Ultrasonic characterization of selected renal tissues

Ultrasoundin Med. &Biol.Vol. 15, No. 3, pp. 241-253, 1989

0301-5629/89 $3.00+ .00 © 1989PergamonPresspie

Printed in the U.S.A.

OOriginal Contributions ULTRASONIC

CHARACTERIZATION

OF SELECTED

RENAL

TISSUES

D. H. TURNBULL, Jf S. R. WILSON,:~ A. L. HINE:~§ a n d F. S. FOSTERt II ~Ontario Cancer Institute and Department of Medical Biophysics, University of Toronto, 500 Sherbourne Street, Toronto, Canada M4X I K9 ~Department of Radiology, Toronto General Hospital, Toronto, Canada (Received 12 May 1988; in final form 3 October 1988)

Abstract--Velocity, attenuation, and hackscatter of ultrasound were measured in human renal tissues over a frequency range relevant to clinical imaging (3.5-7 MHz). Normal renal tissues, as well as three types of mass (angiomyofipoma, renal cell carcinoma, and oucocytuma) were studied, and comparisons made of the appearance of the tissues in clinical images to their ultrasonic and pathological properties. The results showed angiomyoHpema had high attenuation and hackseatter coefficients due to acoustic impedance differences between fat and smooth muscle components of the tumour. The renal cell carcinomas were indistinguishable from normal kidney tissue, except in one case where infiltration by fatlike macrophages led to high attenuation and hackscatter coefficients. This finding also supports the conclusion that fat/nonfat interfaces are a dominant scatter mechanism in renal tissues.

Key Words: Ultrasound, Renal tissue characterization, Angiomyollpoma, Renal cell carcinoma, Oncocytoma, Fat, Velocity, Attenuation, Backscatter.

INTRODUCTION

couraged this course based on the findings that angiomyolipoma does not always appear completely echogenic, and that the malignant tumour, renal cell carcinoma occasionally does appear echogenic (Hartman et al. 1981). We felt that a quantitative study of the ultrasound properties of renal tissues, including angiomyolipoma and renal cell carcinoma, would provide data relevant to renal tissue characterization studies, and would help to resolve uncertainties regarding angiomyolipoma. The general approach taken in this study is summarized in Fig. 2. In vivo (diagnostic) images of kidneys were obtained for each patient included in this study. Surgically removed kidneys, containing a mass, were scanned in a water bath to obtain in vitro images which were used to corroborate findings in the in vivo images. Then one half of the kidney was used for pathologic evaluation while the other half underwent quantitative evaluation of the ultrasonic velocity, attenuation, and backseatter coefficients using the ultrasound macroscope (Foster et al. 1984; D'Astous and Foster 1986). With this approach, we have studied the properties of normal renal tissues, as well as three different types of renal mass (angiomyolipoma, renal cell carcinoma, and oncocytoma), and compared the appearance of the tissues in clinical images to their ultrasonic and pathologic properties.

Quantitative tissue characterization with ultrasound has been the motivation for many past investigations, the eventual goal being to increase the diagnostic potential of clinical ultrasound imaging devices. Several human tissues have been the subject of extensive characterization studies, and a general review of this work can be found in Greenleaf (1986). Human renal tissues have not been subjected to the same scrutiny, although measurements have been made of ultrasound velocity, absorption, and attenuation in kidney tissue from other mammals (Bowen et al. 1979; Goss et al. 1979; Le Croissette et al. 1979; Lele et al. 1976). The present study was motivated by repeated clinical observations that a benign renal tumour, angiomyolipoma, has a characteristic echogenic appearance (Fig. 1) in ultrasound images (Behan and Kazam 1978; Bret et al. 1985; Hartman et al. 1981; Seheible et al. 1978). Some investigators have suggested that this echogenic appearance may be used to diagnose the tumour and allow conservative treatment (Scheible et al. 1978). Others have strongly dis§Current address: X-Ray Department, Central Middlesex Hospital, London, LIK. NResearch scholar of the National Cancer Institute of Canada. 241

242

Ultrasound in Medicineand Biology

Volume15, Number 3, 1989

0)

To doto collection system

F ~ / w / Quortz opticol

,55u.

\

Sphericollyfocused tronsducer(f/2.5,SMHz) / /Roster 5con

_ _ \ V _ _//

/ .... / I

\\

\\

Omm

Mocroscope ottenuotion imoge

Fig. 1. Sagittal scan (3.5 MHz) of a right kidney with echogenie angiomyolipoma (arrow).

0 ®

METHODS

Experimental system The ultrasound macroscope, which has been described in detail by Foster et al. (1984) and D' Astous and Foster (1986), is depicted schematically in Fig. 3. A tissue sample approximately 4 cm X 4 cm X 5 mm was placed between an optical quartz reflector and a thin plastic membrane in a constant temperature (37°C) bath of isotonic saline. The sample was positioned to lie in the focal plane of a 5 MHz, 2-cm diameter, f/2.5 PZT transducer with a lateral beam width at half maximum amplitude of 0.75 mm and a depth of field of approximately 13 ram. The first step in the measurement procedure consisted of moving the transducer in a raster fashion over a 256 × 256 grid (4 cm X 4 cm field of view), and generating

Diagnostic J Image. Ultrasound

l

Jln j Imiges

Vitro

3.5,5 MHz

5,7MIZ

Pathology /

Sectional

Images

Attenuation Velocity Bxckscatter 4 X4 x 1/2cm 5 MHz

37" C

Detailed Frequency Analysis

Attenuation Backscatter 5x5x5mm 3.5 -7 MHz

37° C

Fig. 2. Overview of renal tissue characterization study showing inputs taken from clinical images, pathological examinations, and quantitative macroscope studies.

