Repeatability and Reproducibility of 3D Magnetic Resonance Fingerprinting Relaxometry Measurements in Normal Breast Tissue

Abstract

Background:

3D Breast Magnetic Resonance Fingerprinting (MRF) technique enables T₁ and T₂ mapping in breast tissues. Combined repeatability and reproducibility studies on T₁ and T₂ relaxometry are lacking.

Purpose:

To assess test-retest and two-visit repeatability and inter-scanner reproducibility of 3D Breast MRF technique in a single-institution setting.

Study type:

Prospective

Subjects:

Eighteen women(median age 29 years, 22-33 years)underwent visit 1 scans on scanner 1. Ten of these women underwent test-retest scan repositioning after 10-minute interval. 13 women had visit 2 scans within 7-15 days in same menstrual cycle. Remaining 5 women had visit 2 scans in same menstrual phase in next menstrual cycle. Five women were also scanner at scanner 2 at both visits for inter-scanner reproducibility.

Field Strength/Sequence:

Two 3T MR scanners with 3D Breast MRF technique

Assessment:

T₁ and T₂ MRF maps of both breasts.

Statistical Tests:

Mean T₁ and T₂ values for normal fibroglandular tissues were quantified at all scans. For variability, between and within-subjects coefficients of variation (bCV and wCV respectively) were assessed. Repeatability was assessed with Bland-Altman analysis and coefficient of repeatability (CR). Reproducibility was assessed with inter-scanner CoV and Wilcoxon signed-rank test.

Results:

The bCV at test-retest scans was 9–12% for T₁, 7-17% for T₂, wCV was<4% for T₁, and<7% for T₂. For two visits in same menstrual cycle, bCV was 10–15% for T₁, 13-17% for T₂, wcV was< 7% for T₁ and< 5% for T₂ .For two visits in same menstrual phase, bCV was 6-14% for T₁, 15-18% for T₂, wCV was< 7% for T₁ and < 9% for T₂. For test-retest scans, CR for T₁ and T₂ were 130 ms and 11 ms. For two visit scans, CR was <290 ms for T₁ and 10-14 ms for T₂.Inter-scanner CoV was 3.3-3.6% for T₁ and 5.1-6.6% for T₂ with no differences between inter-scanner measurements (p=1.00 for T₁, p=0.344 for T₂).

Conclusion:

3D Breast MRF measurements are repeatable across scan timings and scanners and may be useful in clinical applications in breast imaging.

INTRODUCTION

Quantitative property mapping techniques can complement conventional anatomical magnetic resonance (MR) imaging by allowing more objective assessment of pathophysiology and facilitating development of imaging biomarkers(1–3). Diffusion weighted imaging is routinely employed in breast MR imaging as a quantitative measure of tissue cellularity while dynamic contrast enhanced-MRI is primarily used as a semi-quantitative technique in breast imaging perfusion. Measurement of T₁ and T₂ relaxation times can provide additional information about normal and diseased breast tissue(4, 5). Previous studies on breast relaxometry have observed significant differences in T₁ and T₂ relaxation times between normal fibroglandular breast tissue and benign and malignant breast tumors(5, 6). Changes in T₂ relaxation time has also been used to characterize response to chemotherapy in breast cancers(7–9). However the T₁ and T₂ values reported for normal breast tissue show a wide variation due to the differences in mapping techniques and field strengths(5, 10, 11). Similarly, studies on response assessment to treatment used different T₂ mapping techniques, which affects comparability of these results(7–9). For quantitative imaging techniques to be successfully used in practice, assessment of measurement variability and repeatability are crucial to interpret results. While studies have been published on repeatability and reproducibility of breast tissue apparent diffusion coefficient (ADC) values, fibroglandular tissue density and fibroglandular tissue enhancement (12–14) and a recent study assessed repeatability and reproducibility metrics for both T₁ mapping and ADC in breast(15), similar repeatability and reproducibility studies on combined T₁ and T₂ breast relaxometry are lacking.

