|Home | About | Journals | Submit | Contact Us | Français|
To evaluate different imaging protocols, especially with respect to radio frequency (RF) receiver coil location, for their suitability in providing least squares derived quantitative T2 values of ovarian follicular fluid for investigations of basic ovarian physiology.
The ovaries of 10 women were imaged via magnetic resonance imaging (MRI) using externally positioned and intravaginally placed RF receiver coils. Half-Fourier acquisition with single-shot turbo spin-echo (HASTE), multiple-echo T2, Dixon, turbo spin-echo, and 3-dimensional (3D) fast imaging with steady-state precession (FISP) and time-reversed FISP (PSIF) sequences were used. Quantitative T2 nuclear spin relaxation rate information from the ovarian follicles between data acquired with the external and intravaginal coils were compared. Additionally, the amount of ovarian follicle and corpora lutea structural detail visible was qualitatively assessed.
The T2 computations indicated that there was no difference in the follicular fluid T2 values or in the heterogeneity (spatial variance) of the T2 values between data acquired with the external RF coil and data acquired with the intravaginal RF coil. The best sequences for the visualization of ovarian internal structure were the 3D PSIF sequences and the multiple-echo T2-weighted images, confirming our earlier imaging work on excised cow ovaries.
It is best to use an externally placed RF coil for quantitative MRI study of ovarian physiology given the lack of difference in quantitative T2 information and the difficulty associated with imaging the ovaries using an intravaginal RF probe.
Quantitative magnetic resonance imaging (qMRI) is becoming an increasingly important aspect of diagnostic magnetic resonance imaging (MRI) and is widely used in research applications.1 In particular, the quantitative measurement of a MRI signal’s spin-spin, or T2, exponential decay (relaxation) rate in ovarian follicular fluid will be important for the study of basic ovarian physiology in women. Previous bovine ovarian studies have shown that T2 relaxation rates in ovarian follicular structures are correlated with the ovarian physiologic state.2,3 This paper discusses finding an optimal way of measuring T2 relaxation rates in ovarian follicular fluid. Specifically, the objective of the present study was to determine whether better quantification of T2 is possible with an intravaginally placed versus an externally placed radio frequency (RF) coil and to verify that imaging sequences previously used for the quantitative study of bovine ovarian physiology in vitro are suitable for the in vivo quantitative study of ovarian physiology in women.
At the beginning of every ovarian cycle, a cohort of several ovarian follicles begins to grow in size. Later in the cycle, one of the follicles in the cohort will become the largest and be termed dominant. That dominant follicle will either become atretic or it will ovulate, and knowing beforehand which route that follicle will take will be useful in assisted reproduction therapy, where a decision would be made to start another cycle if the dominant follicle were known to be atretic. Our previous work has shown that T2 values and the heterogeneity (spatial variance or image texture) of T2 values in the ovarian follicular fluid can be used to differentiate between dominant ovarian follicles that will become atretic and those that will ovulate. Atretic dominant follicles have a higher T2 value and lower T2 heterogeneity in the follicular antrum than do ovulatory dominant follicles.2
We also found, in our previous study of bovine ovaries, that the T2 heterogeneity value in the follicular antrum is positively correlated with follicular fluid progesterone concentration, meaning that quantitative follicular progesterone measurements may be possible from quantitative MRI data.
Quantitative information from the follicle wall can also distinguish between ovulatory and atretic follicles, with atretic follicles having a quantitatively brighter wall overall, with a larger linear variation of brightness across the wall, than have ovulatory follicles.3
If we are to observe similar quantitative correlations in women, it is necessary to identify appropriate imaging sequences in combination with the best RF coil placement. The optimum protocol will allow precise T2 quantification of the follicular fluid while providing visualization of ovarian structure in enough detail to make spatial size and position measurements of ovarian follicles and corpa lutea in time series investigations. We were interested to know whether intravaginally placed coils could provide superior quantitative T2 data for application to physiologic study of the ovary. The use of intravaginal coils for uterine imaging has been previously reported,4 but no previous formal evaluation of the use of intravaginal coils for quantitative ovarian imaging has been done. Our rationale was that internally placed coils would have the advantage of providing a higher signal-to-noise ratio for quantitative purposes than an externally placed RF coil if the internal coil could be placed close enough to the organs of interest. MRI protocols for the qualitative evaluation of pathology in the female pelvis are well established (see Wilber et al5 and references therein) and are useful for the identification of fibroids or adenomyosis, uterine anomalies, endometriosis, hemorrhagic ovarian cysts, and dermoid tumours. However, no standard imaging protocol exists for the quantitative study of the physiology ovarian follicles and corpora lutea in women. Our previous bovine work provided us with several MRI protocols that we wished to evaluate in vivo in preparation for an in vivo MRI study of ovarian physiology.6,7 These previously identified protocols provided the starting point for the present work.
