- Email: [email protected]
Ultrasonic doppler study of the hormonal response of blood flow in the normal human breast
Uhrasound in Med. & Blol Vol. 13, No. 3, pp. 121-129, 1987 Printed in the U.S.A.
0301-5629/87 $3.00 + .00 (c~ 1987 Pergamon Journals Ltd.
OOriginal Contribution ULTRASONIC OF
DOPPLER
BLOOD
FLOW
STUDY IN
THE
OF
THE
NORMAL
HORMONAL HUMAN
RESPONSE BREAST
M. SAMBROOK,J. C. BAMBER,H. MINASIANtand C. R. HILL Physics Department, Institute of Cancer Research: Royal Marsden Hospital, Downs Road, Sutton, Surrey, SM2 5PT, UK
(Received 28 Februao, 1986; in final form 3 September 1986) Abstract--The breasts of seven normal female volunteers were examined using a continuous wave, directional 10 MHz ultrasonic Doppler system. A range of quantitative features were extracted from recorded Doppler signals by first computing an average, single cardiac cycle sonogram from 4-6 overlayed cardiac cycles of sonogram data taken from each recording. Substantial variations were observed to occur in both frequency and amplitude characteristics of the Doppler signals during the menstrual cycle and pregnancy. For each subject the two breasts behaved similarly and the fluctuations correlated with known variations in blood hormone levels and breast surface temperature. In the one case of pregnancy, the mammary blood flow appeared to increase throughout pregnancy, beginning very shortly after conception. It is concluded that the normal fluctuations of the blood flow in the breast may make a large contribution to the variance of Doppler-derived blood flow features for the pre-menopausal breast. Use of the contralateral breast as a control is advocated for studies of the application of the Doppler method to the diagnosis and measurement of therapeutic response of breast cancer in young women. The usefulness of the contralateral breast as such a control might be enhanced by performing Doppler examinations only at about the midcycle. If the presence of a tumour were to alter these fluctuations there may be a possibility of using the effect to advantage alongside other methods for early diagnosis of breast cancer.
INTRODUCTION
White and Cledgett, 1978) and for measurement of therapeutic response (Minasian and Bamber, 1982; Bamber et al., 1983) in pre-menopausal patients. It was indeed noted in our previous studies that the features of Doppler signals obtained from blood flowing in the breast of one pre-menopausal patient were completely uncharacteristic of those from the rest of the patients, who were all post-menopausal. A knowledge of the natural variations in breast tumour blood flow might also be helpful for optimising the timing of breast surgery or treatment by other methods, as suggested by Thomlinson (1982) with regard to chemotherapy.
As part of a general study to further our understanding of the biology of breast cancer, and to assess the potential of Doppler techniques for studying tumour blood flow, a 10 MHz continuous wave ultrasonic Doppler system is being used to obtain information about the blood flow characteristics of the normal human breast and of malignant breast turnouts (Minasian and Bamber, 1982; Bamber et al., 1983). In this paper, we report the results of an attempt to document any changes in m a m m a r y blood flow which takes place during the menstrual cycle. The results of a study on a single pregnant volunteer are also presented. The work was aimed at improving our understanding of breast physiology and the manner in which hormonal changes influence the mammary blood flow, but it was also designed to provide reference data for the possible use of the Doppler method for diagnosis (Boyd et al., 1985; Burns et al., 1982; Wells et al., 1977;
M A T E R I A L S AND M E T H O D S
Stlbjc¢Is Nine normal female volunteers, who were not using oral contraceptives, were initially subjected to this study. One of these volunteers became pregnant and was found to have conceived at about the time, or just after, the first measurement. When it was discovered that tiffs subject was pregnant the possibility of discontinuing the study was suggested to her. She requested
t Breast Unit, Royal Marsden Hospital, London SW3, UK. Present address: Newham General Hospital, London E 13, UK. 121
122
Ultrasound in Medicine and Biology
that the study be continued so that the valuable data which we had already obtained, and were likely to obtain, would not be lost. Two of the studies terminated, for reasons personal to the subjects, before a useful time sequence of data could be gathered. Thus we are able to present results for six normal females throughout the menstrual cycle and one example throughout pregnancy.
Data acquisition The apparatus and method used to obtain and record Doppler signals from blood flowing in the breast has been described in detail previously (Minasian and Bamber, 1982). Briefly, the system consisted of a purpose built 10 MHz, continuous wave, direction resolving Doppler flowmeter, t a maximum frequency follower, headphones, and a high quality stereo tape recorder. Two transducers were used; both Parks,z~ nominally 10 MHz pencil probes. The same transducer was used for all measurements on any one volunteer. The main features of this system are that the flowmeter appears to possess excellent sensitivity and signal to noise ratio--necessary for the small blood vessels under study and the high attenuating medium of the breast. Direction of flow is resolved in this flowmeter using the method of phase-quadrature demodulation followed by frequency domain processing (Atkinson and Woodcock, 1982, p. 67)with a heterodyning frequency of 3.4 kHz. Flow in one direction results in a lower side-band signal and zero flow is represented at 3.4 kHz. Choice of which sideband contains the signal corresponding to the direction of predominant flow is allowed by means of a channel reversing switch. Use of the maximum frequency follower, in combination with listening to the signals, permits the position and orientation of the transducer to be adjusted so as to maximise the Doppler shift frequencies. A calibrated attenuator in place of the record level and balance controis of the tape recorder allows the relative signal amplitude also to be measured. During the early part of the study, volunteers were examined at intervals of approximately 3-4 days over at least two menstrual cycles. A preliminary analysis of the results from the first three subjects (Bamber et al., 1983) demonstrated that it would be more desirable to make measurements over at least three menstrual cycles and at intervals of 2-3 days. This latter scheme was used in the case of the remaining three menstrual cycle studies to be discussed in this report. The vol-
t Built by the department of Medical Physics, Bristol General Hospital, Bristol, UK. z~Parks Medical Electronics Incorp. Beaverton, Oregon, USA.
March 1987, Volume 13, Number 3
unteer who became pregnant was studied at intervals of 2-4 weeks from conception until the pregnancy was almost full term. Each examination was scheduled to occur at the same time of day for a given volunteer (the possibility of diurnal rhythms has not been investigated). A standarised examination posture with the patient lying relaxed in a supine position was adopted and the examination was rescheduled for another day if unusual physiological conditions existed for the subject on a particular day. (All measurements were conducted in the afternoon and the subjects were asked when they had last eaten, consumed alcohol, smoked tobacco, and undergone stressful exercise.) In this respect we usually found it necessary to disregard data from the first examination, as it was generally atypical of the trends followed by subsequent data. The first examination of each study served two purposes only; (a) to allow the volunteers to become familiar with the examination procedure and thereby remove any nervous tension which they might otherwise have had, and (b) to permit a search of both breasts using the Doppler instrumentation so that suitable sites for measurement could be noted and marked as described below. With the hand-held transducer applied to the skin using ultrasonic coupling jelly, and whilst listening to the signals and observing the output of the maximum frequency follower, a general survey of each breast was made; beginning at the nipple and working systematically out in circles of increasing radius. The sites chosen for examination were restricted to between three and six per breast by the requirements of not making more than one measurement per blood vessel and of staying clear of the three main supply arteries which enter from the axillary, medial and caudal margins of the breast. At each selected site the method used previously, for studying variations in tumour blood flow, was adopted (Minasian and Bamber, 1982). The essential features of this are as follows. About 10 cardiac cycles of Doppler-difference frequency signals were recorded, having first attempted to orient the transducer so as to maximise the Doppler shift. This procedure was adopted so as to aid repetition of the beam-vessel orientation on subsequent examination, although the approximate direction and inclination of the transducer was also noted with reference to the point of contact of the transducer with the skin and the tangent to the skin surface. The position of each measurement site was marked on a flexible transparent acetate foil (of the kind used when lecturing with an "overhead projector") and a hole was punched through the foil at these points so that the breast could be re-marked at the time of subsequent examinations using the foil
Blood flow in the normal human breast • M. SAMBROOKet al. overlaying the breast. Absolute relocation of the overlay was accomplished by also marking the positions of natural reference points on the skin, such as the nipple, moles and naevi. The oral temperature of each subject was also recorded at each examination and was found not to have varied by more than 1°C throughout each series of examination.
