Investigation of Acoustic Changes Resulting from Contrast Enhancement in Through-Transmission Ultrasonic Imaging

Ultrasound in Med. & Biol., Vol. 36, No. 9, pp. 1395–1404, 2010 Copyright Ó 2010 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/$ - see front matter

doi:10.1016/j.ultrasmedbio.2010.05.024

d

Original Contribution INVESTIGATION OF ACOUSTIC CHANGES RESULTING FROM CONTRAST ENHANCEMENT IN THROUGH-TRANSMISSION ULTRASONIC IMAGING TAMARA ROTHSTEIN,* DIANA GAITINI,y ZAHAVA GALLIMIDI,y and HAIM AZHARI* * Technion IIT, Faculty of Biomedical Engineering, Haifa, Israel; and y Rambam Medical Center, Department of Medical Imaging, Haifa, Israel (Received 23 May 2009; revised 9 May 2010; in final form 20 May 2010)

Abstract—Through-transmitted ultrasonic waves can be used for computed projection imaging of the breast. The goal of this research was to analyze the acoustic properties changes associated with the propagation of ultrasonic waves through media before and after ultrasound contrast agent (UCA) injection and to study the feasibility of a new imaging method combining projection imaging and UCA. Two transmission techniques were examined: Gaussian pulses and pulse inversion. In the latter, three different double inverted pulses were studied: double Gaussian, double square and double sine. A computerized automatic ultrasonic scanning system was used for imaging. To simulate blood vessels, a phantom, consisting of a latex tube through which saline was circulated, was assembled. The phantom was placed within the scanner and sets of acoustic projection images were acquired. Then, a suspension of the UCA DefinityÔ was added to the saline and a new set of images was obtained. The pre and postcontrast images were quantitatively compared in terms of amplitude and time-of-flight (TOF). In addition, nonlinearity was evaluated by comparing the relative alteration of the positive and negative parts of the signal. Statistically significant (p , 0.001) changes in the projection images resulting from the UCA injection were observed in wave amplitude (22% ± 13%), TOF (7.9 ns ± 6.3 ns) and nonlinear properties (35% ± 32% and 56% ± 17% for Gausian pulses and pulse inversion, respectively). One in vivo study of a female breast is also presented and its preliminary outcomes discussed. Together, these results indicate the technical feasibility of the suggested method and its potential to detect breast tumors. (E-mail:[email protected]) Ó 2010 World Federation for Ultrasound in Medicine & Biology. Key Words: Contrast enhanced ultrasound, Coded excitation, Through-transmission, Amplitude attenuation, Time-of-flight, Nonlinearity.

overcomes part of the above-mentioned limitations. Nevertheless, B-scan presents even more complex restrictions. First, geometry and distance estimates rely on the assumption of constant speed of sound (SOS) in all tissues; since this is obviously not true, an average SOS is used. Second, due to the small acoustic impedance differences between soft tissues, reflection coefficients are small, thus leading to a low signal to noise ratio (SNR) and creating the need for strong amplification and filtering of the signals. Last, since the attenuation and transmission coefficients on the echo’s path to the target tissue and back to the transducer are not known, the signals received from elements located deep inside the tissue do not allow for quantification of acoustic properties of the tissue (e.g., SOS and attenuation coefficients), information that may be crucial for tumor characterization (Azhari 2010). Imaging with through-transmission (TT) provides solutions for these three problems. The use of ultrasound contrast agents (UCA) for medical imaging was proposed by Gramiak and Shah

INTRODUCTION Ultrasound is an imaging modality widely used in medical diagnosis and in medical research in general. The most commonly used ultrasonic imaging method is the ‘‘B-scan’’, which is based on the pulse-echo technique. However, one of its drawbacks is the fact that B-scans are usually performed using handheld transducers, which implies that: (1) the scan cannot be done systematically, (2) the quality is operator-dependent and (3) identical cross-sections can hardly be reproduced. To overcome these limitations, in the field of breast imaging, for example, automated scanning B-scan systems have been suggested in the past and new modern systems have been recently introduced (Chiang et al. 2006; Shipley et al. 2005). Automation indeed

Address correspondence to: Assoc. Prof. Haim Azhari, Faculty of Biomedical Engineering, Technion IIT, Haifa 32000, Israel. E-mail: [email protected] 1395

