Advertisement

Feasibility study of 2D Dixon-Magnetic Resonance Fingerprinting (MRF) of breast cancer

  • Eloisa Zanderigo
    Correspondence
    Corresponding author at: Department of Diagnostic and Interventional Radiology, UKT Tübingen University Hospital, Hoppe-Seyler-Straße 3, 72076 Tübingen, Germany.
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany

    Department of Diagnostic and Interventional Radiology, UKT Tübingen University Hospital, Hoppe-Seyler-Straße 3, 72076 Tübingen, Germany
    Search for articles by this author
  • Luisa Huck
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Martina Distelmaier
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Ebba Dethlefsen
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Mirjam Maywald
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Daniel Truhn
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Timm Dirrichs
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Mariya Doneva
    Affiliations
    Tomographic Imaging Systems, Philips Research Europe, Hamburg, Germany
    Search for articles by this author
  • Volkmar Schulz
    Affiliations
    Physics of Molecular Imaging Systems, Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany

    Hyperion Hybrid Imaging Systems GmbH, Aachen, Germany

    Physics Institute III B, RWTH Aachen University, Aachen, Germany

    Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany
    Search for articles by this author
  • Christiane K. Kuhl
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany
    Search for articles by this author
  • Teresa Nolte
    Affiliations
    Department of Diagnostic and Interventional Radiology, RWTH Aachen University Hospital, Pauwelsstr. 30, 52074 Aachen, Germany

    Physics of Molecular Imaging Systems, Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany
    Search for articles by this author
Open AccessPublished:November 16, 2022DOI:https://doi.org/10.1016/j.ejro.2022.100453

      Abstract

      Purpose

      Application of MRF to evaluate the feasibility of 2D Dixon blurring-corrected MRF (2DDb-cMRF) to differentiate breast cancer (BC) from normal fibroglandular tissue (FGT).

      Methods

      Prospective study on 14 patients with unilateral BC on 1.5 T system/axial T2w-TSE sequence, 2DDb-cMRF, B1 map, dynamic contrast-enhanced (DCE) T1-w GE-series. Mean T1 and T2 values and standard deviations were computed in the BC-/FGT-ROI on pre-/post-contrast MRF-maps and their differences were tested by two-tailed student t-test.
      Accuracy and repeatability of MRF were evaluated in a phantom experiment with gelatin with Primovist surrounded by fat.
      The T1 reduction between pre-/post-contrast MRF-maps was correlated to DCE signal enhancement in the last image post-contrast through the Pearson´s correlation coefficient (r) and for the phantom validation experiment through the Lin’s concordance correlation coefficient (CCC).
      Visual evaluation of cancers on MRF-Maps was performed by rating each MRF-Map by 3 radiologists.

      Results

      T1- and T2-MRF values of BC vs. FGT were for T1 and T2 pre-contrast respectively: 1147 ± 1 ms vs. 1052 ± 9 ms (p = 0.007) and 83 ± 1 ms vs. 73 ± 1 ms (p = 0.03); post-contrast respectively: 367.3 ± 121.5 ms vs. 690.3 ± 200.3 ms (p = 0.0005) and 76.9 ± 11.5 ms vs. 69.8 ± 15.2 ms (p = 0.12). r was positive (FGT r = 0.7; BC r = 0.6). CCC was 0.999 for T1 and 0.994 for T2. In the T1- and T2-MRF-Maps before contrast respectively (7,7,8)/14 and (5,9,8)/14 cancers were visible to the readers; afterwards, (11,12,12)/14 and (5,6,11)/14.

      Conclusions

      MRF is promising for distinction between BC and FGT as well as for analyzing pre-/post-contrast T1 changes. However, its potential for differential diagnosis warrants further studies.