Regionof interest Fig. 3. (a) Schematic diagram of the ultrasound macroscope. A tissue sample is placed in the focal zone of a spherically focussedtransducer. As the transducer is moved in raster fashion, data related to attenuation, velocity or backscatter are accumulated and converted into images. (b) From these images, regions of homogeneous tissue are selected for the measurement of frequency dependent attenuation and backscatter. attenuation, velocity, and (optionally) peak backscatter images of the entire tissue sample. From these images, 0.5 cm X 0.5 regions of homogeneous tissue were selected for further investigation. The tissue in the region of interest was rescanned over an 8 x 8 grid. At each point in the grid, the gated scatter signal and the reflected signal from the quartz were recorded by a Biomation 8100 transient digitizer (Gould Inc., Santa Clara, CA) at a sampling rate of 100 MHz. These signals were used to compute frequency dependent attenuation and backscatter coefficients for the tissue in the region of interest. Computation and display were performed on a VAX 11/780 computer (Digital Equipment Corporation, Maynard, MA) with a GENISCO display system. After scanning, the tissue samples were fixed in 10% formalin and retained for histological examination.

Measurement theory The measurement theory used in this study follows closely that of D'Astous and Foster (1986), and will not be described in detail here. The velocity at some point in the tissue is computed from time of

Ultrasonic characterization of selected renal tissues • D. H. TURNBULL el al.

flight data (Foster et al. 1984). The average frequency dependent attenuation coefficient a(p) corresponding to a region of interest is obtained by comparing the magnitude spectra of tissue attenuated rfsignals, a(x~, yj, t), to the magnitude spectrum of a reference signal ao(t), which does not pass through the tissue, according to the formula: ot(l))

=

18{ o j=l

- 27" loglo

ladv) l

t

8

dB/mm,

8 j=l

[

ff~q

× 2r(1 - cos Or)

IS'(xi, yj, lO[ 2 IA0(v)l 2

backscatter coefficients, because of a lack of accepted standards, we have been unable to absolutely confirm the accuracy of the measurement procedure. However, we can state that the results reported here were reproducible, and in a relative sense reflect the true properties of the tissues.

Tissue samples

where l is the thickness (in mm) of the tissue sample, (x~, yj) are the spatial coordinates of the 8 × 8 grid covering the region of interest, and IA(xi, yj, v) l and IA0(v) l are the magnitude spectra of a(xi, yj, t) and ao(t), respectively. To compute the frequency dependent backscatter coefficient corresponding to a region of interest, the focus of the transducer is moved to the centre of the tissue sample, and a scatter signal s(x~, yj, t) is collected, gated over approximately 2 mm of tissue. This scatter signal is then corrected for attenuation by applying an "inverse attenuation" filter in a piecewise fashion (D'Astous and Foster 1986). The average backscatter coefficient ~(~) corresponding to the attenuation corrected scatter signals s'(x~,yj,t) is computed according to the formula: 1

243

1 (z2 --

} ZI)

'

where Y/q is the intensity reflectance of the water quartz interface, 0r, is the half angle of the transducer subtended at its focus, IS'(xi, yj, z,)l2 is the energy spectrum of the attenuation corrected scatter signal s'(x~, yj, t), IA00,)l 2 is the energy spectrum of the reference signal ao(t), and z~ and z2 are the distances in mm corresponding to the beginning and end of the gated scatter signal. In this formulation the backscatter coefficient is computed in terms of the energy of the scattered radiation from the tissue relative to the energy of the reference signal from a plane reflector. The derivation in D'Astous and Foster (1986) is based on a single frequency backscatter coefficient given by Foster et al. (1984), which was in turn based on a derivation for an unfocussed transducer geometry given by Nicholas et al. (1982). The accuracy of the attenuation measurements made in the macroscope has been evaluated by comparing results from castor oil to previously reported data (D'Astous and Foster 1986). In the case of the

From September 1986 to June 1987, two patients with angiomyolipoma, and eight patients with renal masses of other types (six renal cell carcinoma, and two oncocytoma) undergoing radical nephrectomy were entered in the study (see Table 1). The only factor which affected selection was mass size, with preference given to smaller renal tumours. This selection criterion allowed us to obtain normal kidney data from areas of the renal parenchyma adjacent to the tumour sites, and enabled comparisons to be made between different tumour types similar in size to the two available cases of angiomyolipoma. All patients underwent in vivo, routine sonographic evaluation of the appropriate kidney using commercially available equipment (Siemens Sonoline II, Erlangen, West Germany) with 3.5 and 5-MHz transducers. Immediately following surgical removal, in vitro water bath scans were performed on the entire kidney using an Aloka 520 (Tokyo, Japan) with 5 and 7 MHz transducers. In general, good agreement was obtained between clinical evaluations from the in vivo and in vitro images. Then half the kidney was retained for pathological examination, while the other half was immediately frozen and subsequently sectioned for examination of the ultrasound properties using the macroscope. Tissue samples were sectioned to a uniform thickness of 5 mm using a rotary blade meat cutter, and then cut to the approximate dimensions 3.5 cm × 3.5 cm to fit into the 4 cm × 4 cm field of view of the macroscope. Sectioned samples were taken from the freezer and placed immediately into a constant temperature (37°C) bath of degassed, isotonic saline solution (8.78-g NaC1 per 1 1 of distilled water). The samples were gently massaged after thawing to remove any trapped air bubbles, and the scanning procedure was initiated 10-20 min after placing the sample in the tank. While it is difficult to assess whether this procedure was sufficient to remove very small air bubbles, in several samples where this procedure was not followed the presence of air bubbles was easily detected from the resulting extremely high scatter signals. The scan time for each sample was approximately 4 h; 3 h to generate the attenuation, velocity and peak backscatter images of the whole sample, and 1 h to collect data from the regions of interest.