1396

Ultrasound in Medicine and Biology

(1968). Intravenous injection of UCA is known to increase blood echogenicity and, thus, improve image quality on B-scan and Doppler scanning (Madjar 2001). The use of UCA combined with the B-scan technique has also been proposed as a means of improving the differentiation of breast tumors in ultrasonic imaging (Jung et al. 2005; Kettenbach et al. 2005) and also for kidney functional imaging (Verbeek et al. 2001). UCA is actually a microbubble suspension. The microbubbles typically have a mean diameter of few micrometers and are filled with high molecular-weight gases, which dissolve poorly in blood or air (de Jong et al. 2000). Current agents are stabilized either by encapsulation within a thin shell or by using hydrophobic, almost insoluble perfluor gases. Yet, even bubbles composed of insoluble gases need some coating to avoid rapid disappearance (Uhlendorf et al. 2000). Moreover, microbubble behavior is highly dependent on the acoustic pressure of the transmitted ultrasound. This is characterized by the mechanical index (MI). The MI is defined as (Apfel and Holland 1991): pneg MI 5 pffiffiffi fc

(1)

where pneg is the peak negative pressure of the transmitted ultrasound pulse and fc is its central frequency. For a given central frequency, at low acoustic pressures (MI , 0.1), microbubbles act as efficient scatterers, due to the large difference between their acoustic impedance and that of the surrounding media and their oscillations are synchronous with the incident ultrasound wave. At intermediate acoustic pressures (0.1 , MI , 0.5), oscillations become larger in amplitude and asynchronous with the incident wave, generating a nonlinear behavior where the bubbles are more easily expanded than compressed. In this range, the signals received from the nonlinear dynamics of the UCA can be potentially enhanced and differentiated from those received from the surrounding linear tissue. At high acoustic pressures (MI . 0.5), the oscillations increase even more in amplitude and microbubble disruption occurs (Correas et al. 2001). The ability to detect microbubbles presence in the tissue is the aim in nearly all medical applications of UCA. Among the most sensitive techniques available for detecting and imaging microbubbles in tissue is harmonic detection. This approach takes advantage of the fact that, as opposed to tissue, under specific transmission conditions microbubbles constitute a highly nonlinear medium for ultrasound waves (Simpson et al. 1999). The non-linear scattering exhibited by the microbubbles introduces harmonics in the ultrasound signal, which carry substantial energy (especially the second harmonic). However, when using a single-transducer ultrasound system, transducer bandwidth limitations will

Volume 36, Number 9, 2010

cause transmit and receive passbands to overlap. Consequently, the linear response from tissue will leak into the harmonic signal, reducing agent-to-tissue contrast. To solve this issue and increase contrast, both transmit and receive bandwidths should be narrowed, spoiling axial resolution (Verbeek et al. 2000). The use of harmonic detection in TT mode poses another technical challenge. The SNR is typically so high that even moderate amplification leads occasionally to signal saturation. This may hinder the detection of sub- or second-harmonic components. Hence, in this study harmonic detection was not implemented. Pulse inversion (PI) is a new technique that considerably overcomes the tradeoff between contrast detectability and imaging resolution imposed by harmonic imaging. Its main advantage is that it removes the fundamental signal and displays only the harmonic signal, using the full bandwidth of the transducer. PI has been shown to improve image quality when applied along with pulse-echo (Verbeek et al. 2000) and Doppler imaging (Simpson et al. 1999). Thus, in this study we have chosen to implement the PI technique to augment the detectability of the nonlinear effects stemming from the UCA. We have developed a computerized ultrasonic scanning system, which performs automatic and systematic scans using through-transmitted ultrasonic waves, to provide quantitative images depicting different acoustic parameters (Azhari and Sazbon 1999; Azhari and Stolarski 1997). We have utilized this system to perform scans before and after the injection of microbubbles, using both regular Gaussian pulses and the PI technique and studied the resulting changes in the acoustic properties of the waves. The purpose of this research was to investigate and quantify these changes and to study the feasibility of a new breast imaging method combining acoustic projection imaging and UCA injection. Our hypothesis is that UCA attenuates wave amplitude and causes an increase in the time-of-flight (TOF), i.e., the time it takes the waves to travel from the transmitter to the receiver, due to its gaseous content. In addition, we expect to observe a nonlinear behavior of the microbubbles when working at intermediate mechanical indices. MATERIALS AND METHODS Through-transmission (TT) imaging TT mode is mainly implemented in ultrasonic computed tomography of the breast (Schreiman et al. 1984). In the TT mode, scans are usually performed in the raster scanning trajectory where the transducers move synchronously along parallel lines. The signal in TT is in fact an acoustic projection and contains integrative information about the scanned object.

Contrast enhanced through-transmission imaging d T. ROTHSTEIN et al.