Data analysis The product of the n u m b e r of sites per examination, the n u m b e r of examinations per subject, and the number of subjects resulted in approximately 1200 recordings requiring analysis. Sonograms corresponding to 4 s of data selected from each of these recordings were produced using a tRadionics 8000 discrete Fourier transform, real-time audio spectrum analyser. This device is able (on a 0-5 kHz frequency range) to compute simultaneously (for two input channels) two 64 point spectra every 16.6 ms and displays the resulting two real-time sonograms by means o f a 4 bit-deep digital frame store with a serialised, analogue (standardvideo) output. The non-directional signal from the flowmeter was used for the analysis in this study, usually with the 64 points distributed over a 0-5 kHz frequency range. Due to the limited dynamic range of the displayed information, and to permit comparison of the signal amplitudes from different recordings, the input level control of the analyser was deactivated and replaced with a calibrated attenuator. In fact, it was rarely necessary to use this facility since, as described above, the signal levels had been approximately equalised at the time of recording. The sonogram data were then transferred to an inexpensive microcomputer,~ via a serial interface, for further analysis. A variety of both amplitude and frequency related features were extracted from the sonograms, averaged over all sites for a given examination of a volunteer and studied for any expression of deterministic fluctuations during the period of observation. Table 1 lists the features that were computed for each recording. Before these features were extracted, some initial data reduction and statistical averaging was performed by creating an average, single cycle, sonogram from the mean of the 4-5 consecutive cycles present on the frozen display. The reference point used to define the registration of each of the oveflayed cycles of sonogram data was the time at end diastole and was manually specified to the computer by means of the built-in display cursor on the spectrum analyser. The average time interval between these registration points
t Radionics Medical, Scarborough, Ontario, Canada. :~Sinclair Spectrum, Sinclair Research Ltd., Cambridge, U.K.
123
Table 1. Ultrasonic Doppler features of breast blood flow computed from averaged, single cycle, sonograms such as that shown in Fig. 1. Group
Feature
Symbol or Formula
1
Time averagedmean frequency tf ! Time averagedmaximum frequency ax)
2
Maximumfrequencyat peak systole Maximum frequencyat end diastole "Mean" frequency Range Ratio
3
Normalisedarea under the spectrum E "Mean" frequency (f Smax(f ) "f )/ (f f )
4
Pulsatilityindex Relative flow index
5
Heart rate (mean time between peak systole)
A B (A + B)/2 A- B A/B
(A - B)_/(f) (E). ( f )
The features have been divided into five groups: (1) those from the complete sonogram, (2) those computed from the maximum frequency envelope of the sonogram. (3) those from the maximum spectral envelope, (3) those from the complete sonogram, (4) combinations of parameters from the above groups, (5) and a miscellaneous category. ( ) around a quantity signifies averaging of the instantaneous value of that quantity over the cardiac cycle.
was used as a measure of the heart rate. Feature extraction was then performed on single cycle average sonograms, such as that illustrated in Figure lb. To appreciate the logic of the feature classification scheme used in Table 1, it is helpful to remember that a sonogram is essentially a three dimensional surface, S(f, t), where the instantaneous spectral amplitude (S) is a function of two coordinates; frequency ( f ) and time (t). At any instant in time there exists an instantaneous spectrum S ( f ) , which m a y be characterised by various features including the instantaneous mean, if), and maximum, (fm~x), frequencies. Averaging such quantities over all spectra within the cardiac cycle leads to features in group 1 of the table. Group 2 consists of features of the m a x i m u m frequency envelope,fm~x(t), which we have used previously (Minasian and Bamber, 1982) and should require no further explanation. By analogy to the m a x i m u m frequency envelope one can form the m a x i m u m amplitude spectral envelope, Smax(f), where Smaxis the maxi m u m spectral amplitude occurring over all spectra in the cardiac cycle at a fixed frequency. This curve may then be described by features such as those listed in group 3. These are also consistent with features used in our earlier work (Bamber et al., 1983), although previously, Smax(f) was obtained as the outline of the trace resulting from a long time-exposure photograph of an oscilloscope continuously displaying the instan-
24
Ultrasound in Medicine and Biology
March 1987, Volume 13, Number 3
(a)
(b)
Blood flow in the normal h u m a n breast • M. SAMBROOK et al.
taneous spectra. The mean frequency measured was an unweighted mean, (i.e., no frequency components were excluded from the calculation). Group 4 arises because some features are necessarily combinations of features from more than one of the groups defined above. The pulsatility index (Atkinson and Woodcock, 1982) is a quantity often used to infer differences in downstream vascular impedance. The "relative flow index" was defined by us (Bamber et al., 1983) in an attempt to arrive at a parameter which might more closely reflect changes in volume flow rate. Under conditions of a plane ultrasonic wave of uniform intensity across the diameter of a blood vessel the volume flow rate within that vessel is given by the product of the cross-sectional area of the vessel lumen and the mean flow velocity, the latter being proportional to the first m o m e n t of the Doppler power spectrum and the cosine of the angle between the direction of flow and the direction of sound propagation (Baker et al•, 1978).
L
0
121
A
0
.i
.m
., •.. • ,.•'.. '•" :
Many of the features extracted were correlated with one another, the correlations being particularly strong between those features which were frequencybased (r = 0.88 to 0.97). As noted previously, (Minasian and Bamber, 1982) measures of the pulsatile nature of the flow did not vary greatly between individuals or with time, and were thus found not to be useful for the present purposes. The mean value for the ratio of the instantaneous peak frequency to the instantaneous mean frequency was found to be roughly 1.3 _+ 0.2 (over the cardiac cycle and for all the volunteers). It therefore appears that the flow in the vessels under study in these normal breasts was well behaved with a profile across the vessel which was reasonably close to parabolic (Atkinson and Woodcock, 1982, p. 121). Before a combination of a frequency-based and amplitude based feature is utilised (e.g. in our relative flow index feature) it is, of course, important to know how well such features correlate with each other. Figure 2 demonstrates the observed relationship between the total instantaneous spectral power (averaged over the cardiac cycle) and the instantaneous mean frequency (also averaged over the cardiac cycle). On a linear scale
. '.,
, • :,t
• •
.:::'...:....--
E
"""
"'::"':
• .-~q:.:.':,:.,',.:
F-0 0
.
.. *."
.". : , . 5)".'>'::.' ': ". ~>... .:. ¢-~*'='~'i"
.:.,. :.
. -..
,
'"
..
..,
• :
~,.~._~.~,~.:cj,o....~ • '1" 1500
I
I
!
I
3 000
Time Averaged Mean Frequency(kHz)
RESULTS
Interrelationships between features
125
Fig. 2. Observed relationship between the total spectral power and the mean Doppler shift frequency for all of the data obtained. Both were computed as average values over the cardiac cycle•
no correlation is immediately apparent, but following a double logarithmic transformation a highly significant correlation coefficient is obtained r = 0.75 (the standard error of the gradient of the fitted curve P ( f ) _ 0 . 5 f 2"65
~ 10
is +3%). The technical or physiological sig-
nificance of this result is not understood at present, but it would appear that there may be some benefit from using these two features in combination. This will be illustrated later.