Recently, a 3D Magnetic Resonance fingerprinting (MRF) technique was described for breast imaging that allows simultaneous and volumetric quantification of T₁ and T₂ relaxation times for both breasts in less than 6 minutes of scan time (16), at a resolution that is comparable to that for clinically employed ADC mapping. The ability to obtain rapid, volumetric and simultaneous T₁ and T₂ maps represents a significant advantage over previously described breast relaxometry methods, that typically measured only property at a time, required long acquisition times and had limited organ coverage. The T₁ and T₂ values measured with 3D breast MRF have shown to be accurate and comparable to literature with phantom and in-vivo validation respectively(16). Clinically, measurements obtained from 3D breast MRF may be useful for lesion characterization or assessment of treatment response. However, prior to using this technique for clinical applications, it is important to understand the variability and repeatability of this technique. Hence, the objective of this study was to assess the repeatability and reproducibility of 3D breast MRF technique in normal breast tissues.

MATERIALS AND METHODS

Scanning Protocol

In this Institution Review Board approved prospective study, 18 asymptomatic pre-menopausal women were scanned with 3D breast MRF after obtaining written informed consent. Scans were done on two 3T MR scanners (Siemens, Verio (MR1) and Skyra (MR2) as detailed below.

Scans Performed at Visit 1:

1. Baseline scans: All 18 women underwent scans on scanner 1 (MR1) for baseline data

2. Test-retest repeatability: Ten of these women also underwent test-retest scans on MR1, in which 3D breast MRF was performed before and after repositioning with a 10 minute inter-scan interval to assess test-retest repeatability.

3. Inter-scanner reproducibility: Five of these eighteen women also underwent same-day scans on second scanner (MR2).

Scans Performed at Visit 2:

1. Two-visit repeatability in different menstrual phases: Thirteen women were scanned again on MR1 within 7–15 days in different menstrual phases within the same menstrual cycle.

2. Two-visit repeatability in same menstrual phase: Remaining five women were scanned again on MR1 in the next menstrual cycle in the same menstrual phase as the baseline scans.

3. The same five women scanned for inter-scanner reproducibility in the first visit also underwent same-day scans on MR2 for second-visit assessment of inter-scanner reproducibility.

All scans were performed using a breast coil with eight receive channels. The 3D MRF sequence used in this study was based on a 2D Fast Imaging with Steady State Free Precession (FISP) based MRF acquisition(17), modified for 3D breast imaging (16). The data were acquired sequentially in 48 partitions, with each partition containing 12 segments; each segment comprising of a) magnetization-preparation module (three inversion recovery modules with inversion recovery times of 20, 100 or 250 ms and six T2 preparation modules with effective echo times of 50 or 90 ms), b) spectral-selective fat-saturation module to suppress the fat signal in breast adipose tissue and c) data acquisition window with variable flip angles ranging from 5° to 12°. Other settings included: FOV: 400×400 mm², matrix size 256 × 256 mm², TR/TE: 6.1/0.9ms, in-plane resolution: 1.6×1.6 mm², partition thickness: 3 mm, craniocaudal coverage: 14 cm, number of partitions: 48, partial Fourier in the partition direction: 6/8(16). The overall acquisition time for 48 partitions was approximately 5 minutes 30 seconds. No subject-specific shimming sequence was performed for the MRF acquisition, as the FISP readout is less sensitive to B₀ field inhomogeneity as compared to the balanced steady state free precession (SSFP) readout(18). As described previously (16), the effects of B₁ heterogeneity are also compensated in the MRF acquisition scheme.

In addition to the 3D breast MRF sequence, a 3D gradient echo T₁w fat saturated (FS) image was obtained for anatomic localization (scan parameters: FOV 300 × 300 mm², TR/TE: 4.76/1.62 ms, flip angle: 10 degrees, fat saturation, resolution 0.8 × 0.9 × 1.0 mm³, acquisition time: 1 minute 29 seconds). The total table time per subject including the sagittal localizer, 3D T₁w FS sequence and 3D breast MRF in each sitting was approximately 9 minutes.