      Keywords

      1. Introduction

      Dynamic contrast enhanced (DCE)-MRI is currently the most sensitive diagnostic tool for the detection of breast cancer [
      • Kuhl C.K.
      Current status of breast MR imaging: Part 2. Clinical applications.
      ,
      • Bennani-Baiti B.
      • Bennani-Baiti N.
      • Baltzer P.A.
      Diagnostic performance of breast magnetic resonance imaging in non-calcified equivocal breast findings: results from a systematic review and meta-analysis.
      ,
      • Mann R.M.
      • Cho N.
      • Moy L.
      Reviews and commentary • State of the art.
      ,
      • Sardanelli F.
      • Podo F.
      • Santoro F.
      • et al.
      Multicenter surveillance of women at high genetic breast cancer risk using mammography, ultrasonography, and contrast-enhanced magnetic resonance imaging (the high breast cancer risk Italian 1 study): final results.
      ,
      • Bakker M.F.
      • de Lange S. v
      • Pijnappel R.M.
      • et al.
      Supplemental MRI screening for women with extremely dense breast tissue.
      ,
      • Comstock C.E.
      • Gatsonis C.
      • Newstead G.M.
      • et al.
      Comparison of abbreviated breast MRI vs digital breast tomosynthesis for breast cancer detection among women with dense breasts undergoing screening.
      ]. However its cost, the use of a contrast agent and the difficulties in interpreting the images for non-experienced radiologists limit the use of breast MRI.
      A tool which promises to solve some these problems is quantitative MRI (qMRI). qMRI allows the reconstruction of parametric maps, assigning to every voxel a quantitative tissue property. These intrinsic tissue properties accessible by MRI are the tissue relaxation times T1 and T2, the proton density, the diffusion coefficient and many more. These data are unique because they provide an absolute, quantitative value, independently form the subjective interpretation of the image, which inevitably relies on the experience of the radiologist, and it could be used for the further development of artificial intelligence in radiology. Moreover they can be obtained without the use of contrast material, which that would aid to overcome the first two problems mentioned above and could be helpful for the development of new native protocols in the future. In the field of breast imaging the most widely used quantitative parameter is the ADC value, obtained through a DWI-sequence [
      • White N.S.
      • McDonald C.R.
      • Farid N.
      • et al.
      Diffusion-weighted imaging in cancer: physical foundations and applications of restriction spectrum imaging.
      ,
      • Charles-Edwards E.M.
      • de Souza N.M.
      Diffusion-weighted magnetic resonance imaging and its application to cancer.
      ,
      • Partridge S.C.
      • Amornsiripanitch N.
      DWI in the assessment of breast lesions.
      ,
      • Chu W.
      • Jin W.
      • Liu D.
      • et al.
      Diffusion-weighted imaging in identifying breast cancer pathological response to neoadjuvant chemotherapy: a metaanalysis.
      ,
      • Partridge S.C.
      • Zhang Z.
      • Newitt D.C.
      • et al.
      Diffusion-weighted MRI findings predict pathologic response in neoadjuvant treatment of breast cancer: the ACRIN 6698 multicenter trial.
      ,
      • Baltzer P.
      • Mann R.M.
      • Iima M.
      • et al.
      Diffusion-weighted imaging of the breast—a consensus and mission statement from the EUSOBI International Breast Diffusion-Weighted Imaging working group.
      ]. Another interesting qMRI method that has been applied in the context of breast imaging is represented by T1 and T2 relaxometry [

      T.E. Merchant, G.R.P. Thelissen, P.W. de Graaf, Nieuwenhuizen CWEA: Acta Radiologica Application of a Mixed Imaging Sequence for MR Imaging Characterization of Human Breast Disease, 2010.