Our imaging protocol began with the standard scout and half-Fourier acquisition with single-shot turbo spin-echo (HASTE) imaging.5 Various T2-weighted imaging sequences and 3-dimensional (3D) sequences, shown to be effective for previous in vitro bovine imaging,6,7 were used, following the identification of the ovaries in the HASTE images. Multiple-echo T2 sequences provided the data necessary for T2 quantification. Three-dimensional data allow for more accurate volumetric and spatial position measurement of ovarian structures than is possible with 2-dimensional data. Specific details follow.
The study was approved by the University of Saskatchewan Advisory Committee on Ethics in Human Experimentation, approval number BMC1999-249. Each participant gave written informed consent before she was imaged. Ten women ranging in age from 20 to 42 years, with a mean age of 29.8 years and not using hormonal contraception, participated in the study. The phase of the ovarian cycle of each participant was randomized so that the efficacy of the imaging protocols could be evaluated across the ovarian cycle. The earliest day of the ovarian cycle imaged was 5 days after ovulation, the latest 32 days, with a mean of 15.2 days after ovulation.
The women were imaged using a 1.5 Tesla Siemens Symphony (Erlangen, Germany) magnetic resonance imager. Every woman was first imaged with a curved external RF receiver coil placed on the anterior of the abdomen; this was followed by imaging with a specialized internal RF coil. The external coil was a circularly polarized array coil and was used in conjunction with RF coil elements permanently installed in the imager patient table to create a coil pair having a diameter roughly equal to the thickness of the woman being imaged. The internal coil (Medrad, Pittsburgh, PA) was manufactured to image the cervix endorectally but was applied intravaginally for this study. The internal coil was directed toward the ovaries, when placed, using positional information from the external coil images. Air bladders in the internal coil were also inflated in an attempt to prevent coil motion. We ran several imaging sequences, described below, using each RF coil. RF transmission was through the body coil with the external or internal coils described above used for RF reception only. All sequences used were the stock sequences provided with the imager.
Using the external RF coil., the first imaging sequence provided 3 scout images in each of the axial and coronal planes for general localization of the ovaries. The second imaging sequence, with the imaging slices positioned using the scout images, acquired 13 images in the sagittal plane, using the HASTE sequence.8 The HASTE sequences provided more precise ovarian localization information over the scout images for positioning the slices in the subsequent imaging sequences. The imaging parameters used in all sequences are given in Table 1.
The HASTE images were used for positioning slices for the multiple-echo T2 imaging sequences applied next. The multiple-echo T2 sequence used 180° RF refocusing pulses that were 90° out of phase with the initial 90° RF pulse, and no stimulated echo artifact (image banding) was observed. No presaturation RF pulses were used with the first applied 16-echo train sequence for 9 subjects (Table 2). Two further 16-echo imaging sequences that used fat presaturation9 and magnetization transfer presaturation10 were acquired for 8 subjects. An off-resonance magnetization transfer pulse was used to saturate bound water protons in the sequences that used magnetization transfer presaturation.
Continuing with the external RF coil, we ran additional sequences for the purpose of evaluating the ability of various sequences to provide visualization of internal ovarian structure. For 7 of the 10 women, image slices from a Dixon water-fat spectroscopic separation imaging sequence11 were obtained.
Turbo spin-echo (TSE) sequences without fat presaturation and with a “turbo factor” of 11 were used to obtain axial slices in 4 of the 10 women. In addition, a turbo spin-echo sequence with fat presaturation was used on 4 subjects, with magnetization transfer presaturation on 3 subjects.
Finally, two 3D imaging sequences were tested. A fast imaging with steady-state precession (FISP) sequence was used on 2 subjects. A 3D time-reversed FISP (PSIF) was also used on 2 subjects.
The imaging sequences used with the internal RF coil were similar to the sequences used with the external coil, except that different numbers of sequences were used for the 2 coils (see Table 2). The limitation of keeping the subject in the MRI for less than 90 minutes prevented us from applying all sequences to all subjects.
The 16-echo T2 sequence images from both the external RF coil and internal RF coil were used to produce maps of T2 nuclear spin relaxation-rate values through least squares fit to an exponential decay.2,3 Explicitly, the fit was made to
pixel by pixel using locally written software where S(tn) is the pixel intensity at echo n, tn is the echo time of echo n, and C is a constant. More complex curve fits were not used because this study was focused on finding optimal imaging approaches as opposed to a systematic investigation of ovarian physiology. Separate maps were made from data obtained with no RF presaturation, with fat presaturation, and with magnetization transfer presaturation. For each map made from external RF coil data, the largest follicle in each of the left and right ovaries was identified, and square regions of interest (ROI) were traced completely within the follicular antrum of those follicles for every image slice that contained the follicle. The ROIs were traced with commercial medical image-processing software (Cheshire, Hayden Image Processing Group, Waltham, MA).