For example, by subtracting pulse transmission time from reception time, time-of-flight (TOF) is obtained. TOF is the integration of the reciprocal distribution of the SOS in the examined object and may be expressed as: TOFðx; yÞ 5

eCðx;1y; zÞdz

(2)

where z is the direction of the transmission, (x,y) constitute the projection plane and Cðx; y; zÞ is the SOS at each point in space. SOS is of great relevance in breast tumor characterization (Greenleaf and Bahn 1981; Edmonds et al. 1991). In addition, the amplitude of the received waves also constitutes important information on the nature of the object (Azhari 2010) and can be expressed as: Aðx; y; f Þ 5 A0 ,e2emðx;y;z;f Þdz

(3)

where A0 and Aðx; y; f Þ are the transmitted and received amplitudes, respectively, and mðx; y; z; f Þ is the attenuation coefficient at each point in space, which is a function of space (x,y,z) and frequency (f). Pulse inversion (PI) imaging In PI imaging, two consecutive ultrasound pulses are transmitted in each A-line. The second pulse is an inverted copy of the first and is transmitted after an appropriate delay. The summed responses from the pair are used to form one image line. If the medium is linear, then the response from the second pulse will be an inverted copy of the response from the first one and the result of their summation will be a null signal. For a nonlinear medium, such as UCA microbubbles, the positive and negative halfcycles of the wave experience different acoustic velocities, causing a distortion in the wave profiles (Verbeek et al. 2000). As a result, the responses are not inverted replicas of each other; hence, their sum is not zero. In this case, the magnitude of the result of their summation is related to the degree of nonlinearity (de Jong et al. 2000). Ultrasound contrast agent (UCA) The UCA used in the experiments is DefinityÔ (Bristol-Myers Squibb Medical Imaging, N. Billerica, MA, USA). It comprises phospholipid-encapsulated perfluoropropane microbubbles. According to the manufacturer, samples of these microbubbles have mean diameter ranging from 1.1 mm to 3.3 mm with 2% of them larger than 10 mm. Nevertheless, Kimmel et al. (2007) have reported that more than 98% of the bubbles have diameters below 1.3 mm and only 0.0008% of the bubbles are larger than 10 mm in diameter. Experimental set-up The computerized ultrasonic scanning system is schematically depicted in Figure 1. It uses the TT

1397

technique and can perform automatic scans, thus, ensuring that acquisitions can be precisely reproduced at the same spatial location. The system consists of a pair of focused transducers (V310; Panametrics, currently Olympus IMS, Waltham, MA, USA), submerged in a water tank and facing each other. The distance between the transducers is 20.5 cm. The focal zone is located at the center (a confocal configuration). The central frequency of the transducers is around 5MHz and their diameter is about 6 mm. The transducers were set up on a mechanical assembly with three degrees of freedom of movement, controlled by software developed in our lab. The system can make raster scans of a rectangular user-defined area (up to 8 cm in height and 18 cm in width). Phantom A phantom was assembled to simulate blood vessels. It consisted of a latex tube (8 cm length, 1.5 cm diameter and 0.2 mm wall thickness) connected at each of its ends to a regular 1-L saline bag. One bag was filled with saline and the other was empty. To create circulation, the bag filled with saline was placed higher and the empty bag lower than the tube. The tube was placed in the water tank, equidistant from the two transducers. Transmission technique and measurements Two series of experiments were performed, each using a different transmission technique. Regular R

T

8cm 18cm Pulser/ Receiver

Pulser

A/D

Fig. 1. Computerized ultrasonic scanning system-experimental set-up: An object is placed in a water tank between two ultrasonic transducers. A signal is transmitted from one transducer and detected by the other. The data are registered by an analog to digital (A/D) convertor and processed by a personal computer (PC). An image is obtained by scanning the object in raster mode.

1398

Ultrasound in Medicine and Biology

Volume 36, Number 9, 2010

Gaussian pulses were first used to quantify precontrast vs. postcontrast changes in wave amplitude, TOF and nonlinearity, while the PI technique was used to image the nonlinearity in a different manner. In both cases, the pulse repetition frequency (PRF) was 500 Hz and the ultrasonic signals were sampled at 100 MHz using an A/D digitizer (CompuScope 8247; Gage Applied Technologies, Illinois, USA). The scanning resolution was set to 0.3 mm along the horizontal axis and 1 mm along the vertical axis.