      ,
      • Seo M.
      • Ryu J.K.
      • Jahng G.H.
      • et al.
      Estimation of T2* relaxation time of breast cancer: correlation with clinical, imaging and pathological features.
      ,
      • Liu L.
      • Yin B.
      • Shek K.
      • et al.
      Role of quantitative analysis of T2 relaxation time in differentiating benign from malignant breast lesions.
      ,
      • Manton D.J.
      • Chaturvedi A.
      • Hubbard A.
      • et al.
      Neoadjuvant chemotherapy in breast cancer: early response prediction with quantitative MR imaging and spectroscopy.
      ,
      • Tan P.C.
      • Pickles M.D.
      • Lowry M.
      • Manton D.J.
      • Turnbull L.W.
      Lesion T2 relaxation times and volumes predict the response of malignant breast lesions to neoadjuvant chemotherapy.
      ,
      • Liu L.
      • Yin B.
      • Geng D.Y.
      • Lu Y.P.
      • Peng W.J.
      Changes of T2 relaxation time from neoadjuvant chemotherapy in breast cancer lesions.
      ]; however, relaxometry sequences are often characterized by long acquisition times. Faster multiparametric relaxometry techniques, which permit simultaneous production of T1 and T2 maps, are for example Synthetic MRI (SyMRI) [
      • Jung Y.
      • Gho S.-M.
      • Back S.N.
      • Ha T.
      • Kang D.K.
      • Kim T.H.
      The feasibility of synthetic MRI in breast cancer patients: comparison of T2 relaxation time with multiecho spin echo T2 mapping method.
      ,
      • Meng T.
      • He N.
      • He H.
      • et al.
      The diagnostic performance of quantitative mapping in breast cancer patients: a preliminary study using synthetic MRI.
      ] or Magnetic Resonance Fingerprinting (MRF) [
      • Ma D.
      • Gulani V.
      • Seiberlich N.
      • et al.
      Magnetic resonance fingerprinting.
      ,
      • Chen Y.
      • Panda A.
      • Pahwa S.
      • et al.
      Three-dimensional MR fingerprinting for quantitative breast imaging.
      ].
      MRF consists of three phases: (i) the acquisition of an image series with variable sequence parameters, to which different tissues respond with distinctive signal evolutions, (ii) the simulation of a dictionary of possible signal evolutions, and (iii) the pattern matching which identifies for the acquired signal evolution in a voxel the best matching dictionary entry and hence its T1 and T2 value.
      The use of fingerprinting to differentiate cancer from healthy tissue in breast has been discussed by Chen et al. [
      • Chen Y.
      • Panda A.
      • Pahwa S.
      • et al.
      Three-dimensional MR fingerprinting for quantitative breast imaging.
      ], who tested a 3D MRF technique with fat suppression, and the repeatability and reproducibility of this method has been proved successfully by Panda et al. [
      • Panda A.
      • Chen Y.
      • Ropella‐Panagis K.
      • et al.
      Repeatability and reproducibility of 3D MR fingerprinting relaxometry measurements in normal breast tissue.
      ]. Recently Nolte et al. [
      • Nolte T.
      • Gross-Weege N.
      • Doneva M.
      • et al.
      Spiral blurring correction with water–fat separation for magnetic resonance fingerprinting in the breast.
      ] developed an alternative MRF approach in which they examined 2D MRF with Dixon water–fat separation and a spiral blurring correction in the female breast, validating it in both phantom and healthy volunteers. The blurring correction allowed for a reduction in bias in the quantitative T1 and T2 values due to the blurring of fat into water as well as an improvement in clarity of fat-water interfaces. Moreover, the Dixon water-fat separation allows for separate evaluation of water and fat data without the use for fat suppression pulses.
      The purpose of this study is to evaluate the feasibility of 2D blurring-corrected MRF for breast cancer diagnosis, to investigate whether quantitative features of breast cancer differ from those of normal fibroglandular tissue (FGT) and to assess whether cancers are detectable on the reconstructed MRF parameter maps.

      2. Materials and methods

      2.1 Study design

      This feasibility study includes phantom and in vivo measurements. It was performed in accordance with institutional review board requirements and was conducted over a period of 2 months in a university hospital.

      2.2 MR-Fingerprinting (MRF)

      The MRF sequence acquired 500 undersampled images after an initial 180° inversion pulse, employing the flip angle excitation pattern described by Sommer et al. [

      K. Sommer, T. Amthor, P. Koken, et al., Determination of the optimum pattern length of MRF sequences, in: Proceedings of the 25th Annual Meeting of ISMRM, 2017.