The mean and standard deviation of the T2 value within the sum of the ROIs for each subject was computed. The mean T2 value so obtained represented the measurement of follicular fluid T2 using external RF coil data for the subject. The standard deviation of the T2 value so obtained represented the measurement of follicular fluid T2 heterogeneity, a mixture of image texture and noise, using external RF coil data for the subject.
The spatial coordinates of the follicles, identified in the external RF coil images, were used to locate the same follicles in the internal RF coil images. Once the follicles were identified in the internal RF coil images. ROIs were defined as in the external RF coil images, and follicular fluid values for T2 and T2 heterogeneity were computed for each subject. The measured T2 value was composed of the true value plus noise, so similar T2 measurements indicate similar signal-to-noise ratios for 2 measurements. More critically, similar variances in the measured values of T2 indicate similar signal-to-noise ratios. The T2 heterogeneity values measure T2 variance spatially, and that variance comprises variance owing to noise and variance owing to true spatial variance of T2. Since we were interested in true spatial T2 variance, or texture, as well as the true T2 value, we computed and compared both values as described next.
The values of T2 and T2 heterogeneity for the external RF coil versus the internal RF coil were compared for each of the no RF presaturation, fat presaturation, and magnetization transfer presaturation datasets. A chi-square frequency test was used to test the data distribution for normality. If the data were normally distributed, the data were compared using Student’s t-test. If the data were determined to be not normally distributed, a nonparametric Mann–Whitney rank sum test was used to compare the external RF coil and internal RF coil datasets.
At least one ovary could always be identified easily in the HASTE localization images made with the externally placed RF coil, but they were frequently difficult to locate in the HASTE localization images made with the intravaginal RF coil because the ovaries were too distant from the coil. When the ovaries were identified, the image contrast provided by the various imaging sequences were similar between external RF coil and internal RF coil data. Both ovaries were ultimately identified in all image sequences except for the following cases: in subject 2 only the right ovary could be identified, in subject 4 the left ovary could not be identified in the images acquired with the intravaginal coil, and in subject 5 neither ovary could be identified in the images acquired with the intravaginal coil.
The results of the comparison of the quantitative T2 values are summarized in Figure 1. ROIs could be defined in 8½ subjects with the external RF coil and in 5½ subjects with the intravaginal RF coil for the multiple-echo T2 images obtained without fat presaturation (½ because only one ovary was observed in subjects 2 and 4). No differences in T2 values (p = 0.844) or T2 heterogeneity values (p = 0.814) were observed; both were normally distributed. Usable ROIs could be defined in 8 subjects with the external RF coil and in 5 subjects with the intravaginal RF coil for the multiple-echo T2 images obtained with fat presaturation. The Mann–Whitney rank sum test revealed no differences in T2 values (p = 0.833) and the Student’s t-test revealed no differences in T2 heterogeneity values (p = 0.698). Usable ROIs could be defined in 8 subjects with the external RF coil and in 6 subjects with the intravaginal RF coil for the multiple-echo T2 images obtained with magnetization transfer presaturation. No differences in T2 values (p = 0.438) or T2 heterogeneity values (p = 0.499) were observed; both were normally distributed. Missing data from the images acquired with the intravaginal coil were due to the ovaries being too far from the coil to provide usable signal.
Typical images obtained with the multiple-echo T2 sequences are shown in Figure 2. When images from sequences that used no fat presaturation, fat presaturation, and magnetization transfer presaturation were compared, the image artifact owing to breathing motion was minimal in the images where fat presaturation was used, likely because the amplitude of motion caused by breathing is larger for nearby fatty tissue then for nonfatty tissue. Ovarian structure (visualization of follicles and corpora lutea) was more obvious (higher contrast) in the images with TE (echo time) greater than 100 ms. The image contrast and the visibility of ovarian follicles in the Dixon water images were less than that of the T2-weighted images. However, the Dixon images were often useful for confirming the position of the ovaries when they were not readily apparent in the HASTE images.
The amount of structural detail visible in the ovaries was greatest with the 512-image matrix turbo spin-echo sequence among all the sequences tried. As with the multiple-echo T2 images, the image artifact from motion was least when fat presaturation was used.
In the 3D image sequences, the image contrast for ovarian structures was greater in the PSIF images, compared with the FISP images. Image artifact owing to motion was also least in the PSIF images. Typical image slices from the 3D data are shown (Figure 3).