%Damplitude 5

amplitudepre 2amplitudepost 3100 (4) amplitudepre

The TOF subtraction image depicts the absolute time delay experienced by the ultrasonic waves in the presence of the contrast agent, relative to the baseline (no contrast agent): DTOF 5 TOFpost 2TOFpre

In addition, to check for the nonlinear behavior of the UCA, the positive and negative parts of the signal were analyzed separately. By analogy with eqn (4), changein-amplitudes images, %Damplpos and %Damplneg , were generated separately for the positive and then for the negative parts of the signal, where %Damplpos and %Damplneg represent the percent amplitude changes in the positive and negative parts, respectively. The attenuation undergone by the positive part of the signal was compared with the attenuation undergone by the negative part. To allow a quantitative comparison, the following normalized index was used:

Gaussian pulses. Two computer-controlled pulserreceivers (5800PR; Panametrics (Olympus IMS), Waltham, MA, USA) were used for signal transmission and detection. The scans were performed at four different pulse energy levels: 12.5 mjoules, 25 mjoules, 50 mjoules and 100 mjoules. These energy levels are presets of the 5800PR system and refer to the electrical pulses energizing the transducer. The MI was calculated using eqn (1), after measuring the actual acoustic pressure in water at the focal zone with a calibrated hydrophone (HNR-1000 needle hydrophone; Onda, Sunnyvale CA, USA). The MI values obtained for these four energy levels were: 0.10 (for 12.5 mjoules), 0.18 (for 25 mjoules), 0.31 (for 50 mjoules) and 0.40 (for 100 mjoules). (It is assumed that these values represent the values at the tube’s lumen since the attenuation of the thin latex wall is negligible). The obtained raw data were first band-pass filtered to reduce measurement noise. For each received line-of-sight signal (A-line), which corresponds to one pixel in the final image, the amplitude and TOF were measured, yielding two corresponding pixel levels in the images. Next, for each energy level and for each acoustic property (amplitude and TOF), the image of the postcontrast scan was subtracted from the precontrast one. The amplitude subtraction image was further normalized by the precontrast one, yielding the percent change in amplitude following contrast agent injection:

index 5

%Damplneg 2%Damplpos %Damplpos

1

1

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

0

-0.2

-0.2

-0.4

-0.4

-0.6

-0.6

-0.6

-0.8

-0.8

-0.8

-1

-1 1

2

3

4

5

6

7

8

9

10

(6)

Pulse inversion. Three different double inverted pulses were used: double Gaussian, double square and double sine (Fig. 2). In all cases, the second half of the pulse is an inverted copy of the first. The pulses were designed in MatlabÒ (The MathWorks, Inc., Natick, MA), transmitted through an arbitrary waveform generator (8025; Tabor Electronics Ltd., Haifa, Israel) and received using a computer controlled pulser-receiver (5800PR; Panametrics). The MI-values were always kept in the intermediate range (0.1 , MI , 0.5). The frequencies of the pulses were 5.5 MHz, 5.3 MHz and 5.5 MHz and the time delays between the two inverted portions were

0.8

0

(5)

0.6 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1

0.4 0.2 0 -0.2 -0.4 2.5

3

3.5

4

4.5

5

-1 0

0.5

1

1.5

2

2.5

3

3.5

4

0

5

10

15

time [µsec]

time [µsec]

time [µsec]

GAUSSIAN

SQUARE

SINE

20

Fig. 2. The three types of double pulses used in the pulse inversion technique. For clarity, a zoomed view of one of the sine waves is shown in the right figure.

25

Contrast enhanced through-transmission imaging d T. ROTHSTEIN et al.

5 ms, 5 ms and 2 ms for the double Gaussian, double square and double sine, respectively. The double sine was 44 cycles long (each half). For all three types of double pulses used, the received signals were analyzed for nonlinearity by applying the following steps: (1) The signal was first band pass filtered using one of the preset integrated filters of the Panametrics 5800PR system (through a combination of a low-pass filter of 35MHz and a high-pass filter of 300kHz). After band-pass filtering, each received line-of-sight signal (A-line) was cut between the two inverted portions. The second half was then re-inverted so that it then had the same phase as the first one. An algorithm was used to slide one portion toward the other until the point of maximum cross-correlation between them was found. (2) The values of the two portions were then subtracted at every point, resulting in a new signal termed difference of pulses (DOP). This signal was a measure of the discrepancy between the two halves: the higher it is, the higher the nonlinearity in the medium. (3) Nonlinearity was then assessed by measuring three different parameters of the DOP: ‘‘max’’ (the maximum absolute value of the DOP vector), ‘‘sum’’ (the sum of all absolute values of the DOP vector) and ‘‘rms’’ (root mean square of the DOP vector). All three were normalized by the original signal’s amplitude, which was taken as the average between the largest absolute value of the two halves. Using this method, three different images depicting nonlinearity were obtained. (4) Finally, the pre and postcontrast DOP images were compared. Procedure In the first stage, saline was circulated through the tube and reference scans were performed (precontrast). Gain was set in the beginning of each experiment and kept constant throughout the experimental process. Different gains, though, were used for different transmission techniques or different energy levels of the Gaussian pulses to avoid signal saturation. Next, a DefinityÔ vial, which had been kept refrigerated at 4  C, was brought to room temperature and activated by a shaker (VialmixÔ; Bristol-Myers Squibb Medical Imaging, N. Billerica, MA, USA). A suspension of the UCA was then added to the upper saline bag at the concentration of 0.3 mL/L and a new set of scans was performed (postcontrast). In vivo study A female volunteer with suspicious findings in one breast was assessed to study the clinical applicability of

