Britijh .lournal of Obstetrics arid Gynaecologv July 1990, Vol 97, pp 595-602
Cyclic changes in composition and volume of the breast during the menstrual cycle, measured by magnetic resonance imaging P. A . FOWLER, C. H. KNIGHT
C. E. CASEY, G. G. C A M E R O N ,
M. A . FOSTER,
Summary. The volumes and spin-lattice (T,) relaxation timer of breast tissues and parenchymal water content were measured non-inva\ively by magnetic resonance imaging (MRI) in eight healthy women during four to eight consecutive menstrual cycles. Total breast volume, and parenchymal volume, TI relaxation time and water content were lowest between days 6 and 15. Between days 16 and 28, parenchymal volume, T, relaxation time and water content rose sharply by 38.9”/:,,15.1% and 24.5%. respectively, and peaked after day 2.5. Within 5 days of the oncet of menses, parenchymal volume fell sharply by 3@3?0, while watcr content declined by 17.5%. Rising parenchymal volume in the second half of the menstrual cycle is not solely due to increased tissue water content and provides in vivo evidence for both growth and increaqed tissue fluid at this time.
Changes in qize, histology and activity of thc breast during the menstrual cycle in adult women have dttracted considerable clinical attention (McCance et al. 1937; Ferguson & Anderbon 1981a, h ; Vogel et a1 1981; Anderson ef al 1982: Going at u l 1988). ‘I’echniques for volume estimation, including water displacement (Hytten 1951; Milligan rf ul. 1975) m d
University of Aberdeen. Foresterhill, Aherdeen AH9 2ZD Department of Obstetrics and Gynaecolog? P A I-OWLEK Research Fellow Department ot Medicine and rhWdpCutiCS C b CASEY Lectriier Department of Bio-Medical Phy\ics G G CAMERON Rewurch Assistant M A FOSTER Senior Lecturer Hannah Research Institute, Ayr KA6 SHL C H KNlCiHT Senior Sczerifrfic Offrcer
Conespondence P A Fowler
plaster casting (Ingleby 1949), provide no information about changes within the hreast, while techniques such as mammography and ultrasound do not readily provide volumetric data. Detailed studies of conlposition or activity require invasive techniques such as the excision of healthy tissue during surgery (Potten et nl. 1988). Magnetic resonance imaging (MRI) has been used to characterize breast tissues, including studies of spin-lattice (TI) relaxation during the menstrual cycle (Nelsonetul. 1985;Marlin & Yousef 1986). The T, relaxation time is a measure of the rate of energy loss from tissues following NMR excitation. This cnergy loss is dependent upon the biophysical environmcnt of the excited protons. A short T,, t h e r e h e . indicates the presence of lipids and/or organic structures which can bind cellular water tightly. A longer TI is expected from tissues of greater hydration and with the greatest amount of ccllular water. Volumetric studies of the breast have been made using MRI (Foster & Knight 1983), and the technique can also be employed
595
596
P. A . Fowler et al.
to estimate non-invasively the water content of nun-fatty tissues (MacLennan et al. 1986). The aim of the present pilot study was to investigate MRI-detectable patterns of change, including alteration in parenchymal water content, within the breast during the mcnstrual cycle. Differences duc to parity and the use of oral contraceptives were also examined. Subjects and methods
Eight healthy women, mean age 34.4 (SD 7.6) years, took part in the study. The average bodyweight was 56.7 (SD 8.2) kg. mean height wab 1.62 (SD 0.03) m, and the average BMI (bodymass index = kg/m2) was 21.7 (SD 2.6). Four women were nulliparous; three of the four parous women had breast-Led their children Lor 3-9 months, but not in thc 12-month period preceding the study. Two nulliparous and two parous women were taking oral contraceptives. Thcsc contraccptives wcre: (a) Logynon (Schcring) which gives a progesterone (levonorgestrel: 5CL125 pg/day) dose rising during the cycle and highest oestrogcn just aftcr mid-cycle (ethinyloestradiol: 40 pgiday on days 1418, otherwise 30 pgiday), and (b) Norinyl-1 (Syntex) which gives constant oestrogeniprogestcrone doses (mestranol 50 pg/day, norethisterone 1 mgiday). In both cases no dosage is taken on days 1-7 of the menstrual cycle. The study was approved by the Grampian Health Board Joint Ethical Committcc and informcd conscnt was obtained from all subjects.
Water displacernent. I n three women breast volumes were measured by water displacement on three, six and six occasions, using a tcchnique based on that described by Milligan etal. (1975). These measurements were made within 411 of MRI measurements.