      ] with 500 variable flip angles between 0° and 60°. The sequence is based on a gradient echo (GRE) sequence with gradient spoiling (4·2π over the slice). 3 MRF trains were sequentially acquired, using in- and out-of-phase echo times (TE1/TE2/TE3) = (4.6/6.9/9.2) ms for the (first/second/third) MRF train, respectively, and a constant TR = 21 ms for all MRF trains. In between MRF trains, a delay of 7.5 ms was set to ensure relaxation of the magnetization to equilibrium. A spiral interleaf (7 ms acquisition time) was rotated between consecutive TR intervals by 14.4° to acquire k-space data with an undersampling factor of R = 25. At a slice thickness of 3 mm, acquiring 2 interleaves/image yielded an acceptable homogeneity of the parametric maps and was hence employed throughout the study. A separation of the MRF data into aqueous and fatty tissue fingerprints by a three-point Dixon approach as well as spiral off-resonance deblurring was performed as described by Nolte et al. [
      • Nolte T.
      • Gross-Weege N.
      • Doneva M.
      • et al.
      Spiral blurring correction with water–fat separation for magnetic resonance fingerprinting in the breast.
      ]. The MRF dictionary contained simulated signals with T1 values ranging between 5 and 2000 ms (between 5 and 200 ms in steps of 5 ms, between 210 and 500 ms in steps of 10 ms, between 520 and 2000 ms in steps of 20 ms). T2 values ranged from 2 to 500 ms (between 2 and 100 ms in steps of 2 ms, between 105 and 200 ms in steps of 5 ms, between 210 and 500 ms in steps of 10 ms). B1 correction factors, included into the signal model as a constant multiplicative factor to all flip angles, ranged from 0.7 to 1.3 in steps of 0.1. Parametric maps (MRF-maps) were reconstructed offline by inner product (IP) matching after performing water-fat separation and spiral deblurring. The IP is a value between 0 and 1, where 1 indicates a perfect match between measured and simulated signal evolution. For every voxel, B1 correction was carried out by restricting the matching process to the subset of simulated signal evolutions that had the closest B1 correction factor to the measured one, i.e., that was read from the B1 map. The subsequent analysis of T1 and T2 values was performed solely on the ‘water’ dataset to avoid partial volume artifacts of fat, i.e., fatty tissue fingerprints were not analyzed. To discard non-valid voxels that predominantly contain noise, the temporal mean of the water MRF dataset Iw,mean and of the combined dataset Iwf,mean were calculated and voxels for which Iw, mean (x,y) < 0.3 ∙ max (Iwf,mean) were discarded.

      2.3 Phantom experiments

      2.3.1 MRI acquisition

      The MRF acquisition and post-processing were validated in a phantom consisting of 8 vials with mixtures of gelatin and Primovist in different concentrations surrounded by lard (pig fat). A test-retest experiment with the MRF-sequence as well as a comparison against inversion recovery T1 mapping (10 inversion times TI between 100 and 5000 ms, TR/TE = (10,000/3.6) ms, duration 1 min 40 s per TI) and multi-echo spin echo T2 mapping (30 echoes with a spacing of 35 ms, TR = 10,000 ms, duration 25 min 40 ms) were conducted.

      2.3.2 Data analysis

      For the phantom experiment, mean values within the phantom vials were calculated for the MRF-maps as well as for the reference maps.

      2.3.3 Statistical analysis

      Lin’s concordance correlation coefficient (CCC) was calculated using Excel.

      2.4 In vivo experiments

      2.4.1 Patient cohort

      From February 2019 to March 2019, a total of 17 women participated in our study. We included consecutive women with a known, histologically proven breast cancer, who underwent breast MRI for staging purposes, and who consented to participate. Exclusion criteria were presence of a biopsy clip inside the cancer and previous contralateral mastectomy.

      2.4.2 MRI acquisition

      Each patient received a bilateral breast examination on a 1.5 Tesla MR System (Philips Achieva, Best, The Netherlands) with a 4-channel breast coil (Invivo, Orlando, Florida) in prone position and with immobilization of the breast in the craniocaudal (CC)-direction by the use of two parallel panels (Noras, Würzburg, Germany). All the patients received a standard DCE-MRI protocol, consisting of an axial T2-Turbo spin-echo (TSE) and a coronal T1-TSE sequence, followed by a dynamic DCE-T1-GRE sequence which acquired one image before and four images after the application of 0,1 mmol/kg-BW Gadobutrol (Bayer AG, Germany). The MRF sequence was performed before and after the dynamic sequence. The 2D blurring-corrected MRF is a two-dimensional technique, which so far allows the analysis of one preselected slide only. Therefore, the slice where the lesion was best evident was selected on the basis of the pre-contrast, anatomic/structural T2-TSE images. The location of the tumor was identified on pre-contrast images, which was also compared with the patient's existing ultrasound and mammography images. The MRF sequence was then performed only on this pre-selected slide. In addition to the MRF scan, a B1 map with voxel size and location exactly matching with the MRF scan was acquired using actual flip angle imaging technique [
      • Yarnykh V.L.
      Actual flip-angle imaging in the pulsed steady state: a method for rapid three-dimensional mapping of the transmitted radiofrequency field.
      ]. This was done to correct for in-plane transmit (B1) field inhomogeneities (c.f. paragraph 2 of Material and Methods) that are known to be important in breast MRI [
      • Kuhl C.K.
      • Kooijman H.
      • Gieseke J.
      • et al.
      Effect of B1 inhomogeneity on breast MR imaging at 3.0 T [1].
      ]. Details of the standard breast MRI protocol are provided in Table 1.
      Table 1Details of the multiparametric MRI protocol (TR: Repetition Time; TE: Echo Time; TSE: Turbo spin echo; GE: Gradient echo; NA: not applicable).
      Details of Study Protocol Pulse-Sequence Parameters
      HardwareDescription
      Type of magnet1.5 T Intera (Philips Medical Systems, Best, the Netherlands)
      Surface coilDedicated multielement four-channel breast coil (inVivo, Gainesville, FL)
      Breast immobilizationFixation plates; fixation in craniocaudal direction (Noras Medical Systems, Hoechberg, Germany)
      Type of contrast agentGadobutrol (Bayer Healthcare, Leverkusen, Germany))
      Dose of contrast agent0.1 mmol per kg body weight
      Injection protocol3 mL per second power injection, followed by 20 mL saline
      Pulse-Sequence Protocol
      Pulse-Sequence