1399

the proposed method, as a noninvasive method for breast cancer detection. This procedure was approved by our institutional ethical committee and a written consent was obtained from the patient before the study. The subject was positioned with respect to the scan plane by lying prone on the scanner bed. The breast being examined was immersed in degassed water contained within a small cylindrical rubber tank. The rubber tank was immersed in a larger water tank where the ultrasonic transducers, the scanner mechanical assembly and the motors were located. There was no physical contact between the transducers and the breast. An intravenous flexible plastic cannula was introduced into the patient hand vein. The breast was scanned with the Gaussian pulses at 25 mjoules (MI 5 0.18). Scans were obtained once using a lateral view and then using a cranio-caudal view. Each acquisition required about 4 min. Next, a DefinityÔ vial (Bristol-Myers Squibb Medical Imaging, N. Billerica, MA, USA) was activated by a shaker (VialmixÔ, BristolMyers Squibb Medical Imaging, N. Billerica, MA, USA). One mL of activated DefinityÔ, which contains a maximum of 1.2 3 1010 perflutren lipid microspheres, was diluted in 4 mL of normal saline to form a 5 mL suspension. Before acquiring each of the views, 2.5 mL of the suspension was manually injected into a hand vein as a bolus at a rate of about 0.5 mL/s, followed by a 5 mL saline flush. Image acquisition started 30 s after each injection. The patient was asked not to move during the entire examination. Projection images of the TT waves before and after UCA injection were then generated as explained above. RESULTS Gaussian pulses Amplitude. Significant decreases in wave amplitude resulting from UCA injection were observed in the phantom [%Damplitude5 22% 6 13% (mean 6 std)]. The results are shown in Figure 3. As can be noted, these changes were MI-dependent and maximal changes occurred for MI 5 0.18 (corresponding to pulse energy of 25 uJ). However, in the image regions corresponding to water (i.e., outside the tube) there were no significant changes in amplitude (m 5 0, p , 0.001). TOF. Significant changes were also detected in TOF in the phantom because of the UCA injection (TOF 5 7.9 ns 6 6.3 ns). These changes are shown in Figure 4. TOF changes were also MI-dependent and maximal changes occurred for MI 5 0.18. Again, there were no significant changes in TOF (m 5 0, p , 0.001) in the regions of the image corresponding to water (i.e., outside the tube). Nonlinearity. By comparing the positive and negative parts of the signal separately, we calculated the amount of change in amplitude for each part. The

1400

Ultrasound in Medicine and Biology Decrease in amplitude [%]

a

50

30

Decrease in amplitude [%]

60

38±1

40

Volume 36, Number 9, 2010

negative 48 ± 1

50

28±1

positive

40 15±4

20

9±2

10

30 ± 1

30

27 ± 1

28 ± 2 21 ± 1

20

0 0.10

0.18

0.31

0.40

14 ± 2 10 ± 1

10

10 ± 2

0

mechanichal index (MI)

0.10

Fig. 3. Percent change in ultrasound wave amplitude resulting from ultrasound contrast agents (UCA) injection, depicted for the four mechanical index (MI) levels. Data are presented in terms of mean values. The error bars correspond to the standard deviation. Positive values represent a decrease in amplitude. (p , 0.001 for all MI levels)

0.18

0.31

0.40

mechanichal index (MI)

b

Index of non-linearity [%]

100

73 ± 14

80

51 ± 21

60

corresponding percent decreases in amplitude are depicted in Figure 5a (%Damplneg 5 27% 6 14%, %Damplpos 5 20% 6 8%). The discrepancy between the decreases in the positive and negative parts were found to be significant (p , 0.001), except for the highest MI (0.4). Calculating the corresponding indices of nonlinearity [eqn (6)], maximal effect was again detected for MI 5 0.18, as shown in Figure 5b. Pulse inversion Figure 6 shows typical scan images depicting DOP distribution at and around the phantom (in this case, using the double Gaussian pulse). Each row of the figure shows the DOP image calculated by a different method as explained above (i.e., max, sum and RMS). In the first column, images of precontrast scans are depicted. As can be observed, these images are homogeneous. In the second column, images of postcontrast scans are shown, where the silhouette of the latex tube is rather well defined. The third column illustrates the percent increase in DOP. For this specific scan, this increase was approximately zero outside the tube and approximately 70% inside the tube. Increase in TOF [nsec] 17.3 ± 3.1

20 15

6.1 ± 2.8

10 5

4.1 ± 1.2

4.3 ± 2.6

0 0.10

0.18

0.31

40 20

13 ± 7

4 ± 13

0 -20 0.10

0.18

0.31

0.40

mechanichal index (MI)

Fig. 5. Nonlinearity measurements with Gaussian pulses. (a) Percent change in amplitude of positive vs. negative parts of the ultrasound pulse resulting from ultrasound contrast agents (UCA) injection, where positive values represent a decrease in amplitude. (b) Index of nonlinearity [eqn (6)]. Data are presented in terms of mean values. The error bars correspond to the standard deviation.