MRI methods The subjects were imagcd using the Aberdeen 3.4 MHz (0-08 Tesla) NMR proton imager. Three imaging pulse sequences were used for the following three separate types of measurement.
Volume measurements. Volumes of both breasts combined were calculated from serial images using an inversion recovery (IR) pulse sequcnce (repetition time, Tr = 370 ms, interval, Ti = 170 ms). This pulse sequence gives a contrast of 1:2.8
between muscle and fat 5ignal intensities (Fowler et al. 1988) and image collection time is 47s Serial I R images were collected through the chests of supine subjects, from anterior to posterior limits of the brcasts, at 2-cm intcrvals. The relaxation time of parenchyma was determined in a subject before the start of the study. Bottles ranging between 164 and 1190 ml in volume. and containing CuSO, solutions, of similar T, relaxation to parcnchyma, were uscd to estimate the error in MRI calculation of volumes. The bottle5 were imaged at 2-cm intervals.
Estirnatiotz of water content. A single saturation recovery (SR) image (Tr = 1500 ms) was collccted through thc middle of the breasts of each woman during each imaging session. The women had an 80:20 watcridcuterium tcst object on their chests during image collection. The Tr of this pulse sequence was five times longer than parcnchymal T, relaxation timc. enabling the collection of images which had signal intensities proportional to the water content of non-fatty tissues (MacLennan et ( I / . 1986). Since dcuterium does not contribute to the NMR signal at proton frequencies, it can be used to dilute water to obtain tissue-equivalent concentrations. Wdter/dcutcrium test objects, with wateridcuterium ratios of 60:40, 70:30, 80:20, 90:lO and 100:0, were imaged at the start of each imaging session, providing a standard curve for NMRsignal intensity of the SK images against water content. All SR image signal intensity values wcre expressed as a ratio of the 80:20 waterideuterium tcst object signal intensities to compensate for day-to-day variation in imager performance. In order to obtain a correction factor between MRI estimated and actual watcr contents of lean tissues, 5 kg of lean muscle, wrapped around an 8 W 0 waterideuterium test object, was imaged 15 times using the SR pulse sequence. Subsequently 35 strips of 5 g each were cut from the muscle and dried to constant weight at 60°C, the difference between initial and final weights providing the water content of the muscle. The water content of muscle is around 70-80%, conipared to parenchymal water content of 70.6 (SD 20.6)% (Soman et al. 1970). Since emphasis in the present study is placed on the way in which water content changes, rather than specific values for water content, any crrors in estimating water content are therefore compensated for when subjects and stages of the mcnstrual cycle are compared.
Cyclical breast clianges T, relaxation lime mea.wrement. A T, rclaxation time map of the breast was calculated from an interleaved SlUIK image (Tr = 1000 ms, Ti = 200 mb collcctcd through the middle of the breasts of each woman on each occasion. T, relaxation time values for fat and parenchyma were then measured from specified regions of interest.
Image analysis. The images werc analyscd with a mini-computer and in-house raster graphics etnulating display system. On the I R images (Fig. l a ) fatty tissues were white while lean tissues were dark. Automatic and interactive image analysis techniques were combincd to obtain quantitative information, including area (cm2), image intensity or T, relaxation time, from regions of intercst. The term ’parenchyma’ was used for all non-fatty tissues of the breast, including lobular and non-fatty stroma. On each SR image (Fig. lb) two areas (2 cm’) of parenchymal tissue in each breast were outlined and SR image intensity recorded. From each map (Fig. lc) two areas of fatty and parenchymal tissues (each 2 cm’) werc used to determine the in-
597
vivo relaxation times of fat and parenchyma. Using a simple truncated cone model (Kvist etal. 1987), volumes were extrapolated from the measured areas of test objects or tissues in consecutive 1R images. Thc error of volume calculation from MRI images was determined by subtracting MRI-calculated from the known volumes of the test objects. and expressing the product as a percentage of the known volumes. The repeatability oP the methods used was determined by analysing the same set of images four times, calculating volumes and exprcssing the result as a Percentage deviatioo from the mean volume.
Statistical analysk The menstrual cycle was divided into three periods of roughly equal number of days (days 1-10, 11-20, 21-28: day 1 = onset of menses) and differences due to subjects and day5 in the raw data werc analysed by two-way analysis of variance (two-way ANOVA). The number of day-groups was litnited to three to maximize the number of replicates in each group. To examine
(d) Right breast
Lung
I
Heart Pare yc h y rn a
other leall tissues
canal
Fig. 1. Examples of the three MRI image typcs used: (a) IR image for volume calculation, (b) SR image for parenchymal water content estimation, (c) 1,image for measurement of fatty tissue and parcnchymal T, relaxation times; (d) drawing showing main features visible in the images. to. Test object; ma, motion artifact; blurring of heart and surrounding tissues due to heart beat.