Data Processing and Image Analysis

A dictionary including the signal evolutions from a wide range of T₁ and T₂ values (T₁, 60 to 5000 ms; T₂, 10 to 500 ms) was first calculated using Bloch equation simulations. The raw MR fingerprinting data were processed offline using Matlab (Matlab 2014a; MathWorks, Natick, Mass) and co-registered T₁, T₂ and proton density maps were generated using a pattern-matching process(16). The 3D breast MRF partition containing the largest homogenous area of normal fibroglandular breast tissue was selected. Region of Interests (ROIs) were manually drawn by one radiologist (8 years radiology experience) on proton density maps. Separate ROIs were drawn for right and left breast (Figure 1) and both T₁ and T₂ values were simultaneously obtained. For retest and repeatability analysis, ROIs were drawn on the partition showing similar distribution of breast fibroglandular tissue as compared to the reference scan. The cross-sectional anatomic landmarks (cardiac/breast) were additionally used as references for confirming the accuracy of partitions selected for repeat ROI annotations. The T₁ and T₂ values from all voxels within each ROI were extracted from quantitative T₁ and T₂ maps and averaged to obtain the mean T₁ and T₂ for each ROI.

Figure 1:

Representative T₁ and T₂ MR Fingerprinting color maps of normal breast tissue in a healthy volunteer. Regions of interest (ROIs) were drawn over both breasts on the proton density maps and both T₁ and T₂ relaxation times were simultaneously obtained.

Statistical Analysis

The mean T₁ and T₂ values were used for statistical analysis. The means of T₁ and T₂ were compared between right and left breast using Wilcoxon tests for related samples. The variability and repeatability statistics in this study are similar to those used previously in studies in normal breasts(12, 15). The variability of T₁ and T₂ measurements was assessed by coefficient of variation (CV). The between-subject CV (bCV) at each visit and for each scan session was expressed as a percentage ratio of standard deviation (SD) to the mean of measurements for each breast and for all breasts, the equation is as follows:

$$bCV=\mathit{100}\mathit{\times }SD∕mean$$

The within-subject CV (wCV) was defined as the percentage ratio of the within-subject SD (wSD) to the overall mean of measurements at each visit and each scan session. For measurements 1 and 2 (m1 and m2 respectively) at both visit 1/visit 2 and for test/retest scans for a total of n measurements in each analysis, the equations are as follows

$$\begin{gathered} wSD=\sqrt{\frac{\Sigma (m1-m2{)}^2}{2n}};\mathrm{overall}\mathrm{mean}=\frac{\Sigma (m1-m2{)}^2}{2n}, \\ wCV=\mathit{100}\mathit{\times }\mathrm{wSD}∕\mathrm{overall}\mathrm{mean} \end{gathered}$$

The bCV and wCV values were computed separately for each side and both sides together.

The repeatability was assessed using Bland-Altman analysis(19). For both T₁ and T₂, scatterplots of the difference at two visits and for test-retest scans were plotted against the mean of measurements at two visits and for test-retest scans respectively. The SD of the difference (SD_diff) was estimated and upper and lower limits of agreement (LOA) was defined as follows: Upper LOA = mean difference + 1.96SD_diff; Lower LOA = mean difference - 1.96SD_diff. The 95% confidence intervals for mean difference; upper and lower LOAs were also reported using the appropriate statistics(19). Differences in measurements that fell within the LOA were considered clinically unimportant. The coefficient of repeatability (CR) was computed as 1.96 x √2 x SD_diff (2.77 x SD_diff). Kendall Tau Test was used to assess if there was any correlation between the difference and the mean of measurements at test/retest and two-visit scans.