244

Ultrasound in Medicine and Biology

Volume 15, Number 3, 1989

Table 1. Tabulation of patients, samples and regions of tissue in which the ultrasound velocity and frequency dependent attenuation and backscatter coefficients were measured [AML = angiomyolipoma; RCC ffi renal cell carcinoma; ONC = oncocytoma] Patient

M/F

Age

1

F

68

2

F

39

3

M

66

4

M

62

5

M

70

6

M

73

7

M

75

8

M

44

9

M

66

l0

M

76

Clinical ultrasound observations

Path

3.5-cm mass with echogenic and hypoechoiccomponents Impression: possible AML 7-cm highly echogenic mass with echopoor centre Impression: AML with possible central myomatous component or hemorrhage 3.5-cm quite echogenic mass Impression: Indeterminate solid mass, possible AML 4-cm complex solid and cystic mass Impression: Cystic RCC 4-cm isoechoic nonuniform mass Impression: RCC 2-cm isoechoic mass Impression: possible RCC 7-cm nonuniform solid mass Impression: RCC End stage kidney with 2.5-cm isoechoic mass Impression: RCC 1.5-cm mass with echogenic rim Impression: possible RCC Aspiration biopsy--positive cytology 3.5-cm isoechoic mass with calcium Impression: RCC

Pathological evaluation o f selected renal tissues was performed on both the tissue samples scanned in the macroscope, and on tissue samples from the other half of the kidney, which had not been frozen or subjected to the 4 h o f scanning. T o check that the ultrasound properties o f the tissue were not being significantly altered during the course o f the experiment, several m e a s u r e m e n t s were m a d e on selected regions o f various renal tissues before and after the 4 h scan time. There was no significant change in the velocity or frequency dependent attenuation and backscatter coefficients in these control experiments. In addition, D'Astous and Foster (1986) performed control experiments to determine the effect o f the freeze-thaw process on samples o f breast tissue, and concluded that the resulting change

Samples

Regions/tissue

AML

2

3/AML fat 2/AML muscle

AML

2

1/Kidney 3/AML fat 6/AML muscle

ONC

7

7/Kidney 9/Renal fat 4/ONC

Cystic RCC

4

RCC

5

ONC

4

RCC

4

2/Kidney 2/Renal fat 5/RCC 9/Kidney I/Renal fat 2/RCC 3/Kidney 3/Renal fat l/ONe 2/Kidney 7/RCC

RCC

2

2/RCC

RCC

5

3/Kidney 6/Renal fat 2/RCC +2 Echogenic RCC

RCC

5

3/Kidney 4/Renal fat 2/RCC

in attenuation coefficient was ___6%, while the change in backscatter coefficient was +9%. The effect o f the freeze-thaw process on h u m a n renal tissues was not evaluated.

Experimental protocol: A case study T h e protocol followed in this study is best described by presenting one case o f angiomyolipoma. Patient 1 (Table 1), a 68-year-old female, presented with s y m p t o m s referable to the gastrointestinal tract. Ultrasound demonstrated several masses related to her stomach as well as an inhomogeneous renal mass with clearly defined echogenic a n d less eehogenic c o m p o n e n t s (Fig. 4(a)). A radical n e p h r e c t o m y was performed. The t u m o u r was found to be an angiom y o l i p o m a , the echogenic c o m p o n e n t o f the mass

Ultrasonic characterization of selected renal tissues • D. H. TURNBULL el al.

(a)

(b)

Fig. 4. (a) Sagittal scan (3.5 MHz) of a left kidney (dots) with an angiomyolipoma in the lower pole (+). The tumour has well-defined echogenic (arrow) and less echogenic regions. (b) Gross pathology from radical nephrectomy, showing the angiomyolipoma composed of a fat component correlating with the echogenic region, and a smooth muscle component correlating with the less echogenic region.

correlating well with the fatty portion of the tumour, and the less echogenic part corresponding to the myomatous or muscular elements (Fig. 4(b)). A tissue sample from the angiomyolipoma was scanned in the macroscope, resulting in the 5 MHz, 256 X 256 X 8 bit velocity and attenuation images shown in Figs. 5(a) and (b), respectively. The velocity image ranges from 1420 m/s (black) to 1580 m/s (white). The attenuation image ranges from 0 d B / m m (black) to l d B / m m (white). Figure 5(c) shows a whole mount section (Haematoxylin and Eosin [H and E] stain) through the same tissue sample, illustrating the fat component (corresponding to low velocity and high attenuation) and the muscle component (high velocity and lower attenuation) of this angiomyolipoma. It should be noted that some of the structure seen in the attenuation image (Fig. 5(b)), particularly the high attenuation near the boundaries between the fat and muscle components, is probably a phase cancellation artefact due to the different velocities of the two tissues (Bamber 1986). We attempted to avoid these boundary areas when selecting regions of interest for detailed analyses. Four regions of interest, shown in Fig 5(b), were selected for analyzing the frequency dependent attenuation and backscatter coefficients of the tissue. Regions 1 and 2 correspond to homogeneous regions in the fat component of the tumour, while 3 and 4 correspond to homogeneous regions in the muscle component. The measured average attenuation and backscatter coefficients are shown in Figs. 6(a) and

