signal is superimposed until to be completely submerged in the transmission ringingeffect signal (Fig. 5a), and the 6 quality of the images turned to be poor. Therefore, we further processed as follows in order to get clearer images.
As shown in Fig. 5c, the transmission ringingeffect signal lasted the first 2μs, within which this signal interfered with the distance of the pulse echo signal about 1.5mm (the speed of ultrasound in water is 1540m/s). Since the transmission ringingeffect signal was stable and can be acquired, we processed the images using transmission ringingeffect elimination method to discern the bone samples' genuine profile at close distance (Fig. 5d).
After eliminating the transmission ringingeffect signal, the holes with diameter of 5mm, 3mm and 2mm were exhibited clearer on sound field images than before.
However, the hole with diameter of 1mm still could not be distinguished (Fig. 4b).

B-mode imaging and optimization
For enabling to employ the mature image processing methods, the original sound field images were transformed into B-mode images. We processed all the above attained images with Hilbert transform method to achieve B-mode images of the bone samples.
Then a logarithmic function based gray-scale mapping method was used to further enhance the contrast of the images, solving the problem that the strong reflection points on the bone chips and the background focus on the similar gray level. The results revealed that the images of the holes with diameter of 5mm, 3mm and 2mm became clearer than before at all testing distances. However, the holes with diameter of 1mm still could not be detected at any testing distance (Fig. 6).

Aperture measurements
According to Fig. 7  For the holes with diameters of 5mm, 3mm and 2mm, our study showed that maximum apertures measured by transducer from four different distances differed significantly with the real diameters. Although the cause is unclear, it is speculated that the measured value of hole aperture is affected by the scanning sequence of the transducer. The corresponding ultrasonic measured value should plus 1/2 of the lateral size of transducer at each side respectively. The final correctional measured values have no statistical differences with the actual apertures (P>0.05). In addition, the correctional measured values of holes with diameter of 5mm, 3mm and 2mm have no statistical difference between four different testing distances (4mm, 3mm, 2mm, 1mm) (P>0.05) ( Table 1). And the correctional measured values detected from the same testing distance have no significant difference between different specimens with the same size holes within.

Table 1 The mean correctional measured values of holes apertures from different testing distances
The hole with diameter of 1mm could not be detected at any testing distance.

Discussion
In terms of transducer, the attenuation increases with the higher transmitted frequency, causing the returned signals to be flooded within the dark current noise. In addition, the bone tissue have much higher acoustic impedance than soft tissue, vasculature and other visceral tissue, which prevents the high frequency (>20MHz) ultrasound transducer from penetrating and imaging of the trabecular and cortical bone. But it is well known that the higher frequency can get higher image quality(high resolution), shorter near-field region and smaller transducer size [15]. In selected cases, it could even visualize fissure-like fractures in cadaveric bone [10]. Theoretically the resolution of our new-fabricated 30MHz transducer in this study can be increased by 50% and the size decreased by about 50% compared with 20MHz counterpart [16].
Such advantage of high resolution on structure information of borehole surface can be combined and integrated with low frequency transducer which has a capability of adequate penetration depth, developing into a new method/system based on the dual frequency transducer.
In our study, the 30MHz transducer was designed to be a cuboid with transverse surface of 1mm×1mm and side emitting surface of 3mm×1mm, which could accommodate the narrow pilot hole in pedicle screw placement procedure. Since the emitting surface on the side approaches was quite close to the bone wall, the effect of near-field would be encountered inevitably. Although the direct mathematics formula was hard to work out, we applied finite element analysis to derive the whole near-field length of the 30MHz sensor as 2.5mm, in which the chaos length is 1mm. It can explain the phenomena that the imaging at distance of 1mm cannot be discerned.
Developing a more miniaturized transducer may simultaneously enhance the capability of imaging within shorter near-field area.
After mechanical scanning on the holes，a band-pass filter based on transducer frequency bandwidth was applied to the ultrasonic echo signal to achieve higher signal-to-noise ratio. After that, the elimination method of transmission ringingeffect signal was adopted for the raw data in which the echo signal and near-field signal overlap. We subsequently processed the filtered signal by Hilbert transform and a specific gray-scale mapping [17,18], and B-mode images of the holes were finally achieved. Due to the strong ultrasonic reflection at the bone/water interface [13], we proposed a new gray-scale mapping method based on the logarithmic function in this study. Combined with statistical analysis, we determined the proper threshold of gray-scale mapping curve to finally obtain high-quality B-mode images of bone surfaces.
Kantelhardt et al [10] reported that the smallest diameter of the breach identified by intraosseousultrasonography was 4mm by 20MHz intravascular ultrasound transducer. In this study, we were able to detect a 2mm-sized hole at the distance of 1mm from the bone surface, which met the clinical need for pedicle perforation detection effectively. The breach less than 2mm means that the tip of the drill just touches or penetrates the cortical wall of the pilot-hole in pedicle, which is considered within the "safe zone" [2,19], and to have a low risk of neural and vascular damage.
In addition, the diameter of the nutrient artery entrance to vertebra is about 2mm [20].
Therefore, in the process of ultrasound exploration, utilizing the holes with diameter of less than 2mm as criteria of wall perforation may have false positive. The criteria of holes with diameter of 2mm or greater can discern the breach of pedicle wall effectively and avoid false positivity caused by nutrient blood vessel.
There are two reasons as to why smaller holes in the bone chips could not be clearly exhibited. One possible reason is the heterogeneous structure of the bone tissue, which results in a wide range of pulse-echo signal intensity in the different part of the same bone sample. Another possible cause is the ratio of the dimension of the transducer to the hole. If the size of transducer is much smaller than the hole, the transducer can be assumed to be a point. However, when the size of the transducer is close to that of the hole, the size of the transducer should be considered. Some of the emitted ultrasonic beam will hit on the bony surface of the sample and some other will enter into the hole, by which mixed pulse echo signals will generate and inevitably obscure the boundary of the holes.