Inter-scanner Reproducibility

Inter –scanner reproducibility of 3D breast MRF T₁ and T₂ was assessed on both scanners using Wilcoxon signed rank test to look for differences in measurements on two scanners. The coefficient of variation was estimated for all measurements on MR1 (CV_MR1) and MR2 (CV_MR2). The inter-scanner coefficient of variation (CV_MR1-MR2) for all measurements calculated as percentage ratio of inter-scanner standard deviation [SD(MR1-MR2)] versus overall mean of measurements on both scanners. For measurements on MR1 (s1) and MR2 (s2) for a total of n measurements, the equations are as follows:

$$\begin{gathered} \text{SD(MR1}-\text{MR2)}=\sqrt{\frac{\Sigma (s1-s2{)}^2}{2n}};\mathrm{overall}\mathrm{mean}=\frac{\Sigma (s1-s2{)}^2}{2n}, \\ {\text{CV}}_{\text{MR1-MR2}}=\mathit{100}\mathit{\times }\mathrm{wSD}∕\mathrm{overall}\mathrm{mean} \end{gathered}$$

All statistical analysis was performed using statistical software package (SPSStatistics, version 20, IBM®).

RESULTS (TABLES AND PLOTS)

Subject Characteristics and Mean Statistics

The subjects’ ages ranged from 22–33 years with median age of 29 years. Table 1 details the mean T₁ and T₂ values for right and left breasts for all 18 subjects at visit 1. There was no statistical difference in mean T₁ and T₂ values between right vs. left breasts at baseline visit 1 (right vs. left; P = 0.89 for T₁, P = 0.25 for T₂). Figure 2 shows representative T₁ and T₂ MRF color maps from visit 1 test/retest and visit 2 scans from one subject. There were also no differences in the means of T₁ and T₂ for right and left breasts at both test/retest scans and two-visit repeatability scans (Table 2).

Table 1:

Subject characteristics, T₁ and T₂ values for fibroglandular tissue for right and left breasts at Visit 1 and scanner 1

				T1 at Visit 1 (ms) (Mean ± SD)		T2 at Visit 1 (ms) (Mean ± SD)
Subject #	Age (years)	Menstrual phase for Visit 1	Menstrual phase for Visit 2	Right Breast	Left Breast	Right Breast	Left Breast
1	33	Luteal	Proliferative	1443 ± 524	1646 ± 327	50 ± 8	56 ± 13
2	29	Luteal	Menstrual	1187 ± 268	1110 ± 190	54 ± 8	51 ± 14
3	29	Luteal	Proliferative	1143 ± 141	1103 ± 201	41 ± 7	45 ± 9
4	29	Proliferative	Luteal	1282 ± 303	1220 ± 213	57 ± 8	56 ± 9
5	32	Proliferative	Follicular	1537 ± 240	1520 ± 203	62 ± 10	67 ± 12
6	31	Menstrual	Proliferative	1179 ± 212	1343 ± 267	48 ± 10	53 ± 9
7	28	Luteal	Menstrual	1311 ± 321	1104 ± 315	46 ± 11	57 ± 15
8	28	Luteal	Proliferative	1250 ± 303	1123 ± 341	47 ± 11	51 ± 11
9	30	Menstrual	Proliferative	1079 ± 380	1074 ± 384	40 ± 13	48 ± 16
10	22	Proliferative	Luteal	1127 ± 221	1128 ± 199	44 ± 8	41 ± 8
11	30	Luteal	Menstrual	1320 ± 235	1390 ± 221	53 ± 9	48 ± 9
12	26	Proliferative	Luteal	1115 ± 169	1151 ± 172	35 ± 5	44 ± 4
13	22	Proliferative	Luteal	1186 ± 242	1066 ± 117	55 ± 8	51 ± 5
14	23	Luteal	Luteal	1269 ± 189	1416 ± 212	47 ± 9	51 ± 8
15	28	Luteal	Luteal	1003 ± 239	1021 ± 185	49 ± 11	41 ± 7
16	30	Proliferative	Proliferative	1315 ± 206	1337 ± 262	49 ± 10	55 ± 9
17	30	Menstrual	Menstrual	1066 ± 162	1133 ± 150	36 ± 6	39 ± 6
18	24	Luteal	Luteal	1150 ± 262	1093 ± 294	56 ± 10	57 ± 10
All Subjects				1220 ± 135	1228 ± 174	48 ± 8	51 ± 7.0

Volunteers 1-13 underwent second visit scans within 7-15 days in the same menstrual cycle but different menstrual phases. Volunteers 14-18 were rescanned in the same menstrual phase over two successive menstrual cycles.