245

(b), respectively, plotted over the frequency range 3.5-7 MHz. As expected from the 5-MHz attenuation image, the attenuation coefficients measured in the fat regions are higher than those measured in the muscle regions (Fig. 6(a)). The fat and muscle components of this angiomyolipoma are similarly distinguishable in terms of their backscatter coefficients, with the fat regions showing higher backscatter coefficients than the muscle regions (Fig. 6(b)). The error bars shown in Figs. 6(a) and (b) correspond to the standard deviation of the coefficients measured at the 64 points in the region of interest. It is evident that there is considerable statistical variation in these data, especially in the backscatter coefficients measured in this sample of angiomyolipoma.

a

b ¸¸¸¸9::i¸

~ ~!i~.

C

Fig. 5. Macroscope images of the angiomyolipoma from Fig. 4 compared to a whole mount histology section through the same tissue sample. (a) Velocity image of angiomyolipoma [grayscale: black (1420 m/s) to white (1580 m/s)] showing distinct fat and muscle components. (b) Attenuation image [black (0 dB/mm) to white (1 dB/mm)] with four regions selected for analysis of frequency dependent attenuation and backscatter coefficients. (c) Histology section (H and E stain) through the angiomyolipoma, showing fat (F) and muscle (M) components.

246

Ultrasound in Medicine and Biology

10.2 b)

1.25 a)

~

L

1.00

.~ 0.75

1

2

~

/_.

10-3

8.~ 0.50 ~. 10-4

0.25 ~,

;

(~

;

10-53

Frequency(MHz)

92ST

Fig. 6. Frequency dependent attenuation (a) and backscatter (b) measured in the four regions of tissue shown in Fig. 5(b) (1 and 2 in fat component ofangiomyolipoma; 3 and 4 in muscle component).

The magnitude of these errors depends on the homogeneity of the tissue within a region of interest, and the large variation in this sample is probably due to the inheranfly inhomogeneous nature of the angiomylipoma tumour, as will be discussed in the next section of this paper. More homogeneous renal tisues (e.g., normal kidney) showed much less variation over a region of interest than angiomyolipoma. By obtaining these measurements from many tissue samples, we can establish regions over which the frequency dependent attenuation and backscatter coefficients are defined for each renal tissue type. This case of an angiomyolipoma illustrates the general approach we have taken in this study. More details on the angiomyolipomas studied will be given in the next section, including a discussion of the pathological properties of the tissue responsible for the measured ultrasound porperties.

Volume 15, Number 3, 1989

our patient population (mean age = 64, see Table 1), and the development of fibrosis of the medulla with a progressive loss of nephrons (Marchal et al. 1986). Similarly, we could detect no significant differences between the ultrasound properties ofperirenal fat and renal sinus fat, so regions from both fat types were included under the heading of renal fat. The velocity at 5 MHz, and the attenuation and backscatter coefficients over the range 3.5-7 MHz were measured for normal kidney and renal fat. The average velocity _+ the standard deviation for kidney (1571 + 5 m/s) and renal fat (1407 + 13 m/s) are shown in Fig. 7. The frequency dependent attenuation and backscatter coefficients are shown in Fig. 8. Figure 8(a) shows the attenuation coefficients of the kidney and renal fat samples, plotted as two regions. Each region is determined by an upper curve, which is the mean attenuation coefficient plus the standard deviation, and a lower curve, which is the mean attenuation coefficient minus the standard deviation. It can be seen that there is considerable overlap between the attenuation coefficients of kidney tissue and renal fat, although the fat generally has a higher attenuation coefficient than kidney, with the separation being more pronounced with increasing frequency. Figure 8(b) shows the backscatter coefficients of the kidney and renal fat samples, also as regions corresponding to the mean backscatter coefficient _+ the standard deviation. In this case the renal fat is well separated from the kidney tissue, with the fat having a higher backscatter coefficient.

1600

AML Muscle Kidney

RCC

RESULTS AND DISCUSSION

Normal data We identified 30 regions of normal kidney tissue from 8 patients, and 25 regions of (normal) renal fat from 6 patients (see Table 1 for details), and measured the ultrasound properties in these regions. These regions were generally removed from the tumour sites by several centimetres, and were representative of the renal parenchyma in these patients. The normal kidney tissue included both medulla and renal cortex, for which we found no significant differences in terms of the ultrasound properties of velocity, attenuation or backscatter. This can probably be explained as a consequence of the advanced age of

1500 t AML Fot

1400

~ Renol Fat

147T

Fig. 7. Summary of the ultrasound velocities measured in normal kidney, renal fat, angiomyolipoma (AML), renal cell carcinoma (RCC), and oncocytoma (ONC).

Ultrasonic characterization of selected renal tissues • D. H. TURNBULL et aL

1.0, O)

'

'

O-Z

b)

'

'

247

'

0.8

4 .,l> .~ 0.6

~

.~,

g

~ 0.4

~>

,

5

;

,o-% Frequency ( M H z )

,,~,

Fig. 8. S u m m a r y o f f r e q u e n c y d e p e n d e n t a t t e n u a t i o n (a) a n d backscatter (b) coefficients for samples o f n o r m a l k i d n e y a n d renal fat.