Variations during the menstrual cycle Given the above mentioned relationships between the different features it should be no surprise to learn that most of the features provided similar results in demonstrating that the blood flow in the pre-menopausal breast exhibits measurable cyclic fluctuations. For purposes of presenting the results in this paper the
Fig. 1. A single cycle (a), and an average (b) sonogram. In this example (b) resulted from averaging five cardiac cycles of displayed data. (a) is one of the original cycles. The display used is limited to 3 bits and does not do justice to the amplitude resolution available in the averaged data. Nevertheless, the smoothing of the random signal variations, produced by fluctuation scattering by the blood (Atkinson and Berry, 1974), is readily apparent• The gray scale linearly divides the spectral amplitude range, from zero to the maximum value, into eight classes defined by gray levels ranging from black to white.
126
Ultrasound in Medicine and Biology
examples which best demonstrate this point have been chosen. As a demonstration of site-to-site variation in both breasts within one volunteer, Fig. 3 shows how one Doppler feature, the time averaged mean frequency, varies with time during the menstrual cycle at all measured sites. From studying all of our data there was no significant difference in the blood flow of the two breasts of any one subject nor any tendency for specific sites within the breast to behave differently from other sites. The cyclic behaviour of Doppler features of breast blood flow is, of course, better demonstrated by the value of a feature averaged over all measurement sites for any one subject. Our standard data reduction procedure, adopted for each feature and for each volunteer, is therefore, to compute such a mean value at various times during the menstrual cycle. Some smoothing (with a 3 point moving average) of unexplained minor, and more random, fluctuations in the original data is also applied. In Fig. 4, which shows the mean of all data presented in Fig. 3, the error bars represent plus and minus one standard deviation of the results about the current mean value, after each value has been normalised, so that if the same normal•sing factor were applied to the previous points on the graph, they would all become coincident. Mathematically this is expressed as 1
2
N
---~(Fi+Fi_,-~_,-Pi)
(1)
2
1 i= 1
O'c°rrected- N-
where N is the number of measured sites, Fi is the value of the feature at the ith time sample, and a bar over a symbol indicates an average over N. This operation of shifting by an amount equal to the distance
1543 I
A//'•
/ ~
• : o n s e t of menstruation
1127
-kl
a =
onset of rua|lon
-r x,, 1D Unt-13 L ~o847 o~
>
,
~
7
i
I
17
i
i
/1
27
i
37
Time
I
47
i
I
57
~J
I
67
, 77
[days]
Fig. 4. As an example of the data reduction process described in the text this figure presents a condensed version of the data shown in Fig. 3. A 3 point moving average filter was applied to each curve corresponding to a measurement site and the mean value computed over all measurement sites. The vertical bars at each point on the graph represent plus and minus one normalised standard deviation computed according to Eq. 1 (see text).
of the previous point from the previous mean value results in a value for the standard deviation (+80 Hz) which is partially corrected for spatial variations of the features and should provide a more useful means of assessing the significance of the observed fluctuations with time. It may indeed be appropriate to apply a scaling factor to the data as well as a shift, in which case the present error bars would represent an overestimation of the differences between the behaviour with time of the features from each of the measurement sites. Figure 5 demonstrates, using the "relative flow index" as an example, the results from all six volunteers. To aid the clarity of the figure, particularly for comparison of one trace with another, the error bars have been omitted. Their relative magnitudes however, are similar to those illustrated in Fig. 4.
Variations during pregnancy
U
8~ 955 c
March 1987, Volume 13, Number 3
O
367
I
7
17
27
37
47
57
67
I 77
Time [days] Fig. 3. An example of how one feature, the time averaged mean Doppler shift frequency ((f)), varies with time at all measured sites on both breasts of a normal pre-menopausal subject.
Figure 6 demonstrates a continuous increase in blood flow velocity in the breast with gestational age (the abscissa) during a successful pregnancy. Note that the initial values obtained fall centrally in the range of values for the non-pregnant volunteers. Most of the increase appears to occur before the end of the second trimester of pregnancy, after which the graph tends towards a plateau. The rise in blood flow appears to manifest itself very early. Indeed, one might make the speculative observation that increases in blood flow, (measurable by this noninvasive method) occurred at a stage before it could be known that the volunteer was pregnant by normal indications and test methods. Fur-
•
Blood flowin the normal human breast M. SAMBROOK et al. A : onset
m tO
r"
E
0
'
210
J
DISCUSSION
of
z J0
~
' 60
2'0
8
z.'O
6'0
20
30
5
"t3
i 0
o ii
20
;0
60
80
100
0 0
i 10
4 > 0
o ,A,
0 0
20
LO
127
60
80
0
20
/.0
60
Time [days] Fig. 5. Examples of results from all six volunteer females, demonstrating the variation of one Doppler feature, the "relative flow index" (see Table 1), with time during the menstrual cycle. Typical relative magnitudes of site-to-site variations were similar to those shown in Fig. 4. • indicates the estimated time of onset of menstruation. When present, < indicates probable time of ovulation (see discussion).
ther studies are necessary to confirm this, however, since it is possible that the early rise in blood flow in this instance was coincidentally due to the normal premenstrual rise in breast blood flow seen in the women who did not become pregnant.
The menstrual cycle has been the subject of a considerable amount of research in recent years, with the result that the associated hormonal and morphological changes, and their variability within and between individual subjects, are now well documented. Objectives have varied, but many studies have been aimed at assessing ovarian function and increasing the level of understanding of the complicated interactions of hypothalmic hormones, gonadotrophic hormones, oestrogens and progesterone in their control of the menstrual cycle, pregnancy, and infertility. We have extracted just a few points from this literature, when and where we believe them to be relevant to our present study. Oestrogens and progesterone stimulate mammary development (Linzell, 1959); oestrogens causing duct extension and branching, whilst both oestrogens and progesterone cause the formation of alveoli. Our observation of a substantial rise in breast blood flow just before or around the onset of menstruation is consistent with the time at which there is a maximum in the total circulating levels of oestrogens, progesterone and prolactin (Kim el al., 1974). This is also the time of occurrence of the pre-menstrual breast "hypertrophy" and pain which a number of women experience (Cutler, 1961). Furthermore, Simpson et al. (1981) have observed raised breast surface temperature, also at this time in the menstrual cycle. It is thus likely that the hormonal status measurably influences the breast blood flow which fluctuates cyclicly with the same period as the menstrual cycle. One implication of this is that one might expect a particularly large variance for data derived from Doppler blood flow measurements on breast tumours in pre-menopausal patients at arbitrary times during the menstrual cycle. A question which therefore arises is, would it be possible to make corrections to
LEFT
30
N 2"5 /*-'---*
x ~
2.0
~/°X,
15 9~ oiLL
1.0 ~x / x / x ~ ^
RIGHT
~xJ*"---
x
0-51 0
5
1'0
15 ' Time
20
25 '
i
30
A
Lo
[weeks)
Fig. 6. Variation in one Doppler feature, the "mean" frequency (A + B)/2, with gestational age during pregnancy (both breasts of one volunteer).