Figure 2:

Representative T₁ and T₂ MR Fingerprinting color maps from test-retest scans at visit 1 and visit 2 scan from one subject. The maps show similar range of numbers as seen from color-bars despite differences in breast positions and scan timings.

Table 2: Summary of fibroglandular tissue T₁ and T₂ values for right and left breast for all repeatability scans.

There was no difference in the means of T₁ and T₂ for right and left breasts at both visits using Wilcoxon tests for related samples.

		T1 relaxation time (ms) (Mean ± SD)			T2 relaxation time (ms) (Mean ± SD)
	No of subjects	Right Breast	Left Breast	Right vs. Left P-value	Right Breast	Left Breast	Right vs. Left P-value
Test Scan	10	1154 ± 105	1164 ± 142	1.00	45 ± 8	48 ± 6	0.24
Retest Scan		1171 ± 112	1190 ± 135	0.89	44 ± 7	46 ± 7	0.41
Visit 1 Scan	13	1243 ± 134	1259 ± 185	0.31	49 ± 7	52 ± 8	0.92
Visit 2 Scan in same menstrual cycle		1196 ± 170	1285 ± 184	0.92	49 ± 9	52 ± 8	0.70
Visit 1 Scan	5	1161 ± 132	1200 ± 168	0.60	46 ± 8	49± 8	0.47
Visit 2 Scan in same menstrual phase		1183±76	1212 ± 168	0.69	47 ± 8	47 ± 7	0.55

Between-Subject Variation and Within-Subject Variation

Visit 1Test-Retest scans (n = 10)

Figure 3 (a, b) show the scatter of T₁ and T₂ values at test-retest scans for both breasts. T₁ values at test-retest scans ranged from 1003 ms to 1416 ms and T₂ values ranged from 35 ms to 57 ms. The bCV for T₁ at both scans was 9–12% and for T₂ was 7–17% (Table 3). The wCV for T₁ was < 4% and < 7% for T₂ (Table 3).

Figure 3:

Scatter plots of repeated measurements for T₁ and T₂. Top row (a, b) shows results for same-day test-retest scans, middle row (c, d) shows results for two visit scans in different menstrual phases in the same menstrual cycle and bottom row (e, f) shows results for two visit scans in the same menstrual phase in the next menstrual cycle. The black line represents the identity line. Values for right and left breasts are shown separately.

Table 3:

Summary results for variability of fibroglandular tissue T₁ and T₂ relaxation times for test-retest scans.

	T1 relaxation time			T2 relaxation time
	bCV Test Scan (%)	bCV Retest Scan (%)	wCV (%)	bCV Test Scan (%)	bCV Retest Scan (%)	wCV (%)
All Breasts (n = 20)	10.77	10.73	3.16	7.13	6.76	4.92
Right Breast (n = 10)	9.30	9.72	3.01	17.16	14.74	6.48
Left Breast (n = 10)	12.22	11.19	2.87	12.95	15.41	5.80

Two-visit Scans at Different Menstrual Phases in the Same Menstrual Cycle (n = 13)

Figure 3 (c, d) show the scatter of T₁ and T₂ values at two visits for both breasts. T₁ values at both visits ranged from 942 ms to 1646 ms and T₂ values ranged from 32 ms to 67 ms. For two visits at different menstrual phases in the same menstrual cycle, the bCV for T₁ was 10–15% and for T₂ was 13–17% (Table 4). The wCV was < 7% for T₁ and < 5% for T₂ (Table 4).

Table 4:

Summary results for variability of fibroglandular tissue T₁ and T₂ relaxation times for different menstrual phase in the same menstrual cycle two-visits scans.