It should be noted that for each of velocity, attenuation, and backscatter, the mean and standard deviation were computed over the 30 regions of kidney, or the 25 regions of renal fat. In these, and in the renal tissues described later in the paper, analyses were performed over the available number of regions of the tissue type. Since each region was selected for its homogeneity, the measurement procedure extracted average values of the ultrasonic properties of the tissue type. By taking the mean and standard deviations of these average values from a number of regions of the same tissue type from different patients, we have attempted to determine the range and variation of these tissue properties over our patient population. Comparisons can be made between our results and those of past investigations of kidney tissue and fat. The velocity measured in kidney (1571 m/s) agrees well with the value of 1568 m/s obtained by Bowen et al. (1979) for canine kidney at 37°C. The attenuation measured for kidney at 5 MHz and 37°C is 0.22 (_+0.06) dB/mm, which is very close to the value 0.2 d B / m m reported for hog kidney by Le Croissette et al. (1979), but low compared to the values reported by Lele et al. (1976) (0.52 dB/mm at 5 MHz for bovine kidney) and Goss et al. (1979) (0.41 dB/mm at 4 MHz for bovine and feline kidney). The lower value of attenuation coefficient may be indicative for fewer tissue inhomogeneities in our kidney samples. This explanation is consistent with

the observation that the present value of attenuation at 7 MHz (0.29 dB/mm) is much closer to the absorption coefficient reported by Goss et al. (1979) (0.17 dB/mm at 7 MHz) than their attenuation coefficient (0.76 dB/mm at 7 MHz). The attenuation coefficient measured in renal fat at 5 MHz and 37°C is 0.30 dB/mm, which compares to 0.24 dB/mm measured by D'Astous and Foster (1986) in breast fat. The corresponding backscatter coefficient in renal fat (5.3 × 10-4 Sr - l . mm -~) is somewhat higher than their value of 2.0 X 10-4 Sr -~. mm -~ measured in breast fat at 5 MHz and 37°C. Some discussion of the frequency dependence of the attenuation and backscatter coefficients will be given later in the paper. These measurements of the ultrasound velocity, attenuation, and backscatter in normal renal tissues serve as useful baselines for comparison with the tumours described in this paper. In the remainder of this section, the data from three types of renal mass will be discussed.

Angiomyolipomas Two cases ofangiomyolipoma have been studied (patients 1 and 2 of Table 1). The first angiomyolipoma (patient l) has already been discussed in some detail in the Methods section, above. This tumour consisted of well-defined fat and muscle components (see Figs. 4(b) and 5). The second angiomyolipoma (patient 2) contained a large central hematoma which was confined to the muscle tissue in the tumour. As

248

Ultrasound in Medicine and Biology

Volume 15, Number 3, 1989 i0-1

1.0

b) g

0.8

~10-3

"7 0.6 10-4 9a ~a

~04

":'-:::..::. 10-5

0.2

0

.5

4

5

6

I0 -6 7 5 Frequency (MHz)

4

5

6

7 143T

Fig. 9. Summary of frequency dependent attenuation (a) and backscatter (b) coefficients for samples of angiomyolipoma fat and muscle, superimposed on normal renal tissue data. in the first case, regions could be distinguished consisting predominantly of fat or muscle tissue. We were able to identify and measure the ultrasound properties in 6 regions of angiomyolipoma fat and 8 regions of angiomyolipoma muscle tissue from the two tumours (see Table 1). The velocity measurements are included in Fig. 7, while the frequency dependent attenuation and backscatter coefficients are shown in Fig. 9. It can be seen that the fat component of angiomyolipoma has a velocity (1476 + 19 m/s) intermediate between that of normal kidney tissue and renal fat. The muscle component of angiomyolipoma has a velocity (1575 _ 7 m/s) similar to that of normal kidney tissue (see Fig. 7). Figure 9(a) shows the frequency dependent attenuation coefficients of the angiomyolipoma fat and muscle components superimposed on the data from normal kidney and renal fat. Figure 9(b) shows the frequency dependent backscatter coefficients of the two components of angiomyolipoma, also superimposed on the data from the normal renal tissues. The fat component of angiomyolipoma can be differentiated from the other tissues, having higher attenuation and backscatter coefficients. The muscle component of angiomyolipoma has a lower attenuation coefficient, similar to that measured in normal kidney tissue, and a backscatter coefficient intermediate between that of kidney and renal fat. To obtain insight into the characteristics of the angiomyolipoma responsible for the measured ultra-

sound properties, we undertook a more detailed pathological examination of the tissue. Figure 10 shows a micrograph of tissue taken from the fat component of the tumour. It is evident that the tissue is actually composed of a mixture of fat and smooth muscle. It seems likely that the acoustic impedance differences between the muscle (high acoustic impedance) and fat (low acoustic impedance) are responsible for the high scattering and the high attenuation properties of the angiomyolipoma. The high backscatter coefficient in the fatty tissues of angiomyolipoma explains the echogenic appearance of this tu-

Fig. 10. Histology section (H and E stain) from "fat" component of angiomyolipoma, showing a mixture of fat and smooth muscle.

Ultrasonic characterization of selected renal tissues • D. H. TURNBULL et al.