Conclusion
Our study provides a basis for the ability of high-frequency ultrasound transducers (30 MHz) to effectively image bone surfaces at close range for the first time. It also lays a foundation for further study to develop a new prototype system combining the 11 penetration capability of low frequency transducer with the surface imaging capability of high frequency transducer for pedicle interior imaging and navigation.

The fabrication and characteristics test of micro high-frequency transducer
We designed and fabricated a single-element ultrasonic transducer of 30MHz for this research, and the dimension is 3mm×1mm×1mm to adapt to the narrow space of pedicle bore-holes (Fig. 8b). The transducer is a classic three-tiered structure composed of matching layer, piezoelectric layer and backing layer as shown in Fig. 8a.
The matching layer is home-made conductive glue mixture. The piezoelectric material is PZT-5H coating with chromium and Au as electrode. The physical properties of each layer of the transducer are shown in the Table 2. The pulse-echo method was used to evaluate the transducer's performance. A simple test platform is composed of water tank, position adjustment system, DPR500 pulse transmission/reception system, oscilloscope and 5mm acrylic plate. The transducer was immersed in constant temperature water at 20℃ and fixed on the position adjustment system to ensure that the sound wave was perpendicular to the acrylic plate. Then DPR500 system (Imagine company, New York, USA) was used to drive the transducer. The amplitude of the driving pulse was 92.4V and the pulse width was 3.2ns. The receiving bandwidth and gain were set as 5-500MHz and 0dB, respectively. Received pulse signals were displayed and stored in the Tektronix DPO5034 oscilloscope (Tektronix, Inc., OR, USA), whose input impedance was set as 50Ω. The digitalizing bit depth was 12 and corresponding quantification accuracy was 0.002V. The Radio frequency (RF) curve is shown in the black line and its fast Fourier transform spectrum is shown in the red line. The performance of the transducer is displayed in Fig. 9. The center frequency is 29.0564MHz, -6dB bandwidth is 7.74116 MHz/26.4% and the amplitude is 2.232V. The resolution of the transducer is 200µｍ measuring by the wire target.

Specimen preparation
Chainsaw (PROXXON, MBS 240/E) was used to cut the bone chips (Fig. 1a). The size of each specimen was designed as length 3cm, width 1cm and thickness 2mm to emulate the thinned outer cortex of the pedicle. After the dimension was measured and assured with a caliper, an electric drill was used to carve one different size round hole in the center of each bone chip (Fig. 1b, c, d, e).

Measurement method
Although ultrasound B-scan mode is used in the actual clinical application, the C-scan mode was adopted in this study to simplify the experimental data. The bone 13 specimens were plane-scanned along X and Y axis by the transducer with a center frequency of 30MHz. The stepping of 0.3mm was set by preliminary experiment.
DPR500 of 500MHz channel was used to drive the transducer. The oscilloscope received the pulse echo signals and transmits them to the UMS-III system. Finally, the image was displayed on the screen. In order to ensure the millimeter-sized air bubbles did not exist in the water bath, the degassing and deionized water was poured into the water tank and all experiments were performed at room temperature. A band-pass filter with the range of 20MHz to 40MHz was used to reduce the noise of ultrasonic raw data and improve the signal-to-noise ratio. (Fig. 2)

The elimination method of transmission ringingeffect signal
The superimposed signal was attained by scanning bone specimens, and the transmission ringingeffect signal was got through scanning without the bone specimens under the same other parameters. The ultrasound pulse echo signal was achieved by subtracting the transmission ringingeffect signal from the original superimposed signal.

Envelope-detected with Hilbert transform
In this study, the method to detect the envelope of the ultrasound pulse echo signal is based on Hilbert transform [21], which applies to create the analytic expression of the scattered signals of the bone chips. These analytic signals are complex, in which the real part is the original signals and the imaginary part is the envelope of the original signals. The amplitude of this envelope-detected record was logarithmically compressed and processed nonlinearly so that a larger dynamic range of weak to strong echoes was presented in one same image. So the effective B-mode images of the bone chips were achieved from the above ultrasound pulse echo signals.

Gray-scale mapping method
Since the dynamic range of the pulse echo signals of the bone chips is much larger than the conventional B-mode images of biological soft tissue. A specific gray-scale mapping curve was applied in order to achieve images with proper contrast. The gray-scale mapping curve in this study is defined mainly based on the logarithmic function, which accommodates the strong ultrasonic reflection of the bone/water interface. The threshold range of the mapping curve was determined by statistical method according to the actual experimental testing results.

Statistical analyses
Date was analyzed with SPSS 22.0 statistical software (SPSS Inc., Chicago, USA).
After the size measurements on the same hole from different distances were tested to be normally distributed, the statistical analyses were performed using one-way ANOVA. The P value of less than 0.05 was defined as statistically significant. Further Dunnett's t test was used to compare the measured apertures from different distances with real aperture of the same hole, the diameter of which is 5mm, 3mm and 2mm.

Ethics approval and consent to participate
This work was approved by the ethics committees of Suzhou Municipal Hospital.

Consent for publication
All the authors of the paper approved the publication of the article.      are sequentially utilized to achieve B-mode images. The images of the holes with diameter of 5mm, 3mm and 2mm are clearer than before, and the hole with diameter of 1mm still cannot be detected at any distance.