128
Ultrasound in Medicine and Biology
such data based on a knowledge of the time of a measurement relative to the phase of an individual's own cycle? Treating one menstrual cycle as being divided into 360 ° with zero at the onset of menstruation, the cyclic maximum in our data occurs at about 330 ° (averaged over all six subjects). Unfortunately, the standard deviation of this point is about _+88° between individuals and may be as much as _+180 ° even within an individual. Such a level of what might be called "phase variability" is enough to produce data fluctuations of up to 60%. This would make corrections of the kind mentioned above difficult, if not impossible. The "variability" of the menstrual cycle is already well known, and manifests itself in substantial variations of the length of the cycle, and the length of each stage (follicular, luteal, etc.) in consecutive cycles, and the absolute serum hormone levels over consecutive cycles (Ganong, 1983). Generally, a minimum in the mammary blood flow was found to occur mid-cycle, (i.e. at about 180°). For one of the subjects studied (Fig. 5b) this minimum regularly coincided with the time at which the subject complained of pain in the right or left low iliac fossa; suggesting possible ovulation. (N.B. all volunteers were asked whether they felt breast pain and none of the volunteers had any knowledge of the data or results from the study until well after the series of measurements was complete.) If it is assumed that these pains were associated with the recent occurrence of ovulation then, although such pains can be associated with a fibrocystic condition of the breast, it appears that ovulation is accompanied by a reduction in blood flow to the breast (at least in this subject). Furthermore, in the case of this particular individual, about 4 to 5 days following minimum blood flow she complained of painful breasts. This pain continued and tended to increase throughout the period of increasing blood flow, subsiding completely once menstruation had occurred. Over this period, (the luteal phase) the breast size, assessed visually, would also increase markedly and return to a normal size after menstruation. We have no reason to suspect that these changes are necessarily abnormal. They probably represent a somewhat exaggerated version of what happens to breast blood flow in all normal females. It was demonstrated above that it is not likely to be possible to make corrections to data obtained from breast tumours in pre-menopausal patients. This emphasises the importance of using the contralateral breast as a control, as in the methods of previous studies (Burns et al., 1982; Minasian and Bamber, 1982). Another implication of our results occurs because of the existence of the minimum in all of the Doppler features of breast blood flow at 180 ° in the cycle. In studies
March 1987, Volume 13, Number 3 where one is attempting to make diagnostic use of a possible bilateral imbalance of some Doppler feature, due to the presence of a carcinoma in one of the breasts (Burns et al., 1982), the contralateral contrast that one is looking for (if it exists) may well be maximised by performing the examination as close to mid-cycle as possible. This, of course, assumes that the blood flow to the breast containing the tumour remains high, even through the follicular stage of the cycle. Although there is at present no evidence to support this suggestion, there does exist a similar rationale for thermographic examination of the breast, where cooling of the breasts is used in an attempt to improve the normal:abnormal thermographic contrast (Jones, et al., 1975). As a final, highly speculative consideration, it may be suggested that if the presence of, or impending development of, a tumour in a breast were to alter the cyclic pattern of breast blood flow there may be a way of using the effect directly for early diagnosis of breast cancer. However, the Doppler method in its present form is not particularly well suited to continuous monitoring and such a function is more likely to be better served by thermal methods, as suggested by Simpson et al. (1981). CONCLUSION It has been observed that the hormonal status can measurably influence ultrasonic Doppler features of the blood flowing in the normal pre-menopausal breast. This helps to establish further the credibility of our continuous wave Doppler technique as a method for quantitatively studying relative characteristics of breast blood flow (Minasian and Bamber, 1982). Measured characteristics of blood flow in the breasts of non-pregnant young women fluctuate cyclicly in a manner which correlates with other features of menstruation. In the one case of pregnancy encountered in this study the ultrasonic features suggested that blood flow increases throughout pregnancy, beginning sometime during the first week after conception. The magnitudes of the fluctuations observed are similar to previously documented increases in ultrasonic indices of breast blood flow due to the presence of primary carcinoma of the breast, and may well be larger than the variations which occur following endocrine therapy of such tumours. The variation from one person to another in the timing of these fluctuations, relative to the menstrual cycle, is sufficiently large to suggest that it would be very difficult to make corrections to individual data, based only on a knowledge of the time at which a measurement was made. This stresses the importance of using the contralateral breast as a control for studies of the application of the Doppler
Blood flow in the normal human breast • M. SAMBROOK el al.
method to the diagnosis and measurement of therapeutic response of breast cancer in pre-menopausal women. However, although it was observed that the Doppler features of both breasts behave similarly, it has yet to be shown that this remains to be so when one of the breasts contains a tumour. If, in fact, the presence of a tumour were to alter these fluctuations there may be a possibility of using the effect to advantage as a further aid for early diagnosis of breast cancer. Knowledge of this kind regarding the natural variations in breast blood flow should permit more rational planning; both of breast cancer therapy and of diagnostic or other studies on pre-menopausal subjects. By timing the therapy according to the appropriate phase of the menstrual cycle, one might be able to minimise bleeding during surgery or maximise tissue or tumour perfusion during hyperthermia, radiotherapy, or treatment by drugs. Acknowledgments--We wish to express our thanks to the volunteers for their patience and goodwill. The work was supported in part by grants from the Royal Marsden Hospital and the Cancer Research Campaign/Medical Research Council Joint Committee.
REFERENCES Atkinson P. and Berry M. V. (1974) Random noise in ultrasonic echoes diffracted by blood. J. Phys. AT, 1293-1302. Atkinson P. and Woodcock J. P. (1982) Doppler Ultrasound and its use in Clinical Measurement. Academic Press, London. Baker D. W., Forster F. K. and Daigle R. E. (1978) Doppler principles and techniques. In Ultrasound, Its Application in Medicine & Biology Part 1 (Edited by F. J. Fry), Chap. III, pp. 161-287. Elsevier, Amsterdam.
129
Bamber J. C., Sambrook M., Minasian H. and Hill C. R. (1983) Doppler study of blood flow in breast cancer. In Proceedings of the Third International Congress on the Ultrasonic Examination of the Breast (Edited by J. Jellins and T. Kobayashi). John Wiley & Sons, Chichester. Boyd J., Jellins J., KossoffG. and Reeve T. S. (1985) Combined Bmode and Doppler examinations of the breast. In Proceedings of the 4th Meeting World Fed. Ultrasound Med. Biol. (Edited by R. W. Gill and M. J. Dadd), p. 365. Pergamon Press, Sydney. Burns P. N., Halliwell M., Wells P. N. T. and Webb A. J. (1982) Ultrasonic Doppler studies of the breast. Ultrasound Med. BioL 8, 127-143. Cutler M. ( 1961 ) Tumours of the Breast. Pitman Medical Publishing Co. Ltd., London. Ganong W. F. (1983) Review of Medical Physiology, Chap. 23, pp. 354. Lange Medical Publications, CA. Jones C. H., Greening W. P., Davey J. B., McKinna J. A. and Greeves V. J. (1975) Thermography of the female breast. 5 year study in relation to detection and prognosis of cancer. Br. J. RadioL 48, 532-538. Kim M. H., Hosseinian A. H. and Dupon C. (1974) Plasma levels of oestrogens, androgens and progesterone during normal and dexamethasone-treated cycles. J. Clin. Endrocrinol. Metab. 39, 706-712. Linzell J. L. (July, 1959) Physiology of the mammary glands. Physiol. Rev. 39, 534-576. Minasian H. and Bamber J. C. (1982) A preliminary assessment of an ultrasonic Doppler method for the study of blood flow in human breast cancer. Ultrasound Med. Biol. 8, 357-364. Simpson H. W., Much F., Giles C., Grant J. K., Griffiths K., Halberg F., MacEwan J., McNicol A. M., Turbitt M. and Wilson D. (1981) A fresh approach to human breast cancer. In Commentaries on Research in Breast Disease 2 (Edited by R. D. Bulbrook and D. J. Taylor), pp. 133-164. Alan R. Liss Inc., New York. Thomlinson H. (1982) Measurement and management of carcinoma of the breast. Clinical Radiology 33, 481-493. Wells P. N. T., Halliwell M., Skidmore B., Webb A. J. and Woodcock J. P. (1977) Tumour detection by ultrasonic Doppler blood flow signals. Ultrasonics 15, 231-232. White D. N. and Cledgett P. R. (1978) Breast carcinoma detection by ultrasonic Doppler signals. UltrasoundMed. Biol. 4, 329-335.