249

10-2

I.O,

b)

0.8

.q~

~ 10-3 ~~,.~Echogenic

o.6

~ i0-4 .~b

g o.4

~ 10-5

0.2

~ [

05

~ I

4

I

5

Kdiney I

6

7 IO-e3 4 Frequency(MHz) I

~ 5

6

7 ,',,r

Fig. 11. Summary of (a) frequency dependent attenuation and (b) backscatter coefficients for samples of renal cell carcinoma superimposed on normal renal tissue data. Data from a single region of echogenic renal cell carcinoma are shown separately.

mour. The mixture of fat (low velocity) and muscle (high velocity) tissues also explains the intermediate velocity measured in the fat components of angiomyolipoma. Renal cell carcinomas Six cases of renal cell carcinoma have been studied (patients 4, 5, 7, 8, 9, 10 of Table 1). With the exception of patient 9, the appearance of these renal cell carcinomas in clinical ultrasound images show an echogenicity similar to that of normal kidney. In all, 20 regions of renal cell carcinoma were identified, and the ultrasound properties in those regions of tissue were measured (see Table 1 for details). The velocity measurements are included in Fig. 7, while the frequency dependent attenuation and backscatter coefficients are shown in Fig. 11. The measured attenuation (Fig. 1 l(a)) and backscatter (Fig. 1 l(b)) coefficients of these renal cell carcinomas coincide closely with those of normal kidney tissue. The velocity measured in renal cell carcinoma (1569 ___8 m/s) was also very similar to that of normal kidney. Patient 9 was an interesting case of a small (1.5cm diameter) renal cell carcinoma with a well-defined, echogenic rim (see Fig. 12). The 5-MHz macroscope images of the ultrasound attenuation, velocity, and peak backscatter are shown in Fig. 13(a), (b), and (c), respectively. Figure 13(d) shows a whole

mount section (H and E stain) through the same tissue sample, illustrating two well-defined components in the tumour. It is seen that the echogenic rim of this tumour correlates with a region of high attenuation (Fig. 13(a)) and somewhat elevated backscatter (Fig. 13 (c)). Data from the rim of this tumour were placed in a separate category from the other regions of renal cell carcinoma. The frequency dependent attenuation and backscatter coefficients for a region in the echogcnic rim of the renal cell carcinoma from pa-

Fig. 12. Sagittal scan (3.5 MHz) demonstrating a small renal cell carcinoma with an ¢chogenic rim (arrow).

250

Ultrasound in Medicineand Biology a .

.

.

.

.

.

.

.

b

c

Volume 15, Number 3, 1989 none showed similar macrophage infiltration. This mechanism could account for the low incidence of high echogenicity in renal cell carcinomas reported by some investigators (Hartman et al. 1981). One somewhat confusing fact is that the velocity me~.sured in the echogenic rim of this tumour is similar to that measured in other renal cell carcinomas, and not significantly lower as might be expected in the presence of the high lipid content of the macrophages. Since the ultrasound velocity in macrophage cells is unknown, however, this must remain a puzzling mystery for the present.

Oncocytomas

Fig. 13. Macroscope images of the renal cell carcinoma from Fig. 11 compared to a whole mount histology section through the same tissue sample. (a) attenuation image [grayscale: black (0 dB/mm) to white (1 dB/mm)] showing high attenuation rim of tumour; (b) velocity image [black (1500 m/s) to white (1600 m/s)]; (c) peak backscatter image; (d) whole mount section (H and E stain) through tissue sample showing kidney (K) and tumour (RCC) with echogenic rim (E).

tient 9 are shown in Fig. 1 1 (labelled Echogenic RCC). This tissue had a very high attenuation coefficient (Fig. 1 l(a)) as well as an elevated backscatter coefficient (Fig. 11 (b)). The velocity measured in this region of tissue was 1564 m/s, which is in the range of the other renal cell carcinomas studied. A closer pathological examination of the tumour from patient 9 showed some very interesting features. This was found to be a papillary carcinoma, with localized regions of necrosis not correlated with either the echogenic rim or the centre of the tumour. However, correlating well with the echogenic rim was a region of infiltration by lipid containing macrophages. Figure 14 shows a micrograph demonstrating the structure of the tissue in the rim of the tumour. Since the macrophages have a high lipid content (with similar properties to fat tissue), we are led to postulate that the high attenuation and backscatter coefficients measured in the rim of this tumour may be due to high acoustic impedance differences between layers of papillary tumour cells and infiltrating maerophages, similar to the acoustic impedance differences between fat and smooth muscle tissues in angiomyolipoma. This hypothesis is indi:ectly supported by the finding that on reviewing the other five cases of renal cell carcinoma included in this study,

Two cases of a rare benign renal tumour, oncocytoma, were also studied (patients 3 and 6 of Table 1). We were able to identify 5 regions of oncocytoma, and the ultrasound properties were measured in these regions. The measured velocity (1573 + 4 m/s) is included in Fig. 7, and the frequency dependent attenuation and backscatter coefficients are shown in Fig. 15. These oncocytomas had a somewhat lower attenuation coefficient than normal kidney (Fig. 15(a)) and a backscatter coefficient very similar to that of kidney (Fig. 15(b)).

Functional form of attenuation and backscatter frequency dependence Since frequency dependent attenuation and backscatter coefficients are often reported in terms of their functional frequency dependence, we have fitted the data of Fig. 8, 9, 13, and 15 to curves of the form a(p) = al" v*, (attenuation),

Fig. 14. Histology section (H and E stain) of echogenic renal cell carcinoma from Fig. 11, showing carcinoma cells infiltrated by lipid containing macrophages (arrows).

U l t r a s o n i c c h a r a c t e r i z a t i o n o f selected r e n a l tissues • D. H . TURNBULL el al.

1.0

251

10-2

Q) 0.8 I0 -3

b)

Renal Fat~

10-4 ,o .~ 0.4

t

~

10-5 0.2,

...