598
A. Fowler et ul.
?I
changes during the menstrual cycle in more detail the menstrual cycle was divided into 6-day periods (days 1-5, 6-10, 11-15, 16-20, 21-25; day 26 onwards). Differences between the number 01measurements during each day-group in each woman werc compensated for by taking the mean of each variable, during each daygroup, for each woman. Therefore each subject
1200
;.
E
-
Results
a,
c
(u
? Ur
800
-
400
-
Between nine and 18 measurements (mean 12, SD 3) were obtained from each woman during . C X consecutive menstrual cycles (mean 5, SD I ) with an average of three measurements made during each cycle. The individual imaging timc for each subject ranged between 20 and 30 min.
5
0
ld
I
1200
800 Volume (rnl)
400
0
-s
130
I
0 ._ c m
120
;110
c ._
5
100
t.
S
._ G)
m
._
contributed only one data point for cach variable in each day-group. To compensate for the considerable inter-cycle variation and dilferences between individuals, measurements werc expressed as a percentage of the values measured during menses for each subject. Intcrcycle variability was calculated from irnages collected on the same day during successive menstrual cycles in cach subject. Thc volumcs in the first cycle were expressed as a percentage of volumes in the second cycle. Differences between volumcs, water content, o r T, relaxation during different day-groups were tested by paired t-test. The effects of parity or use of oral contraceptives were tested by two-way analysis of variance (two-way ANOVA). Relations between variablcs were tested by least-squares linear regression and correlation coefficient ( r j . MRl and water displacement techniques lor estimating total breast volume wcrc compared using the method of Bland & Altnian (1986).
90 80
60
70
80
90
100
Culibrution o,f MRI meusurenwnts The error of volume calculation from IR images was -5.7 (SD 1.3)%, indicating slight underestimation of volume calculation (Fig. 2a). while rcpcatability, expressed as a percentage of mean volume, was 4.3 (SD 2.7)%. Mean breast volumes as calculated hy MKI (641.3, SLI 239.8 ml) were significantly higher (paired t-test I = 2.93. P
Cyclic changes in composition and volume of the breast during the menstrual cycle, measured by magnetic resonance imaging P. A . FOWLER, C. H. KNIGHT
C. E. CASEY, G. G. C A M E R O N ,
M. A . FOSTER,
Summary. The volumes and spin-lattice (T,) relaxation timer of breast tissues and parenchymal water content were measured non-inva\ively by magnetic resonance imaging (MRI) in eight healthy women during four to eight consecutive menstrual cycles. Total breast volume, and parenchymal volume, TI relaxation time and water content were lowest between days 6 and 15. Between days 16 and 28, parenchymal volume, T, relaxation time and water content rose sharply by 38.9”/:,,15.1% and 24.5%. respectively, and peaked after day 2.5. Within 5 days of the oncet of menses, parenchymal volume fell sharply by 3@3?0, while watcr content declined by 17.5%. Rising parenchymal volume in the second half of the menstrual cycle is not solely due to increased tissue water content and provides in vivo evidence for both growth and increaqed tissue fluid at this time.
Changes in qize, histology and activity of thc breast during the menstrual cycle in adult women have dttracted considerable clinical attention (McCance et al. 1937; Ferguson & Anderbon 1981a, h ; Vogel et a1 1981; Anderson ef al 1982: Going at u l 1988). ‘I’echniques for volume estimation, including water displacement (Hytten 1951; Milligan rf ul. 1975) m d
University of Aberdeen. Foresterhill, Aherdeen AH9 2ZD Department of Obstetrics and Gynaecolog? P A I-OWLEK Research Fellow Department ot Medicine and rhWdpCutiCS C b CASEY Lectriier Department of Bio-Medical Phy\ics G G CAMERON Rewurch Assistant M A FOSTER Senior Lecturer Hannah Research Institute, Ayr KA6 SHL C H KNlCiHT Senior Sczerifrfic Offrcer
Conespondence P A Fowler
plaster casting (Ingleby 1949), provide no information about changes within the hreast, while techniques such as mammography and ultrasound do not readily provide volumetric data. Detailed studies of conlposition or activity require invasive techniques such as the excision of healthy tissue during surgery (Potten et nl. 1988). Magnetic resonance imaging (MRI) has been used to characterize breast tissues, including studies of spin-lattice (TI) relaxation during the menstrual cycle (Nelsonetul. 1985;Marlin & Yousef 1986). The T, relaxation time is a measure of the rate of energy loss from tissues following NMR excitation. This cnergy loss is dependent upon the biophysical environmcnt of the excited protons. A short T,, t h e r e h e . indicates the presence of lipids and/or organic structures which can bind cellular water tightly. A longer TI is expected from tissues of greater hydration and with the greatest amount of ccllular water. Volumetric studies of the breast have been made using MRI (Foster & Knight 1983), and the technique can also be employed