The lack of quantitative improvement in the measured values of T2 and T2 heterogeneity in data measured with the intravaginally placed RF coil over the externally placed coil indicates that there is no advantage to be had by using intravaginal coils, at least by themselves. Conversely, the poorer visual quality of the images acquired with intravaginal coil did not lead to poorer measured values of T2 and T2 heterogeneity with the intravaginal coil. While it was not possible to combine the signal from the intravaginal coil with the signal from the externally placed coil on our equipment, it is likely that the signal-to-noise ratio from the 2 coils together would be higher. However, given that the external coil provided images equal to or superior than images from the intravaginal coil for the measurement of ovarian follicular fluid and the fact that external coil use is more comfortable for the patient, we conclude that the use of an external coil only is sufficient for follicular fluid T2 measurement. Precise measurement of follicular fluid T2 may be important in future clinical applications of assisted reproduction technology where it may be used for the noninvasive determination of follicle ovulatory status and follicular fluid progesterone concentration. The excellent visibility of the ovarian follicles and corpus luteum in the 3D FISP and PSIF images in this first in vivo study confirmed our earlier work with excised bovine ovaries,6,7 with one important difference. The FISP sequence provided the clearest images of ovarian structure and the PSIF images had too-high contrast with the excised bovine ovaries. In this in vivo study in women, we found that the FISP images had too-low image contrast and the PSIF images provided the clearest images of ovarian structure. The PSIF sequence should therefore be used for the purposes of studying 3D follicular dynamics within the ovary in vivo. Time series 3D studies will allow us to test mathematical models of follicular physiology where precise time series measurements of individual follicle surface areas are required.14
We applied magnetization transfer pulses prior to the multiecho T2 pulse trains and found no appreciable affect on the measured T2 value. However, given that the follicular fluid may contain significant amounts of macromolecules at different points in the ovarian cycle,2 it is conceivable that magnetization transfer effects may be used to make inferences about the physiologic state of the follicle. It is further important to note that variable magnetization transfer effects from the slice selection pulses exist in multislice sequences. As shown in Table 1, the sequence parameters were kept fixed between subjects, so any such magnetization transfer effects in our study were constant between subjects.
Previous studies with ultrasonography and with MRI of bovine ovaries in vitro have demonstrated that image parameters are significantly correlated with ovarian physiologic status. Ultrasonographic image texture and intensity in the follicular antrum and wall and in the corpus luteum change with their physiologic state.12 Similar physiologically relevant changes in the values of T2 relaxation rates in MRI images of bovine ovaries made in vitro have also been observed.2,3 We expect similar physiologic correlations with T2* values, especially in the follicle wall, because T2* is sensitive to vascularization change on the basis of varying deoxyhemoglobin.13 It is anticipated that T2 values or T2* values will reflect the vascularization of the theca cell layers in maturing follicles, based on differences in T2 (or T2*) values owing to blood in the ovarian stroma of women imaged at different times in the ovarian cycle. Quantification of T2* can be obtained from multiple-echo gradient-echo sequence data (as opposed to the spin-echo approach used in this study).
We conclude that it is best to use an externally placed RF coil for the physiologic study of ovaries in vivo. The lack of improvement in quantitative T2 measurement combined with the difficulty of locating the ovaries in images made with the intravaginally placed RF coil shows that there is no advantage to the use of an intravaginally placed RF coil. In agreement with our earlier work with excised bovine ovaries,6 considerable anatomic detail may be seen in T2-weighted images (images having TE greater than approximately 100 ms). The T2-weighted turbo spin-echo sequences were found to be useful for visualization of ovarian structure at high resolution (512 squared image matrices). The 16-echo T2 sequences provide lower image resolution (256 squared image matrices) but enable the computation of T2 maps for T2 magnitude and T2 heterogeneity determination. The application of fat presaturation reduced image artifact and blurring due to breathing motion with both the turbo spin-echo and the multiple-echo T2 sequences.
In future studies involving daily imaging of individual women, we expect to observe changes in T2 values associated with ovarian follicles and corpora lutea that are correlated with physiologic changes throughout the ovarian cycle on the basis of our past investigations with excised bovine ovaries. These planned in vivo investigations will focus on the use of multiple-echo T2 and PSIF sequences with an externally placed RF coil, as a result of the study described here.
The numerical image analysis was performed by Jennifer Hadley. Dawn Senko assisted with image acquisition for the first 2 subjects. This research was supported by grants from the Canadian Institutes for Health Research (CIHR) and the Saskatchewan Health Services Utilization and Research Commission (HSURC).
Gordon E Sarty, Departments of Psychology and Medical Imaging, University of Saskatchewan, Saskatoon, Sk.
Angie R Baerwald, Department of Obstetrics, Gynecology and Reproductive Sciences, University of Saskatchewan, Saskatoon, Sk.
John Loewy, Humber River Regional Hospital, Toronto, ON.
Roger A Pierson, Department of Obstetrics, Gynecology and Reproductive Sciences, University of Saskatchewan, Saskatoon, Sk.