0301-5629/87 $3.00 + .00 (c~ 1987 Pergamon Journals Ltd.
OOriginal Contribution ULTRASONIC OF
DOPPLER
BLOOD
FLOW
STUDY IN
THE
OF
THE
NORMAL
HORMONAL HUMAN
RESPONSE BREAST
M. SAMBROOK,J. C. BAMBER,H. MINASIANtand C. R. HILL Physics Department, Institute of Cancer Research: Royal Marsden Hospital, Downs Road, Sutton, Surrey, SM2 5PT, UK
(Received 28 Februao, 1986; in final form 3 September 1986) Abstract--The breasts of seven normal female volunteers were examined using a continuous wave, directional 10 MHz ultrasonic Doppler system. A range of quantitative features were extracted from recorded Doppler signals by first computing an average, single cardiac cycle sonogram from 4-6 overlayed cardiac cycles of sonogram data taken from each recording. Substantial variations were observed to occur in both frequency and amplitude characteristics of the Doppler signals during the menstrual cycle and pregnancy. For each subject the two breasts behaved similarly and the fluctuations correlated with known variations in blood hormone levels and breast surface temperature. In the one case of pregnancy, the mammary blood flow appeared to increase throughout pregnancy, beginning very shortly after conception. It is concluded that the normal fluctuations of the blood flow in the breast may make a large contribution to the variance of Doppler-derived blood flow features for the pre-menopausal breast. Use of the contralateral breast as a control is advocated for studies of the application of the Doppler method to the diagnosis and measurement of therapeutic response of breast cancer in young women. The usefulness of the contralateral breast as such a control might be enhanced by performing Doppler examinations only at about the midcycle. If the presence of a tumour were to alter these fluctuations there may be a possibility of using the effect to advantage alongside other methods for early diagnosis of breast cancer.
INTRODUCTION
White and Cledgett, 1978) and for measurement of therapeutic response (Minasian and Bamber, 1982; Bamber et al., 1983) in pre-menopausal patients. It was indeed noted in our previous studies that the features of Doppler signals obtained from blood flowing in the breast of one pre-menopausal patient were completely uncharacteristic of those from the rest of the patients, who were all post-menopausal. A knowledge of the natural variations in breast tumour blood flow might also be helpful for optimising the timing of breast surgery or treatment by other methods, as suggested by Thomlinson (1982) with regard to chemotherapy.
As part of a general study to further our understanding of the biology of breast cancer, and to assess the potential of Doppler techniques for studying tumour blood flow, a 10 MHz continuous wave ultrasonic Doppler system is being used to obtain information about the blood flow characteristics of the normal human breast and of malignant breast turnouts (Minasian and Bamber, 1982; Bamber et al., 1983). In this paper, we report the results of an attempt to document any changes in m a m m a r y blood flow which takes place during the menstrual cycle. The results of a study on a single pregnant volunteer are also presented. The work was aimed at improving our understanding of breast physiology and the manner in which hormonal changes influence the mammary blood flow, but it was also designed to provide reference data for the possible use of the Doppler method for diagnosis (Boyd et al., 1985; Burns et al., 1982; Wells et al., 1977;
M A T E R I A L S AND M E T H O D S
Stlbjc¢Is Nine normal female volunteers, who were not using oral contraceptives, were initially subjected to this study. One of these volunteers became pregnant and was found to have conceived at about the time, or just after, the first measurement. When it was discovered that tiffs subject was pregnant the possibility of discontinuing the study was suggested to her. She requested
t Breast Unit, Royal Marsden Hospital, London SW3, UK. Present address: Newham General Hospital, London E 13, UK. 121
122
Ultrasound in Medicine and Biology
that the study be continued so that the valuable data which we had already obtained, and were likely to obtain, would not be lost. Two of the studies terminated, for reasons personal to the subjects, before a useful time sequence of data could be gathered. Thus we are able to present results for six normal females throughout the menstrual cycle and one example throughout pregnancy.
Data acquisition The apparatus and method used to obtain and record Doppler signals from blood flowing in the breast has been described in detail previously (Minasian and Bamber, 1982). Briefly, the system consisted of a purpose built 10 MHz, continuous wave, direction resolving Doppler flowmeter, t a maximum frequency follower, headphones, and a high quality stereo tape recorder. Two transducers were used; both Parks,z~ nominally 10 MHz pencil probes. The same transducer was used for all measurements on any one volunteer. The main features of this system are that the flowmeter appears to possess excellent sensitivity and signal to noise ratio--necessary for the small blood vessels under study and the high attenuating medium of the breast. Direction of flow is resolved in this flowmeter using the method of phase-quadrature demodulation followed by frequency domain processing (Atkinson and Woodcock, 1982, p. 67)with a heterodyning frequency of 3.4 kHz. Flow in one direction results in a lower side-band signal and zero flow is represented at 3.4 kHz. Choice of which sideband contains the signal corresponding to the direction of predominant flow is allowed by means of a channel reversing switch. Use of the maximum frequency follower, in combination with listening to the signals, permits the position and orientation of the transducer to be adjusted so as to maximise the Doppler shift frequencies. A calibrated attenuator in place of the record level and balance controis of the tape recorder allows the relative signal amplitude also to be measured. During the early part of the study, volunteers were examined at intervals of approximately 3-4 days over at least two menstrual cycles. A preliminary analysis of the results from the first three subjects (Bamber et al., 1983) demonstrated that it would be more desirable to make measurements over at least three menstrual cycles and at intervals of 2-3 days. This latter scheme was used in the case of the remaining three menstrual cycle studies to be discussed in this report. The vol-
t Built by the department of Medical Physics, Bristol General Hospital, Bristol, UK. z~Parks Medical Electronics Incorp. Beaverton, Oregon, USA.
March 1987, Volume 13, Number 3
unteer who became pregnant was studied at intervals of 2-4 weeks from conception until the pregnancy was almost full term. Each examination was scheduled to occur at the same time of day for a given volunteer (the possibility of diurnal rhythms has not been investigated). A standarised examination posture with the patient lying relaxed in a supine position was adopted and the examination was rescheduled for another day if unusual physiological conditions existed for the subject on a particular day. (All measurements were conducted in the afternoon and the subjects were asked when they had last eaten, consumed alcohol, smoked tobacco, and undergone stressful exercise.) In this respect we usually found it necessary to disregard data from the first examination, as it was generally atypical of the trends followed by subsequent data. The first examination of each study served two purposes only; (a) to allow the volunteers to become familiar with the examination procedure and thereby remove any nervous tension which they might otherwise have had, and (b) to permit a search of both breasts using the Doppler instrumentation so that suitable sites for measurement could be noted and marked as described below. With the hand-held transducer applied to the skin using ultrasonic coupling jelly, and whilst listening to the signals and observing the output of the maximum frequency follower, a general survey of each breast was made; beginning at the nipple and working systematically out in circles of increasing radius. The sites chosen for examination were restricted to between three and six per breast by the requirements of not making more than one measurement per blood vessel and of staying clear of the three main supply arteries which enter from the axillary, medial and caudal margins of the breast. At each selected site the method used previously, for studying variations in tumour blood flow, was adopted (Minasian and Bamber, 1982). The essential features of this are as follows. About 10 cardiac cycles of Doppler-difference frequency signals were recorded, having first attempted to orient the transducer so as to maximise the Doppler shift. This procedure was adopted so as to aid repetition of the beam-vessel orientation on subsequent examination, although the approximate direction and inclination of the transducer was also noted with reference to the point of contact of the transducer with the skin and the tangent to the skin surface. The position of each measurement site was marked on a flexible transparent acetate foil (of the kind used when lecturing with an "overhead projector") and a hole was punched through the foil at these points so that the breast could be re-marked at the time of subsequent examinations using the foil
Blood flow in the normal human breast • M. SAMBROOKet al. overlaying the breast. Absolute relocation of the overlay was accomplished by also marking the positions of natural reference points on the skin, such as the nipple, moles and naevi. The oral temperature of each subject was also recorded at each examination and was found not to have varied by more than 1°C throughout each series of examination.