595
596
P. A . Fowler et al.
to estimate non-invasively the water content of nun-fatty tissues (MacLennan et al. 1986). The aim of the present pilot study was to investigate MRI-detectable patterns of change, including alteration in parenchymal water content, within the breast during the mcnstrual cycle. Differences duc to parity and the use of oral contraceptives were also examined. Subjects and methods
Eight healthy women, mean age 34.4 (SD 7.6) years, took part in the study. The average bodyweight was 56.7 (SD 8.2) kg. mean height wab 1.62 (SD 0.03) m, and the average BMI (bodymass index = kg/m2) was 21.7 (SD 2.6). Four women were nulliparous; three of the four parous women had breast-Led their children Lor 3-9 months, but not in thc 12-month period preceding the study. Two nulliparous and two parous women were taking oral contraceptives. Thcsc contraccptives wcre: (a) Logynon (Schcring) which gives a progesterone (levonorgestrel: 5CL125 pg/day) dose rising during the cycle and highest oestrogcn just aftcr mid-cycle (ethinyloestradiol: 40 pgiday on days 1418, otherwise 30 pgiday), and (b) Norinyl-1 (Syntex) which gives constant oestrogeniprogestcrone doses (mestranol 50 pg/day, norethisterone 1 mgiday). In both cases no dosage is taken on days 1-7 of the menstrual cycle. The study was approved by the Grampian Health Board Joint Ethical Committcc and informcd conscnt was obtained from all subjects.
Water displacernent. I n three women breast volumes were measured by water displacement on three, six and six occasions, using a tcchnique based on that described by Milligan etal. (1975). These measurements were made within 411 of MRI measurements.
MRI methods The subjects were imagcd using the Aberdeen 3.4 MHz (0-08 Tesla) NMR proton imager. Three imaging pulse sequences were used for the following three separate types of measurement.
Volume measurements. Volumes of both breasts combined were calculated from serial images using an inversion recovery (IR) pulse sequcnce (repetition time, Tr = 370 ms, interval, Ti = 170 ms). This pulse sequence gives a contrast of 1:2.8
between muscle and fat 5ignal intensities (Fowler et al. 1988) and image collection time is 47s Serial I R images were collected through the chests of supine subjects, from anterior to posterior limits of the brcasts, at 2-cm intcrvals. The relaxation time of parenchyma was determined in a subject before the start of the study. Bottles ranging between 164 and 1190 ml in volume. and containing CuSO, solutions, of similar T, relaxation to parcnchyma, were uscd to estimate the error in MRI calculation of volumes. The bottle5 were imaged at 2-cm intervals.
Estirnatiotz of water content. A single saturation recovery (SR) image (Tr = 1500 ms) was collccted through thc middle of the breasts of each woman during each imaging session. The women had an 80:20 watcridcuterium tcst object on their chests during image collection. The Tr of this pulse sequence was five times longer than parcnchymal T, relaxation timc. enabling the collection of images which had signal intensities proportional to the water content of non-fatty tissues (MacLennan et ( I / . 1986). Since dcuterium does not contribute to the NMR signal at proton frequencies, it can be used to dilute water to obtain tissue-equivalent concentrations. Wdter/dcutcrium test objects, with wateridcuterium ratios of 60:40, 70:30, 80:20, 90:lO and 100:0, were imaged at the start of each imaging session, providing a standard curve for NMRsignal intensity of the SK images against water content. All SR image signal intensity values wcre expressed as a ratio of the 80:20 waterideuterium tcst object signal intensities to compensate for day-to-day variation in imager performance. In order to obtain a correction factor between MRI estimated and actual watcr contents of lean tissues, 5 kg of lean muscle, wrapped around an 8 W 0 waterideuterium test object, was imaged 15 times using the SR pulse sequence. Subsequently 35 strips of 5 g each were cut from the muscle and dried to constant weight at 60°C, the difference between initial and final weights providing the water content of the muscle. The water content of muscle is around 70-80%, conipared to parenchymal water content of 70.6 (SD 20.6)% (Soman et al. 1970). Since emphasis in the present study is placed on the way in which water content changes, rather than specific values for water content, any crrors in estimating water content are therefore compensated for when subjects and stages of the mcnstrual cycle are compared.