	T1 relaxation time			T2 relaxation time
	bCV Visit 1 (%)	bCV Visit 2 (%)	wCV (%)	bCV Visit 1 (%)	bCV Visit 2 (%)	wCV (%)
All Breasts (n = 26)	12.73	14.38	5.9	14.23	16.69	4.96
Right Breast (n = 13)	10.78	14.19	6.3	15.40	17.33	4.99
Left Breast (n = 13)	14.67	14.30	5.2	13.54	16.33	4.53

Two-visit Scans in Same Menstrual Phase (n = 5)

Figure 3 (e, f) shows the scatter of T₁ and T₂ values at same menstrual phases over two visits for both breasts. T₁ values at both visits ranged from 1021 ms to 1416 ms and T₂ values ranged from 36 ms to 57 ms. For two visits in the same menstrual phases, the bCV for T₁ was 6–14% and for T₂ was 15–18% (Table 5). The wCV was < 7% for T₁ and < 9% for T₂ (Table 5).

Table 5:

Summary results for variability of fibroglandular tissue T₁ and T₂ relaxation times for same menstrual phase two-visits scan.

	T1 relaxation time			T2 relaxation time
	bCV Visit 1 (%)	bCV Visit 2 (%)	wCV (%)	bCV Visit 1 (%)	bCV Visit 2 (%)	wCV (%)
All Breasts (n = 10)	12.75	5.68	4.49	15.87	14.55	4.05
Right Breast (n = 5)	11.36	6.41	6.23	18.2	18.00	5.84
Left Breast (n = 5)	14.03	5.20	4.39	16.82	14.82	8.43

Bland-Altman Analysis

Both T₁ and T₂ were repeatable between test-retest scans and two-visit scans independent of the timing of the second visit with respect to the menstrual cycle or menstrual phase (Figure 4).

Figure 4:

Bland Altman plots for differences in measurements from two scans versus mean of measurements from two scans. Top row shows T₁ (a) and T₂ (b) plots for same-day test-retest scans, middle row shows T₁ (c) and T₂ (d) plots for two-visit scans in different menstrual phases in the same menstrual cycle and bottom row shows T₁ (e) and T₂ (f) plots for two-visit scans in the same menstrual phase in the next menstrual cycle. Upper and lower solid lines represent LOA, the center dashed line is the mean difference in measurement from two scans with 95% CI of mean difference in dotted lines.

Visit 1Test-Retest scans (n = 10)

As shown in the Bland-Altman plots in Figure 4 (a, b), the mean difference for T₁ at test-retest was −21 ms (95% CI −44, 1.8 ms) with upper LOA (95% CI) of 71 ms (109, 33 ms) and lower LOA (95% CI) of −113 ms (−75, −151 ms). The CR for T₁ was 130 ms. The mean difference for T₂ at test-retest was 1.6 ms (95% CI −0.3, 3.4 ms) with upper LOA (95% CI) of 9 ms (12.3, 6.0 ms) and lower LOA (95% CI) of −6 ms (−2.9, −9.1 ms). The CR for T₂ was 11 ms.

Two-visit scans at Different Menstrual Phases in the Same Menstrual Cycle (n = 13)

As shown in the Bland-Altman plots in Figure 4 (c, d), the mean difference for T₁ at two visits was 11.03 ms (95% CI −31, 53 ms) with upper LOA (95% CI) of 216 ms (289,143 ms) and lower LOA (95% CI) of −194 ms (−121, −267 ms). The CR for T₁ was 289 ms. The mean difference for T₂ at two visits was 0.08 ms (95% CI −1.4, 1.5 ms) with upper LOA (95% CI) of 7 ms (9.6, 4.6 ms) and lower LOA (95% CI) of −7 ms (−4.4, −9.4 ms). The CR for T₂ was 10 ms.

Two-visit Scans in Same Menstrual Phase (n = 5)

As shown in the Bland-Altman plots in Figure 4 (e, f), the mean difference for T₁ at two visits was −18 ms (95% CI −85, 50 ms) with upper LOA (95% CI) of 167 ms (284,50 ms) and lower LOA (95% CI) of −202 ms (−86, −319 ms). The CR for T₁ was 261 ms. The mean difference for T₂ at two visits was 0.3 ms (95% CI −3.3, 3.9 ms) with upper LOA (95% CI) of 10 ms (16.6, 3.9 ms) and lower LOA (95% CI) of −10 ms (−3.4, −15.9 ms). The CR for T₂ was 14 ms.