Data analysis The product of the n u m b e r of sites per examination, the n u m b e r of examinations per subject, and the number of subjects resulted in approximately 1200 recordings requiring analysis. Sonograms corresponding to 4 s of data selected from each of these recordings were produced using a tRadionics 8000 discrete Fourier transform, real-time audio spectrum analyser. This device is able (on a 0-5 kHz frequency range) to compute simultaneously (for two input channels) two 64 point spectra every 16.6 ms and displays the resulting two real-time sonograms by means o f a 4 bit-deep digital frame store with a serialised, analogue (standardvideo) output. The non-directional signal from the flowmeter was used for the analysis in this study, usually with the 64 points distributed over a 0-5 kHz frequency range. Due to the limited dynamic range of the displayed information, and to permit comparison of the signal amplitudes from different recordings, the input level control of the analyser was deactivated and replaced with a calibrated attenuator. In fact, it was rarely necessary to use this facility since, as described above, the signal levels had been approximately equalised at the time of recording. The sonogram data were then transferred to an inexpensive microcomputer,~ via a serial interface, for further analysis. A variety of both amplitude and frequency related features were extracted from the sonograms, averaged over all sites for a given examination of a volunteer and studied for any expression of deterministic fluctuations during the period of observation. Table 1 lists the features that were computed for each recording. Before these features were extracted, some initial data reduction and statistical averaging was performed by creating an average, single cycle, sonogram from the mean of the 4-5 consecutive cycles present on the frozen display. The reference point used to define the registration of each of the oveflayed cycles of sonogram data was the time at end diastole and was manually specified to the computer by means of the built-in display cursor on the spectrum analyser. The average time interval between these registration points
t Radionics Medical, Scarborough, Ontario, Canada. :~Sinclair Spectrum, Sinclair Research Ltd., Cambridge, U.K.
123
Table 1. Ultrasonic Doppler features of breast blood flow computed from averaged, single cycle, sonograms such as that shown in Fig. 1. Group
Feature
Symbol or Formula
1
Time averagedmean frequency tf ! Time averagedmaximum frequency ax)
2
Maximumfrequencyat peak systole Maximum frequencyat end diastole "Mean" frequency Range Ratio
3
Normalisedarea under the spectrum E "Mean" frequency (f Smax(f ) "f )/ (f f )
4
Pulsatilityindex Relative flow index
5
Heart rate (mean time between peak systole)
A B (A + B)/2 A- B A/B
(A - B)_/(f) (E). ( f )
The features have been divided into five groups: (1) those from the complete sonogram, (2) those computed from the maximum frequency envelope of the sonogram. (3) those from the maximum spectral envelope, (3) those from the complete sonogram, (4) combinations of parameters from the above groups, (5) and a miscellaneous category. ( ) around a quantity signifies averaging of the instantaneous value of that quantity over the cardiac cycle.
was used as a measure of the heart rate. Feature extraction was then performed on single cycle average sonograms, such as that illustrated in Figure lb. To appreciate the logic of the feature classification scheme used in Table 1, it is helpful to remember that a sonogram is essentially a three dimensional surface, S(f, t), where the instantaneous spectral amplitude (S) is a function of two coordinates; frequency ( f ) and time (t). At any instant in time there exists an instantaneous spectrum S ( f ) , which m a y be characterised by various features including the instantaneous mean, if), and maximum, (fm~x), frequencies. Averaging such quantities over all spectra within the cardiac cycle leads to features in group 1 of the table. Group 2 consists of features of the m a x i m u m frequency envelope,fm~x(t), which we have used previously (Minasian and Bamber, 1982) and should require no further explanation. By analogy to the m a x i m u m frequency envelope one can form the m a x i m u m amplitude spectral envelope, Smax(f), where Smaxis the maxi m u m spectral amplitude occurring over all spectra in the cardiac cycle at a fixed frequency. This curve may then be described by features such as those listed in group 3. These are also consistent with features used in our earlier work (Bamber et al., 1983), although previously, Smax(f) was obtained as the outline of the trace resulting from a long time-exposure photograph of an oscilloscope continuously displaying the instan-
24
Ultrasound in Medicine and Biology
March 1987, Volume 13, Number 3
(a)
(b)
Blood flow in the normal h u m a n breast • M. SAMBROOK et al.
taneous spectra. The mean frequency measured was an unweighted mean, (i.e., no frequency components were excluded from the calculation). Group 4 arises because some features are necessarily combinations of features from more than one of the groups defined above. The pulsatility index (Atkinson and Woodcock, 1982) is a quantity often used to infer differences in downstream vascular impedance. The "relative flow index" was defined by us (Bamber et al., 1983) in an attempt to arrive at a parameter which might more closely reflect changes in volume flow rate. Under conditions of a plane ultrasonic wave of uniform intensity across the diameter of a blood vessel the volume flow rate within that vessel is given by the product of the cross-sectional area of the vessel lumen and the mean flow velocity, the latter being proportional to the first m o m e n t of the Doppler power spectrum and the cosine of the angle between the direction of flow and the direction of sound propagation (Baker et al•, 1978).
L
0
121
A
0
.i
.m
., •.. • ,.•'.. '•" :
Many of the features extracted were correlated with one another, the correlations being particularly strong between those features which were frequencybased (r = 0.88 to 0.97). As noted previously, (Minasian and Bamber, 1982) measures of the pulsatile nature of the flow did not vary greatly between individuals or with time, and were thus found not to be useful for the present purposes. The mean value for the ratio of the instantaneous peak frequency to the instantaneous mean frequency was found to be roughly 1.3 _+ 0.2 (over the cardiac cycle and for all the volunteers). It therefore appears that the flow in the vessels under study in these normal breasts was well behaved with a profile across the vessel which was reasonably close to parabolic (Atkinson and Woodcock, 1982, p. 121). Before a combination of a frequency-based and amplitude based feature is utilised (e.g. in our relative flow index feature) it is, of course, important to know how well such features correlate with each other. Figure 2 demonstrates the observed relationship between the total instantaneous spectral power (averaged over the cardiac cycle) and the instantaneous mean frequency (also averaged over the cardiac cycle). On a linear scale
. '.,
, • :,t
• •
.:::'...:....--
E
"""
"'::"':
• .-~q:.:.':,:.,',.:
F-0 0
.
.. *."
.". : , . 5)".'>'::.' ': ". ~>... .:. ¢-~*'='~'i"
.:.,. :.
. -..
,
'"
..
..,
• :
~,.~._~.~,~.:cj,o....~ • '1" 1500
I
I
!
I
3 000
Time Averaged Mean Frequency(kHz)
RESULTS
Interrelationships between features
125
Fig. 2. Observed relationship between the total spectral power and the mean Doppler shift frequency for all of the data obtained. Both were computed as average values over the cardiac cycle•
no correlation is immediately apparent, but following a double logarithmic transformation a highly significant correlation coefficient is obtained r = 0.75 (the standard error of the gradient of the fitted curve P ( f ) _ 0 . 5 f 2"65
~ 10
is +3%). The technical or physiological sig-
nificance of this result is not understood at present, but it would appear that there may be some benefit from using these two features in combination. This will be illustrated later.