Cyclical breast clianges T, relaxation lime mea.wrement. A T, rclaxation time map of the breast was calculated from an interleaved SlUIK image (Tr = 1000 ms, Ti = 200 mb collcctcd through the middle of the breasts of each woman on each occasion. T, relaxation time values for fat and parenchyma were then measured from specified regions of interest.
Image analysis. The images werc analyscd with a mini-computer and in-house raster graphics etnulating display system. On the I R images (Fig. l a ) fatty tissues were white while lean tissues were dark. Automatic and interactive image analysis techniques were combincd to obtain quantitative information, including area (cm2), image intensity or T, relaxation time, from regions of intercst. The term ’parenchyma’ was used for all non-fatty tissues of the breast, including lobular and non-fatty stroma. On each SR image (Fig. lb) two areas (2 cm’) of parenchymal tissue in each breast were outlined and SR image intensity recorded. From each map (Fig. lc) two areas of fatty and parenchymal tissues (each 2 cm’) werc used to determine the in-
597
vivo relaxation times of fat and parenchyma. Using a simple truncated cone model (Kvist etal. 1987), volumes were extrapolated from the measured areas of test objects or tissues in consecutive 1R images. Thc error of volume calculation from MRI images was determined by subtracting MRI-calculated from the known volumes of the test objects. and expressing the product as a percentage of the known volumes. The repeatability oP the methods used was determined by analysing the same set of images four times, calculating volumes and exprcssing the result as a Percentage deviatioo from the mean volume.
Statistical analysk The menstrual cycle was divided into three periods of roughly equal number of days (days 1-10, 11-20, 21-28: day 1 = onset of menses) and differences due to subjects and day5 in the raw data werc analysed by two-way analysis of variance (two-way ANOVA). The number of day-groups was litnited to three to maximize the number of replicates in each group. To examine
(d) Right breast
Lung
I
Heart Pare yc h y rn a
other leall tissues
canal
Fig. 1. Examples of the three MRI image typcs used: (a) IR image for volume calculation, (b) SR image for parenchymal water content estimation, (c) 1,image for measurement of fatty tissue and parcnchymal T, relaxation times; (d) drawing showing main features visible in the images. to. Test object; ma, motion artifact; blurring of heart and surrounding tissues due to heart beat.
598
A. Fowler et ul.
?I
changes during the menstrual cycle in more detail the menstrual cycle was divided into 6-day periods (days 1-5, 6-10, 11-15, 16-20, 21-25; day 26 onwards). Differences between the number 01measurements during each day-group in each woman werc compensated for by taking the mean of each variable, during each daygroup, for each woman. Therefore each subject
1200
;.
E
-
Results
a,
c
(u
? Ur
800
-
400
-
Between nine and 18 measurements (mean 12, SD 3) were obtained from each woman during . C X consecutive menstrual cycles (mean 5, SD I ) with an average of three measurements made during each cycle. The individual imaging timc for each subject ranged between 20 and 30 min.
5
0
ld
I
1200
800 Volume (rnl)
400
0
-s
130
I
0 ._ c m
120
;110
c ._
5
100
t.
S
._ G)
m
._
contributed only one data point for cach variable in each day-group. To compensate for the considerable inter-cycle variation and dilferences between individuals, measurements werc expressed as a percentage of the values measured during menses for each subject. Intcrcycle variability was calculated from irnages collected on the same day during successive menstrual cycles in cach subject. Thc volumcs in the first cycle were expressed as a percentage of volumes in the second cycle. Differences between volumcs, water content, o r T, relaxation during different day-groups were tested by paired t-test. The effects of parity or use of oral contraceptives were tested by two-way analysis of variance (two-way ANOVA). Relations between variablcs were tested by least-squares linear regression and correlation coefficient ( r j . MRl and water displacement techniques lor estimating total breast volume wcrc compared using the method of Bland & Altnian (1986).
90 80
60
70
80
90
100
Culibrution o,f MRI meusurenwnts The error of volume calculation from IR images was -5.7 (SD 1.3)%, indicating slight underestimation of volume calculation (Fig. 2a). while rcpcatability, expressed as a percentage of mean volume, was 4.3 (SD 2.7)%. Mean breast volumes as calculated hy MKI (641.3, SLI 239.8 ml) were significantly higher (paired t-test I = 2.93. P