Figure 7 and Table 1 summarize the results obtained with the PI technique using the three different pulse shapes. All three types of double pulses and all three parameters (‘‘max’’, ‘‘sum’’ and ‘‘rms’’) show significant increase in DOP after contrast injection in the region corresponding to the phantom tube (DOP 5 56% 6 17%, p , 0.001). It was found that all tested pulses differed significantly from each other (p , 0.001). The double square pulse showed the highest increase in DOP, while the double sine showed the poorest increase. No significant differences were found between the three DOP parameters (‘‘max’’, ‘‘sum’’ and ‘‘rms’’), (p . 0.3), implying that none of these computation methods had an advantage over the other two. Finally, there were no significant changes in DOP outside the phantom tube (m 5 0, p , 0.001), indicating that DOP changes stem only from the UCA.

0.40

mechanichal index (MI)

Fig. 4. Change in ultrasound wave time-of-flight resulting from ultrasound contrast agents (UCA) injection, depicted for the four mechanical index (MI) levels. Data are presented in terms of mean values. The error bars correspond to the standard deviation. Positive values represent an increase in time-of-flight. (p,0.001 for all MI levels)

In vivo: preliminary results The preliminary results obtained for the in vivo case are presented in this section. This patient had a large area of deformed architecture observed by X-ray mammography (Fig. 8, top). Three sampled biopsy samples were taken. The microscopic findings indicated a chronic

Contrast enhanced through-transmission imaging d T. ROTHSTEIN et al.

1401

Fig. 6. Example of scan images of difference of pulses (DOP) using the double square pulse. (a) Precontrast scan. (b) Postcontrast scan. (c) Percent increase in DOP. The estimated borders of the vessel are marked for clarity on the precontrast images by double-headed arrows.

infection with fibrosis (nonmalignant). A visual comparison was then done between the X-ray mammogram and the acoustic mammograms. The corresponding image depicting the relative change in amplitude (Fig. 8, bottom) has shown many highlighted, geometrically undefined regions. However, when studying the changes in SOS (Fig. 8, middle), a large defined area was noted on the left side of the mediolateral view. As can be noted the location of this region matches the location of the suspicious region observed in the X-ray mammogram. (Nevertheless, it should be recalled that the X-ray mammogram was obtained while the patient was standing and the breast was squeezed. On the other hand, the ultrasonic mammograms were obtained when the patient was lying prone on the scanner bed and the breast was not squeezed. Thus, accurate comparison is not possible.)

Increase in DOP [%] 140 max

120

sum

100

rms 77

78

34

An ultrasonic imaging method suitable for breast imaging has been introduced. It combines computerized TT (projection) imaging with UCA injection followed by image subtraction. This proposed method is similar in concept to the commonly used breast imaging technique utilized in magnetic resonance imaging (MRI). The main advantage offered by the ultrasound option is naturally its cost effectiveness. In this article, we demonstrate that quantitative analysis of TT ultrasound before and after UCA injection shows significant differences in amplitude and in TOF. The main reason for amplitude reduction stemming from the UCA injection is the high reflection and scattering coefficients in the transition from the surrounding media to the microbubbles due to their very small acoustic impedance. This implies that more energy is reflected and scattered, so less is transmitted, leading to a decrease in the transmitted wave’s amplitude. The observed changes in the TOF correspond to changes in the SOS through the UCA-filled regions. Evidently, the gaseous filling of the bubbles forms regions

80

71 56

35

DISCUSSION

56

56

40

60 40 20

Table 1. Average percent increase in DOP and SD values obtained for the three PI pulses used here and for the three methods of analysis implemented (see also Fig. 7 for graphical visualization)

0 Double Sine

Double Square

Double sine

Double square

Double Gaussian

40% 6 21% 34% 6 20% 35% 6 20%

71% 6 41% 78% 6 53% 77% 6 51%

56% 6 13% 56% 6 12% 56% 6 12%

Double Gaussian

Fig. 7. Percent increase in difference of pulses (DOP), shown for the three different double pulses used and three computation parameters. Mean and standard deviation values are presented numerically in Table 1. (p , 0.001 in all cases).