Variations during the menstrual cycle Given the above mentioned relationships between the different features it should be no surprise to learn that most of the features provided similar results in demonstrating that the blood flow in the pre-menopausal breast exhibits measurable cyclic fluctuations. For purposes of presenting the results in this paper the
Fig. 1. A single cycle (a), and an average (b) sonogram. In this example (b) resulted from averaging five cardiac cycles of displayed data. (a) is one of the original cycles. The display used is limited to 3 bits and does not do justice to the amplitude resolution available in the averaged data. Nevertheless, the smoothing of the random signal variations, produced by fluctuation scattering by the blood (Atkinson and Berry, 1974), is readily apparent• The gray scale linearly divides the spectral amplitude range, from zero to the maximum value, into eight classes defined by gray levels ranging from black to white.
126
Ultrasound in Medicine and Biology
examples which best demonstrate this point have been chosen. As a demonstration of site-to-site variation in both breasts within one volunteer, Fig. 3 shows how one Doppler feature, the time averaged mean frequency, varies with time during the menstrual cycle at all measured sites. From studying all of our data there was no significant difference in the blood flow of the two breasts of any one subject nor any tendency for specific sites within the breast to behave differently from other sites. The cyclic behaviour of Doppler features of breast blood flow is, of course, better demonstrated by the value of a feature averaged over all measurement sites for any one subject. Our standard data reduction procedure, adopted for each feature and for each volunteer, is therefore, to compute such a mean value at various times during the menstrual cycle. Some smoothing (with a 3 point moving average) of unexplained minor, and more random, fluctuations in the original data is also applied. In Fig. 4, which shows the mean of all data presented in Fig. 3, the error bars represent plus and minus one standard deviation of the results about the current mean value, after each value has been normalised, so that if the same normal•sing factor were applied to the previous points on the graph, they would all become coincident. Mathematically this is expressed as 1
2
N
---~(Fi+Fi_,-~_,-Pi)
(1)
2
1 i= 1
O'c°rrected- N-
where N is the number of measured sites, Fi is the value of the feature at the ith time sample, and a bar over a symbol indicates an average over N. This operation of shifting by an amount equal to the distance
1543 I
A//'•
/ ~
• : o n s e t of menstruation
1127
-kl
a =
onset of rua|lon
-r x,, 1D Unt-13 L ~o847 o~
>
,
~
7
i
I
17
i
i
/1
27
i
37
Time
I
47
i
I
57
~J
I
67
, 77
[days]
Fig. 4. As an example of the data reduction process described in the text this figure presents a condensed version of the data shown in Fig. 3. A 3 point moving average filter was applied to each curve corresponding to a measurement site and the mean value computed over all measurement sites. The vertical bars at each point on the graph represent plus and minus one normalised standard deviation computed according to Eq. 1 (see text).
of the previous point from the previous mean value results in a value for the standard deviation (+80 Hz) which is partially corrected for spatial variations of the features and should provide a more useful means of assessing the significance of the observed fluctuations with time. It may indeed be appropriate to apply a scaling factor to the data as well as a shift, in which case the present error bars would represent an overestimation of the differences between the behaviour with time of the features from each of the measurement sites. Figure 5 demonstrates, using the "relative flow index" as an example, the results from all six volunteers. To aid the clarity of the figure, particularly for comparison of one trace with another, the error bars have been omitted. Their relative magnitudes however, are similar to those illustrated in Fig. 4.
Variations during pregnancy
U
8~ 955 c
March 1987, Volume 13, Number 3
O
367
I
7
17
27
37
47
57
67
I 77
Time [days] Fig. 3. An example of how one feature, the time averaged mean Doppler shift frequency ((f)), varies with time at all measured sites on both breasts of a normal pre-menopausal subject.
Figure 6 demonstrates a continuous increase in blood flow velocity in the breast with gestational age (the abscissa) during a successful pregnancy. Note that the initial values obtained fall centrally in the range of values for the non-pregnant volunteers. Most of the increase appears to occur before the end of the second trimester of pregnancy, after which the graph tends towards a plateau. The rise in blood flow appears to manifest itself very early. Indeed, one might make the speculative observation that increases in blood flow, (measurable by this noninvasive method) occurred at a stage before it could be known that the volunteer was pregnant by normal indications and test methods. Fur-
•
Blood flowin the normal human breast M. SAMBROOK et al. A : onset
m tO
r"
E
0
'
210
J
DISCUSSION
of
z J0
~
' 60
2'0
8
z.'O
6'0
20
30
5
"t3
i 0
o ii
20
;0
60
80
100
0 0
i 10
4 > 0
o ,A,
0 0
20
LO
127
60
80
0
20
/.0
60
Time [days] Fig. 5. Examples of results from all six volunteer females, demonstrating the variation of one Doppler feature, the "relative flow index" (see Table 1), with time during the menstrual cycle. Typical relative magnitudes of site-to-site variations were similar to those shown in Fig. 4. • indicates the estimated time of onset of menstruation. When present, < indicates probable time of ovulation (see discussion).
ther studies are necessary to confirm this, however, since it is possible that the early rise in blood flow in this instance was coincidentally due to the normal premenstrual rise in breast blood flow seen in the women who did not become pregnant.
The menstrual cycle has been the subject of a considerable amount of research in recent years, with the result that the associated hormonal and morphological changes, and their variability within and between individual subjects, are now well documented. Objectives have varied, but many studies have been aimed at assessing ovarian function and increasing the level of understanding of the complicated interactions of hypothalmic hormones, gonadotrophic hormones, oestrogens and progesterone in their control of the menstrual cycle, pregnancy, and infertility. We have extracted just a few points from this literature, when and where we believe them to be relevant to our present study. Oestrogens and progesterone stimulate mammary development (Linzell, 1959); oestrogens causing duct extension and branching, whilst both oestrogens and progesterone cause the formation of alveoli. Our observation of a substantial rise in breast blood flow just before or around the onset of menstruation is consistent with the time at which there is a maximum in the total circulating levels of oestrogens, progesterone and prolactin (Kim el al., 1974). This is also the time of occurrence of the pre-menstrual breast "hypertrophy" and pain which a number of women experience (Cutler, 1961). Furthermore, Simpson et al. (1981) have observed raised breast surface temperature, also at this time in the menstrual cycle. It is thus likely that the hormonal status measurably influences the breast blood flow which fluctuates cyclicly with the same period as the menstrual cycle. One implication of this is that one might expect a particularly large variance for data derived from Doppler blood flow measurements on breast tumours in pre-menopausal patients at arbitrary times during the menstrual cycle. A question which therefore arises is, would it be possible to make corrections to
LEFT
30
N 2"5 /*-'---*
x ~
2.0
~/°X,
15 9~ oiLL
1.0 ~x / x / x ~ ^
RIGHT
~xJ*"---
x
0-51 0
5
1'0
15 ' Time
20
25 '
i
30
A
Lo
[weeks)
Fig. 6. Variation in one Doppler feature, the "mean" frequency (A + B)/2, with gestational age during pregnancy (both breasts of one volunteer).