Kendall Tau analysis did not reveal statistically significant coefficient values for any of the comparisons suggesting that the differences between repeated measurements were independent of the means of repeated measurements for T₁ and T₂.

Inter-scanner Reproducibility Results

Figure 5 shows scatter of T₁ and T₂ measurements on MR2 versus MR1 at both visits. There were no statistical differences between inter-scanner T₁ and T₂ at both visits (Visit 1: (MR1-MR2) T₁, p = 1.000, Visit 1: (MR1-MR2) T₂, p = 0.344, Visit 2: (MR1-MR2) T₁, p = 1.000, Visit 2: (MR1-MR2) T₂, p = 0.109).

Figure 5:

Scatter plots of repeated measurements for T₁ (a) and T₂ (b) on scanner 2 (MR2) versus scanner 1 (MR1). Top row (a, b) shows the relationships at visit 1 and bottom row (c, d) shows the relationships at visit 2. The inter-scanner CoV [CV_(MR1-MR2)] for T₁ and T₂ was 3.6% and 6.6% at visit 1 and 3.3% and 5.1% at visit 2 respectively without significant differences in inter-scanner measurements. The black line represents the identity line.

At visit 1, for T₁, the CV_MR1 , CV_MR2 and (CV_MR1-MR2) were 12.2%, 11.78% and 3.6% respectively. For T₂, the CV_MR1 , CV_MR2 and (CV_MR1-MR2) were 16.7%, 15.8% and 6.6% respectively. At visit 2, for T₁, the CV_MR1 , CV_MR2 and (CV_MR1-MR2) were 5.6%, 6.97% and 3.3% respectively. For T₂, the CV_MR1 , CV_MR2 and (CV_MR1-MR2) were 15.6%, 10.2% and 5.1% respectively.

DISCUSSION

This work shows the repeatability and inter-scanner reproducibility of 3D breast MRF measurements in a single-institution setting. The T₁ and T₂ measured using 3D breast MRF were repeatable across multiple volunteers at different scan timings. The ranges of T₁ and T₂ for normal fibroglandular tissue observed in this study are similar to the mean T₁ of 1198 ± 99 ms and 1445 ± 93 and T₂ of 40 ± 5 ms and 54 ± 9 values reported previously(11, 16) with comparable T₁ and T₂ for right and left breasts for all subjects.

There was a higher bCV as compared to wCV, indicating a greater inter-subject variability but relatively consistent intra-subject measurements. The greater inter-subject variation may be due to the physiological differences in breast density and in timings of menstrual phases between different women. While we are unaware of previous data on bCV for T₁ and T₂ available for comparison, the bCV for 3D breast RF T₁ (9–15%) and T₂ (7–18%) across all sub-groups is comparable to the previously reported bCV of 18% for ADC and much less than bCV of 47% and 61% for fibroglandular tissue enhancement and fibroglandular density respectively (12). Thus breast relaxometry may potentially be a more reliable quantitative measure of breast fibroglandular tissue characteristics as compared to fibroglandular tissue density/enhancement metrics despite differences in scan timings between subjects and subject positioning on repeated scans.