‘‘Max’’ ‘‘Sum’’ ‘‘Rms’’

DOP 5 difference of pulses; SD 5 standard deviation.

1402

Ultrasound in Medicine and Biology

Volume 36, Number 9, 2010

Fig. 8. Results of the in vivo study. (Top) X-Ray mammogram of the breast with the suspicious region marked by a dashed ellipsoid. (Middle) Image depicting the absolute changes in SOS show a large defined area (dashed ellipses) on the left side of the mediolateral view. As can be noted, the location of this region matches the location of the suspicious region shown in the mammogram. (Bottom) The corresponding image depicting the relative amplitude change show many highlighted, geometrically undefined regions.

of medium where the bulk modulus is substantially reduced. Consequently, the SOS is reduced and the corresponding TOF is increased in the postcontrast images. In this study, we have also examined the effect of MI value on the acoustic parameters. As can be concluded from Figures 3, 4 and 5, the effects stemming from the UCA injection depend on the magnitude of the MI. Indeed, all four energy levels used here have mechanical indices that match the theoretical intermediate acoustic pressure range (0.1 , MI , 0.5) mentioned before (Correas et al. 2001). Nevertheless, the results obtained here suggest that MI-values in the range of 0.18 to 0.31

better augment the UCA nonlinear behavior within this intermediate range, while the MI 5 0.1 is below and MI 5 0.40 is above this effective range. The nonlinear behavior of the media containing UCA is clearly visible in Figure 5. Our results show that the negative portion of the wave undergoes significantly higher amplitude attenuation than the positive one up to MI 5 0.4 where the discrepancy is negligible. The combined results stated above regarding the effect of MI on the acoustic parameters studied here clearly imply the existence of an optimal MI value in ultrasonic contrast enhanced TT mode. Our findings indicate that the optimal value is around

Contrast enhanced through-transmission imaging d T. ROTHSTEIN et al.

MI 5 0.18. However, these conclusions should be restricted solely to the use of DefinityÔ. Other UCA materials may have different optimal values and effective range. Significant increases in nonlinearity following UCA injection were also revealed by applying the PI technique with double Gaussian, double square and double sine waveforms (p , 0.001). The results of the experiments show that PI concepts are still valid when working in TT mode. This means that the distortion of the pulse, caused by its positive and negative half-cycles’ experiencing different acoustic velocities, occurs not only when the pulse is reflected by the microbubbles (backscatter), but also when it travels through them (forward scatter). The highest contrast was achieved when using the double square pulse (Fig. 7). This may be attributed to the fact that the square pulse better excites the transfer function of the transducer. Thus, it could be more robust and less susceptible to deformations stemming from noise sources, leaving the nonlinearity deformations more emphasized. Another possible reason is that the double square pulse used here carries more energy on lower frequencies than the two other pulse shapes tested and low frequency waves are more sensitive to the nonlinearity of the medium [(higher MI – see eqn (1)]. As for the in vivo feasibility case, it was noted that only certain regions of the breast have depicted changes in the post injection images at the time of scanning (particularly in the SOS change images). These findings may support our hypothesis that when the UCA microbubbles reach the scanned volume or target tissue (i.e., tumor) through the blood stream, the scatter of ultrasonic waves from this region is temporarily augmented. However, once the contrast material is washed out, the scatter reverts to its previous value. The rate by which contrast material flows into the target region and by which it is washed out, depends on the efficiency of the blood vessels within that region. Our explanation is that since angiogenesis in tumors is unorganized, more contrast materials will be accumulated there. Consequently, we expect to detect significant changes as a function of time in the acoustic properties of these regions. Evidently, there are notable localized changes occurring in the acoustic parameters of the breast tissue as a result of the injected contrast material. However, we still have to accumulate more clinical data and seek characteristic patterns. Another technical point, which should be noted, is that piezoelectric transducers, such as those utilized in this study, are phase sensitive. The electrical response of a phase sensitive receiver is thus proportional to the average complex pressure over the area of the receiver. Naturally, when passing through a complicated structure such as the breast the ultrasonic pressure field over the area of the receiver aperture is nonuniform in terms of

1403

amplitude and especially phase. Consequently, as shown by Buess and Miller 1981, inaccuracies may be induced into the amplitude measurements. As a remedy, they have suggested the use of phase insensitive receiver, which they have shown to provide a more reliable estimate of intensity sensitive parameters. In conclusion, a new ultrasonic imaging method combining acoustic projection imaging and UCA injection was presented. This approach allows quantification of different acoustic properties and, thus, potentially better characterization of the medium. The results obtained here for the phantom together with the preliminary ones from the in vivo study indicate the technical feasibility of the suggested method. We hope that this new technique may provide a novel, cost-effective, noninvasive method for breast tumors detection and characterization that may be complementary or alternative to breast MRI. Acknowledgments—The authors wish to acknowledge the Israel Cancer Research Fund (grant number 2009491) and Israel Cancer Association (grant number 2010851) for their financial support of this research and Aharon Alfassi for his valuable technical support.