128
Ultrasound in Medicine and Biology
such data based on a knowledge of the time of a measurement relative to the phase of an individual's own cycle? Treating one menstrual cycle as being divided into 360 ° with zero at the onset of menstruation, the cyclic maximum in our data occurs at about 330 ° (averaged over all six subjects). Unfortunately, the standard deviation of this point is about _+88° between individuals and may be as much as _+180 ° even within an individual. Such a level of what might be called "phase variability" is enough to produce data fluctuations of up to 60%. This would make corrections of the kind mentioned above difficult, if not impossible. The "variability" of the menstrual cycle is already well known, and manifests itself in substantial variations of the length of the cycle, and the length of each stage (follicular, luteal, etc.) in consecutive cycles, and the absolute serum hormone levels over consecutive cycles (Ganong, 1983). Generally, a minimum in the mammary blood flow was found to occur mid-cycle, (i.e. at about 180°). For one of the subjects studied (Fig. 5b) this minimum regularly coincided with the time at which the subject complained of pain in the right or left low iliac fossa; suggesting possible ovulation. (N.B. all volunteers were asked whether they felt breast pain and none of the volunteers had any knowledge of the data or results from the study until well after the series of measurements was complete.) If it is assumed that these pains were associated with the recent occurrence of ovulation then, although such pains can be associated with a fibrocystic condition of the breast, it appears that ovulation is accompanied by a reduction in blood flow to the breast (at least in this subject). Furthermore, in the case of this particular individual, about 4 to 5 days following minimum blood flow she complained of painful breasts. This pain continued and tended to increase throughout the period of increasing blood flow, subsiding completely once menstruation had occurred. Over this period, (the luteal phase) the breast size, assessed visually, would also increase markedly and return to a normal size after menstruation. We have no reason to suspect that these changes are necessarily abnormal. They probably represent a somewhat exaggerated version of what happens to breast blood flow in all normal females. It was demonstrated above that it is not likely to be possible to make corrections to data obtained from breast tumours in pre-menopausal patients. This emphasises the importance of using the contralateral breast as a control, as in the methods of previous studies (Burns et al., 1982; Minasian and Bamber, 1982). Another implication of our results occurs because of the existence of the minimum in all of the Doppler features of breast blood flow at 180 ° in the cycle. In studies
March 1987, Volume 13, Number 3 where one is attempting to make diagnostic use of a possible bilateral imbalance of some Doppler feature, due to the presence of a carcinoma in one of the breasts (Burns et al., 1982), the contralateral contrast that one is looking for (if it exists) may well be maximised by performing the examination as close to mid-cycle as possible. This, of course, assumes that the blood flow to the breast containing the tumour remains high, even through the follicular stage of the cycle. Although there is at present no evidence to support this suggestion, there does exist a similar rationale for thermographic examination of the breast, where cooling of the breasts is used in an attempt to improve the normal:abnormal thermographic contrast (Jones, et al., 1975). As a final, highly speculative consideration, it may be suggested that if the presence of, or impending development of, a tumour in a breast were to alter the cyclic pattern of breast blood flow there may be a way of using the effect directly for early diagnosis of breast cancer. However, the Doppler method in its present form is not particularly well suited to continuous monitoring and such a function is more likely to be better served by thermal methods, as suggested by Simpson et al. (1981). CONCLUSION It has been observed that the hormonal status can measurably influence ultrasonic Doppler features of the blood flowing in the normal pre-menopausal breast. This helps to establish further the credibility of our continuous wave Doppler technique as a method for quantitatively studying relative characteristics of breast blood flow (Minasian and Bamber, 1982). Measured characteristics of blood flow in the breasts of non-pregnant young women fluctuate cyclicly in a manner which correlates with other features of menstruation. In the one case of pregnancy encountered in this study the ultrasonic features suggested that blood flow increases throughout pregnancy, beginning sometime during the first week after conception. The magnitudes of the fluctuations observed are similar to previously documented increases in ultrasonic indices of breast blood flow due to the presence of primary carcinoma of the breast, and may well be larger than the variations which occur following endocrine therapy of such tumours. The variation from one person to another in the timing of these fluctuations, relative to the menstrual cycle, is sufficiently large to suggest that it would be very difficult to make corrections to individual data, based only on a knowledge of the time at which a measurement was made. This stresses the importance of using the contralateral breast as a control for studies of the application of the Doppler
Blood flow in the normal human breast • M. SAMBROOK el al.
method to the diagnosis and measurement of therapeutic response of breast cancer in pre-menopausal women. However, although it was observed that the Doppler features of both breasts behave similarly, it has yet to be shown that this remains to be so when one of the breasts contains a tumour. If, in fact, the presence of a tumour were to alter these fluctuations there may be a possibility of using the effect to advantage as a further aid for early diagnosis of breast cancer. Knowledge of this kind regarding the natural variations in breast blood flow should permit more rational planning; both of breast cancer therapy and of diagnostic or other studies on pre-menopausal subjects. By timing the therapy according to the appropriate phase of the menstrual cycle, one might be able to minimise bleeding during surgery or maximise tissue or tumour perfusion during hyperthermia, radiotherapy, or treatment by drugs. Acknowledgments--We wish to express our thanks to the volunteers for their patience and goodwill. The work was supported in part by grants from the Royal Marsden Hospital and the Cancer Research Campaign/Medical Research Council Joint Committee.
REFERENCES Atkinson P. and Berry M. V. (1974) Random noise in ultrasonic echoes diffracted by blood. J. Phys. AT, 1293-1302. Atkinson P. and Woodcock J. P. (1982) Doppler Ultrasound and its use in Clinical Measurement. Academic Press, London. Baker D. W., Forster F. K. and Daigle R. E. (1978) Doppler principles and techniques. In Ultrasound, Its Application in Medicine & Biology Part 1 (Edited by F. J. Fry), Chap. III, pp. 161-287. Elsevier, Amsterdam.
129
Bamber J. C., Sambrook M., Minasian H. and Hill C. R. (1983) Doppler study of blood flow in breast cancer. In Proceedings of the Third International Congress on the Ultrasonic Examination of the Breast (Edited by J. Jellins and T. Kobayashi). John Wiley & Sons, Chichester. Boyd J., Jellins J., KossoffG. and Reeve T. S. (1985) Combined Bmode and Doppler examinations of the breast. In Proceedings of the 4th Meeting World Fed. Ultrasound Med. Biol. (Edited by R. W. Gill and M. J. Dadd), p. 365. Pergamon Press, Sydney. Burns P. N., Halliwell M., Wells P. N. T. and Webb A. J. (1982) Ultrasonic Doppler studies of the breast. Ultrasound Med. BioL 8, 127-143. Cutler M. ( 1961 ) Tumours of the Breast. Pitman Medical Publishing Co. Ltd., London. Ganong W. F. (1983) Review of Medical Physiology, Chap. 23, pp. 354. Lange Medical Publications, CA. Jones C. H., Greening W. P., Davey J. B., McKinna J. A. and Greeves V. J. (1975) Thermography of the female breast. 5 year study in relation to detection and prognosis of cancer. Br. J. RadioL 48, 532-538. Kim M. H., Hosseinian A. H. and Dupon C. (1974) Plasma levels of oestrogens, androgens and progesterone during normal and dexamethasone-treated cycles. J. Clin. Endrocrinol. Metab. 39, 706-712. Linzell J. L. (July, 1959) Physiology of the mammary glands. Physiol. Rev. 39, 534-576. Minasian H. and Bamber J. C. (1982) A preliminary assessment of an ultrasonic Doppler method for the study of blood flow in human breast cancer. Ultrasound Med. Biol. 8, 357-364. Simpson H. W., Much F., Giles C., Grant J. K., Griffiths K., Halberg F., MacEwan J., McNicol A. M., Turbitt M. and Wilson D. (1981) A fresh approach to human breast cancer. In Commentaries on Research in Breast Disease 2 (Edited by R. D. Bulbrook and D. J. Taylor), pp. 133-164. Alan R. Liss Inc., New York. Thomlinson H. (1982) Measurement and management of carcinoma of the breast. Clinical Radiology 33, 481-493. Wells P. N. T., Halliwell M., Skidmore B., Webb A. J. and Woodcock J. P. (1977) Tumour detection by ultrasonic Doppler blood flow signals. Ultrasonics 15, 231-232. White D. N. and Cledgett P. R. (1978) Breast carcinoma detection by ultrasonic Doppler signals. UltrasoundMed. Biol. 4, 329-335.