..:::-

4 Frequency (MHz)

;

;

7 ,,~T

Fig. 15. Summary of frequency dependent attenuation (a) and backscatter (b) coefficients for samples of oncocytoma superimposed on normal renal tissue data. and pry) = m" vv, (backscatter). Here a I and p~ represent the corresponding coefficients at 1 MHz, and 3' represents the frequency dependence. The results of the curve fitting are given in Table 2 for all the tissues types studied. The difference between the given curves and the corresponding mean value of attenuation or backscatter is less than 20% over the entire frequency range in all cases. The frequency dependence of the backscatter coefficients of renal fat and the fatty components of angiomyolipoma are higher than the other tissues, so the use of higher frequency ultrasound should enhance the echogenic appearance of the tumour. The very low value of the frequency dependence of the backscatter coefficient in a single region of the echogenic renal

cell carcinoma from patient 9 is interesting in view of the search for a quantitative discriminator between renal cell carcinoma and angiomyolipoma. Comparisons of the frequency dependence of attenuation and backscatter in kidney and fat can be made to those reported by past investigators. The frequency dependence of attenuation in kidney (3' = 1.2) measured in the present study is slightly higher than the value 1.09 reported by Goss et al. (1979) for bovine and feline kidney at 37°C over a frequency range 0.5-7 MHz, and a linear frequency dependence reported by Lele et al. (1976) for bovine kidney from 5 to 9 MHz. The frequency dependence of attenuation in renal fat (7 = 1.5) agrees well with the value 1.7 reported by D'Astous and Foster (1986) for breast fat at 37°C over a frequency range of 3-7 MHz. The corresponding frequency dependence of backscatter in renal fat (3' = 2.4) is significantly higher than the value 1.9 reported in breast fat.

Table 2. Functional form of frequency dependence of attenuation and backscatter coefficients measured in renal tissues

Kidney R e n a l fat Angiomyolipoma: Muscle Fat R e n a l cell c a r c i n o m a E c h o g e n i c r e n a l cell c a r c i n o m a Oncocytoma

Attenuation

Backscatter

0 . 0 2 9 3 dB m m -I M H z -t'2 0 . 0 2 6 3 dB m m -I M H z -1"5

0 . 6 6 3 × 10 -6 Sr -~ m m -~ M H z L7 11.0 × 10 -6 Sr -~ m m -~ M H z -2"4

0 . 0 4 2 0 d B m m -t M H z - H 0 . 0 9 3 9 d B m m -t M H z L2 0.0431 d B m m -I M H z - I ' ° 0 . 1 1 3 d B m m -~ M H z -1"4 0 . 0 1 3 5 d B m m -I M H z - H

5.94 X 10 -6 Sr -I m m -l M a z -1"4 34.9 × 10 6 Sr x m m -~ M H z - z s 0 . 5 6 7 × 10 -6 Sr -~ m m -1 M H z - I s 28.0 × 10 -6 Sr -l m m -~ M n z -°'9 0 . 8 9 8 X 10 -6 Sr -I m m -I M H z -1"2

252

Ultrasound in Medicine and Biology

Table 3. Summary of renal tissue properties measured at 5 MHz, 37°C

Kidney Renalfat Angiomyolipoma: Muscle Fat Renal cell carcinoma Echogenic renal cell carcinoma Oncocytoma

Velocity [m/s]

Attenuation coefficient [dB/mm]

Backscatter coefficient [St -~ • m m -~] × 10-4

1571 +_ 5 1407_+ 13

0.22 + 0.060 0.30_+0.13

0.095 _+ 0.070 5.3_+5.0

1575 -+ 7 1476 -+ 19

0.24 -+ 0.060 0.68 _+ 0.20

0.69 -+ 0.41 20.0 _+ 7.3

1569 _+ 8

0.23 _+0.060

0.11 _+ 0.08

1564 1573 _+ 4

0.69 0.14 _+ 0.020

I. 1 0.061 + 0.030

CONCLUSIONS The ultrasound properties of normal renal tissues and three types of renal tumour have been measured at a temperature of 37°C, and over a frequency range of 3.5-7 MHz. Table 3 is included as a general summary of the renal tissue properties measured at 5 MHz, and 37°C. We have found the ultrasound properties of two angiomyolipomas, namely high attenuation and backscatter coefficients in the fatty components of the tumours, to agree with observations of high echogenicity of angiomyolipoma in clinical ultrasound images. Pathological examination of the fatty components of angiomyolipoma shows the tissue to be composed of a mixture of fat and smooth muscle. We have postulated that the resulting high acoustic impedance variations between these components are responsible for the high scattering properties of angiomyolipoma. Other investigators have shown that structural variations in tissue composition, and the existence of acoustic impedance differences caused by the presence of fat (low acoustic impedance) and collagen (high acoustic impedance) may be responsible for increased scattering properties in some tissues (Bamber et ai. 1981; O'Donnell et ai. 1981). In vitro demonstrations of the echogenic properties of water/fat and protein/fat interfaces have been made by Behan and Kazam (1978) and Davis et al. (1981). We believe the angiomyolipoma tumour to be an in vivo example of the same scattering mechanism in tissue. The renal cell carcinomas studied in this paper had ultrasound properties indistinguishable from the normal kidney studied, with one exception. In one case of renal cell carcinoma we found an echogenic