REFERENCES Apfel RE, Holland CK. Gauging the likelihood of cavitation from short-pulse, low-duty cycle diagnostic ultrasound. Ultrasound Med Biol 1991;17:179–185. Azhari H, Stolarski S. Hybrid ultrasonic computed tomography. Comput Biomed Res 1997;30:35–48. Azhari H, Sazbon D. Volumetric imaging using spiral ultrasonic computed tomography. Radiology 1999;212:270–275. Azhari H. Basics of Biomedical Ultrasound for Engineers. Hoboken, NJ: John Wiley & Sons, Inc., 2010. Busse LJ, Miller JG. Detection of spatially nonuniform ultrasonic radiation with phase sensitive (piezoelectric) and phase insensitive (acoustoelectric) receivers. J Acoust Soc Am 1981;70:1370–1376. Chiang HR, Tiu CM, Chen SP, Chou YH, Chiou SY, Chiou HJ, Wang HK, Yen CH, Chang CY, Davidson DA, Chen J. Automated full-field breast ultrasound possesses potential for screening breast cancer in women with dense breasts. Ultrasound Med Biol 2006; 32:107. Correas JM, Bridal L, Lesavre A, Me´jean A, Claudon M, He´le´non O. Ultrasound contrast agents: Poperties, principles of action, tolerance, and artifacts. Eur J Radiol 2001;11:1316–1328. de Jong N, Frinking PJ, Bouakaz A, Ten Cate FJ. Detection procedures of ultrasound contrast agents. Ultrasonics 2000;38:87–92. Edmonds PD, Moretensen CL, Hill JR, Holland SK, Jensen JF, Schattner P, Valdes AD, Lee RH, Marzoni FA. Ultrasound tissue characterization of breast biopsy specimens. Ultrasonic Imaging 1991;13:162–185. Gramiak R, Shah P. Echocardiography of the aortic root. Invest Radiol 1968;3:356–366. Greenleaf JF, Bahn RC. Clinical imaging with transmissive ultrasonic computerized tomography. IEEE Trans Biomed Eng 1981;28: 177–185. Jung EM, Jungius KP, Rupp N, Gallegos M, Ritter G, Lenhart M, Clevert DA, Kubale R. Contrast enhanced harmonic ultrasound for differentiating breast tumors - First results. Clin Hemorheol Micro 2005;33:109–120. Kettenbach J, Helbich TH, Huber S, Zuna I, Dock W. Computer-assisted quantitative assessment of power Doppler US: Effects of microbubble contrast agent in the differentiation of breast tumors. Eur J Radiol 2005;53:238–244.

1404

Ultrasound in Medicine and Biology

Kimmel E, Krasovitski B, Hoogi A, Razansky D, Adam D. Subharmonic response of encapsulated microbubbles: Conditions for existence and amplification. Ultrasound Med Biol 2007;33:1767–1776. Madjar H. Contrast ultrasound in breast tumor characterization: Present situation and future tracks. Eur J Radiol 2001;11(Suppl 3):E41–E46. Schreiman JS, Gisvold JJ, Greenleaf JF, Bahn RC. Ultrasound computed transmission tomography of the breast. Radiology 1984;150:523–530. Shipley J, Duck F, Goddard D, Hillman M, Halliwell M, Jones M, Thomas B. Automated quantitative volumetric breast ultrasound data-acquisition system. Ultrasound Med Biol 2005;31:905–917. Simpson DH, Chin CT, Burns PN. Pulse inversion Doppler: A new method for detecting nonlinear echoes from microbubble contrast

Volume 36, Number 9, 2010 agents. IEEE Trans Ultrason Ferroelectr Control 1999;46: 372–382. Uhlendorf V, Scholle F-D, Reinhardt M. Acoustic behaviour of current ultrasound contrast agents. Ultrasonics 2000;38:81–86. Verbeek XAAM, Ledoux LAF, Willigers JM, Brands PJ, Hoeks APG. Experimental investigation of the pulse inversion technique for imaging ultrasound contrast agents. J Acoust Soc Am 2000;107: 2281–2290. Verbeek XAAM, Willigers JM, Prinzen FW, Peschar M, Ledoux LAF, Hoeks APG. High-resolution functional imaging with ultrasound contrast agents based on RF processing in an in vivo kidney experiment. Ultrasound Med Biol 2001;27:223–233.