      Parameter
      FingerprintingB1 mapT2-Weighted TSET1-weighted GE (Dynamic Series)
      TR/TE20/(4.6/6.9/9.2) ms60/1.3 ms3954/110 ms253/4.5 ms
      Flip angleVarying between 0° and 60°50°90°90°
      OrientationAxialAxialAxialAxial
      Acquisition matrix256 × 256128 × 128512 × 510512 × 512
      Field of view430 mm430 mm380 mm380 mm
      No. of sections11125–31
      Section thickness3 mm3 mm3 mm3 mm
      No. of dynamicsNANANA1 precontrast; 4 postcontrast
      Acquisition time108 s4 min 23 s230 s80 s (per dynamic)

      2.4.3 Data analysis

      For the in vivo experiment, the tumor ROI was delineated on the DCE subtraction images by a radiologist with 1.5 years of experience in breast diagnostic that correspond to the MRF slice using dedicated software (Intellispace, Philips, The Netherlands) and interpolated onto the MRF voxel grid (Matlab, The Mathworks, Massachusetts). The FGT ROI comprised all valid voxels inside the contralateral, healthy breast. Within the tumor and normal FGT ROI, the mean and standard deviation of T1 and T2 values, acquired pre- and post-contrast injection, were calculated using MATLAB. Enhancement rates of the breast cancer and the normal fibroglandular tissue were evaluated by using the same ROI on the DCE images. Enhancement rates in % were calculated, considering the relative intensity increase of the last post-contrast acquisition compared to the pre-contrast image.

      2.4.4 Visual assessment

      The visual assessment of cancers on T1- and T2-MRF-maps for each patient, acquired before and after contrast injection, was performed independently by 3 radiologists (X1–X3) with 4, 6 and more than 10 years of experience. Readers were blinded to the anamnesis of the patients and to the DCE-MRI sequences and had to evaluate one MRF-map at a time in this sequence: T1-MRF map before contrast injection, T2-MRF map before contrast injection, T1-MRF after contrast injection and at the end T2-MRF map after contrast injection. One patient per time was presented. After the visualization of each single MRF-map, the radiologist was asked to assess on which breast the lesion is and then to point exactly at the lesion with an arrow. To that purpose, T1- and T2-MRF-maps were generated by discarding all voxels for which Iw,mean(x,y) < 0.05 ∙ max(Iw,mean) or IP(x,y) < 0.6. The answers of the radiologists were then compared to the acquired DCE-MRI images, which were used as gold standard. A value of 1 was assigned if the localization of the cancer was detectable on the MRF-maps and of − 1 if not. If the radiologist could identify the breast in which the cancer was, but not the lesion, a value of − 1 was equally assigned.

      2.4.5 Statistical analysis

      The differences between the quantitative mean values of T1 and T2 breast cancer vs. FGT, acquired before and after injection of contrast material, were tested for significance by two tailed student t-tests. A p-value less than 0.05 was considered statistically significant. Pearson´s correlation coefficient (r) between the decrease of T1-relaxation time with the enhancement rate in DCE-MRI was calculated considering the last dynamic DCE-MRI image post contrast. All statistical analyses were performed using Excel.