In this study, wCV was low for both T₁ and T₂. The wCV for 3D breast MRF T₁ is comparable to the wCV of T₁ mapping with variable flip angle technique (15). Additionally, the low wCV for T₂ obtained with 3D breast MRF can allow repeated and reliable mapping of both T₁ and T₂ in the same subjects on follow-up scans. The wCV for 3D breast MRF T₁ and T₂ are also comparable to the 5–11% wCV reported for ADC (12–15, 20) and lesser than 13–22% wCV reported for fibroglandular tissue density and fibroglandular tissue enhancement respectively (12). In this study, we scanned a few women in same menstrual phase and the variability in T₁ and T₂ metrics was similar to repeat scans in different menstrual phases. The small within-subject variability in measurements may make this technique useful for quantitative assessment in longitudinal studies wherein baseline and follow-up clinical scans in patients may be scheduled at different menstrual phases. Prior studies on ADC mapping have shown that the overall fluctuation in normal breast ADC is relatively small (20) and there were no differences in ADC and fibroglandular tissue density in relation to menstrual cycle(12, 13). There is less contemporary evidence on menstrual variation on T₁ and T₂. Older studies have shown no significant alterations in T₁ values in different stages of menstrual cycles (21, 22) while for T₂ values, the results have been mixed, with one study reporting no change (23) while another reporting significant change between second and 3^rd week of cycles(21). However, the studies on T₂ relaxometry were performed using older mapping techniques and at extremely low field scanners (0.02 T) and results may not reflect true breast physiology. Thus 3D breast MRF may also be a useful technique in simultaneously evaluating the physiologic variability in T₁ and T₂ over menstrual cycles on the current high field scanners used in clinical practice.

In general, for test-retest scans, there was lesser variability and lower CR as compared to two-visit scans. The 5 women scanned in same menstrual phase showed a slightly greater difference in T₂ values on follow-up scans compared to 13 women scanned in the different menstrual phase and this is likely due to the smaller number of subjects and lesser number of measurements in the same-menstrual phase group compared to different-menstrual phase cohort. Overall, based on Bland-Altman plots and CR, a T₁ difference of < 290 ms and T₂ difference of 10–14 ms on repeat scanning is likely to be physiologic and clinically unimportant. Conversely, an effect size greater than this would represent a true measurable difference due to pathology or disease.

The 3D breast MRF technique also had a low inter-scanner variation, ranging from 3–4% for T₁ and 5–7% for T₂ at two separate visits. To our knowledge, there has been no prior work on inter-scanner reproducibility of both T₁ and T₂ together. A recent work explored the inter-site reproducibility of B1-corrected variable flip angle T₁ mapping technique and showed an average difference of 8.4% for fibroglandular tissue and 7.5% for adipose tissue(15), both of which are slightly higher than the differences obtained for 3D breast MRF T₁. Moreover, 3D breast MRF does not require prior subject-specific shimming or B₁ correction as the low-flip angles used in the acquisition scheme mitigates the effect of B₁ inhomogeneity (16). This may make the technique less prone to T₁ mapping errors due to B₁ inhomogeneity and easier to apply across scanners and sites in practice (24). As 3D breast MRF acquisition scheme also uses a fat suppression module; breast fat matches to lowest dictionary entry and the maps generated depict the quantitative distribution of T₁ and T₂ fibroglandular tissue. However, imperfect fat suppression or partial volume effects may still theoretically affect measurements and this may be better evaluated with larger sample population in the future.

This study has several limitations. First, all volunteers were healthy young premenopausal women. There may be some physiological variation in relaxation times due to loss of hormonal stimulation or decreased fibroglandular tissue volume with age. While one study showed a slight increase in T₁ relaxation times with age (21), in the study by Chen et al(16), older patients had a slightly lower T₁ in healthy contralateral breast compared to younger, healthy volunteers with a mean difference of < 200 ms while T₂ values were similar in both groups. Second, single-partition ROIs were drawn instead of volumetric ROIs. While partitions selected for drawing repeat ROIs were anatomically similar to the partition used for baseline data, in the future volumetric ROIs may allow more precision in measurements. Third, this was a single-institution, single observer study. However, our sample size is comparable to some of the other previously published single–institution repeatability studies in breast imaging (12, 15, 20). This is an initial feasibility study on the repeatability and reproducibility of this breast relaxometry technique before expanding it to larger populations and multiple institutions, or employing it for clinical applications in breast oncologic imaging.

In conclusion, 3D breast MRF T₁ and T₂ measurements in normal breast tissue are repeatable across various scan timings and scanners and may potentially be useful in clinical applications in breast imaging and longitudinal follow-up of individual subjects.