Volume 15, Number 3, 1989

rim having elevated attenuation and backscatter coefficients, and which correlated with an area of papillary carcinoma infiltrated by lipid containing macrophages. We are led to suggest that the possible acoustic impedance variations between the tumour cells and fatlike macrophages are responsible for the increased scattering observed in this tissue. In view of this observation of high scattering, and the corresponding echogenicity in a renal cell carcinoma, we should urge extreme caution in diagnosing angiomyolipoma based only on high tumour echogenicity. Similar cautions have been issued by other investigators (Hartman et al. 1981). It should be noted that if simultaneous in vivo measurements of backscatter and velocity could be made, then it would probably be possible to confidently discriminate between angiomyolipoma and renal cell carcinoma. The main conclusion to be drawn from this study is that fat/nonfat interfaces seem to be a dominant scattering mechanism in the renal tumours studied, giving rise to the high attenuation and backscatter coefficients measured in the angiomyolipomas and one renal cell carcinoma. Incorporation of simple acoustic impedance interfaces into existing scatter theories (see Sehgal and Greenleaf 1984) may have potential for the future of quantitative tissue scattering models. Acknowledgments--We wish to thank the National Cancer Institute of Canada for the financial support of this work. One of the authors (D.T.) expresses gratitude to the Medical Research Council of Canada for financial support. We are also grateful to Dr. J. Carruthers, Dr. P. Van Nostrand, and Dr. S. Ritchie for their expert help with pathology, and to K. Harasiewicz for technical assistance in resurrecting the macroscope system.

REFERENCES Bamber, J. C.; Hill, C. R., King, J. A. Acoustic properties of normal and cancerous human liver--II. Dependence on tissue structure. Ultrasound in Med. & Biol. 7:135-144; 1981. Bamber, J. C. Chapter 4: Attenuation and absorption. In: Hill, C. R., ed. Physical principles of medical ultrasonics. Chicester, UK: Ellis Horwood Ltd.; 1986:166-168. Behan, M.; Kazam, E. The echographic characteristics of fatty tissues and turnouts. Radiology 129:143-151i 1978. Bowen, T.; Connor, W. G.; Nasoni, R. L., Pifer, A. E.; Sholes, R. R. Measurement of temperature dependence of the velocity of ultrasound in soft tissues. In: Linzer, M., ed. Ultrasonic tissue characterization II. NBS Special Publication 525. Washington, D.C." US Govt. Printing Office, 1979:57-61. Bret, P. M.; Bretagnolle, M.; Galliard, D.; Plauchu, H., Labadie, M., Lapray, J. F.; Roullaud, Y.; Cooperberg, P. Small, asymptomatic angiomyolipomas of the kidney. Radiology 154:7-10, 1985. Davis, P. L.; Filly, R. A.; Goerke, J. In vitro demonstration of an echogenic emulsion: Relationship of lipid particle size to echo detection. J. Clin. Ultrasound 9:263-266, 1981. D'Astous, F. T.; Foster, F. S. Frequency dependence of ultrasound attenuation and backscatter in breast tissue. Ultrasound in Med. & Biol. 12(10):795-808; 1986.

Ultrasonic characterization of selected renal tissues • D. H. TURNBULLe! al. Foster, F. S.; Strban, M.; Austin, G. The ultrasonic macroscope: Initial studies of breast tissue. Ultrasonic Imaging 6:242-26 l; 1984. Goss, S. A.; Frizzell, L. A.; Dunn, F. Ultrasonic absorption and attenuation in mammalian tissues. Ultrasound in Med. & Biol. 5:181-186; 1979. Greenleaf, J. F., editor. Tissue characterization with ultrasound. Boca Raton, FL: CRC Press; 1986.2 vol. Hartman, D. S.; Goldman, S. M.; Friedman, A. C.; Davis, C. J.; Madewell, J. E.; Sherman, J. L. Angiomyolipoma: Ultrasonicpathologic correlation. Radiology 139:451-458; 1981. Le Croissctte, D. H.; Heyser, R. C.; Gammell, P. M.; Roseboro, J. A. The attenuation of selected soft tissue as a function of frequency. In: Linzer, M., ed. Ultrasonic tissue characterization II. NBS Special Publication 525. Washington, D.C.: US Govt. Printing Office; 1979:101-108. Lele, P. P., Mansfield, A. B.; Murphy, A. I.: Namery, J.; Senapati, N. Tissue characterization by ultrasonic frequency-dependent

253

attenuation and scattering. In: Linzer, M., ed. Ultrasonic Tissue Characterization, NBS Special Publication 453, Washington, D.C.: US Govt. Printing Office; 1976:167-196. Marchal, G.; Verbeken, E.; Oyen, R.; Moerman, F.; Baert, A. L.; Lauweryns, J. Ultrasound of the normal kidney: A sonographic, anatomic and histologic correlation. Ultrasound in Med. & Biol. 12(12):999-1009; 1986. Nicholas, D.; Hill, C. R.; Nassiri, D. K. Evaluation of backscattering coefficients for excised human tissues: Principles and techniques. Ultrasound in Med. & Biol. 8(1):7-15; 1982. O'DonneU, M.; Mimbs, J. W.; Miller, J. G. Relationship between collagen and ultrasonic backscatter in myocardial tissue. J. Acoust. Soc. Am. 69(2):580-588; 1981. Scheible, W.; Ellenbogen, P. H.; Leopold, G. R.; Siao, N. T. Lipomatus tumors of the kidney and adrenal: Apparent echographic specificity. Radiology 129:153-156, 1978. Sehgal, C. M.; Greenleaf, J. F. Scattering of ultrasound by tissues. Ultrasonic Imaging 6:60-80; 1984.