      3. Results

      3.1 Patient cohort

      14 Patients (mean age 56, age range 42–72 years) with a known, unilateral breast cancer were analyzed. Out of 17 data sets that we acquired, 3 had to be discarded because the cancer was not included in the MRF section. The type of cancer was: No special type (NST) in 79 % (11/14)), tubular invasive in 7 % (1/14); ductal carcinoma in situ (DCIS) in 14 % (2/14). Further details on the characteristics of cancers are provided in Table 2.
      Table 2Cancer characteristics in DCE-MRI, their histological findings and MRF T1 and T2 quantitative values pre- and post-contrast injection for both cancers (BC) and for normal fibroglandular tissue (FGT). (NST: No special type; DCIS: Ductal carcinoma in situ; ER: Estrogen receptor; PR: progesterone receptor; HER2neu: human epidermal growth factor receptor 2; BC: breast cancer; FGT: fibroglandular tissue).
      PatientAgeDCE-MRIHistologyMRF pre contrastMRF post contrast
      MRI-ACRSize in mmVisual kineticsT2-signal intensityCancerER/PR/

      HER2neu
      Ki67 %T1/ms BCT1/ms FGTT2/ms BCT2/ms FGTT1/ms cancerT1/ms FGTT2/ms cancerT2/ms FGT
      142226 × 15Wash outHypointenseNST-/-/-40–50 %1238 ± 461161 ± 10787 ± 1076 ± 13335 ± 71733 ± 12170 ± 778 ± 8
      254414 × 10Wash outHypointenseTubular-/-/-Up to 14 %982 ± 451074 ± 10166 ± 666 ± 17342 ± 76771 ± 13964 ± 766 ± 17
      348218 × 15Wash outHyperintenseNST+/-/-95 %1324 ± 721164 ± 131496 ± 1283 ± 15445 ± 120968 ± 13394 ± 1395 ± 19
      463138 × 16Wash outHypointenseNST-/-/-70 %1213 ± 99902 ± 10771 ± 1052 ± 19301 ± 47694 ± 11767 ± 852 ± 24
      564133 × 20Wash outHypointenseNST+/-/-40–50 %1099 ± 92944 ± 24084 ± 1681 ± 34478 ± 96853 ± 23676 ± 1379 ± 24
      668173 × 11PlateauIsointenseNST//1139 ± 47959 ± 61100 ± 966 ± 12284 ± 19756 ± 9386 ± 767 ± 12
      771119 × 11Wash outHypointenseNST+/+/-5–15 %1051 ± 781075 ± 9773 ± 778 ± 21283 ± 32956 ± 11268 ± 779 ± 17
      848322 × 20Wash outHyperintenseNST+/+/-Up to 20 %1271 ± 711172 ± 9289 ± 975 ± 15521 ± 65588 ± 12986 ± 972 ± 13
      972220 × 19ProgressiveIsointenseDCIS+/+/1330 ± 891020 ± 0113 ± 1196 ± 0677 ± 210280 ± 076 ± 1236 ± 0
      1050232 × 15PlateauHypointenseNST+

      lobular
      +/+/-Up to 2 %1011 ± 571176 ± 9373 ± 1875 ± 28304 ± 70782 ± 16090 ± 2072 ± 13
      1164226 × 20Wash outHyperintenseNST+/+/+30 %1189 ± 69962 ± 4977 ± 961 ± 17232 ± 8654 ± 6590 ± 672 ± 13
      1257240 × 14Wash outHypointenseNST+/+/-Up to 15 %903 ± 971009 ± 10768 ± 1077 ± 14273 ± 16331 ± 6760 ± 883 ± 11
      1344320 × 8ProgressiveIsointenseDCIS+

      LIN
      +/+/1207 ± 1111120 ± 10496 ± 1170 ± 14320 ± 14731 ± 13685 ± 676 ± 14
      1447119 × 11Wash outIsointenseNST+/+/-3 %1099 ± 112986 ± 9068 ± 1551 ± 11347 ± 58567 ± 12464 ± 1250 ± 10

      3.2 Phantom validation

      Fig. 1 shows the results of the phantom validation experiment. Between MRF and IR/MESE reference measurements, Lin’s CCC is calculated as 0.999 for T1 and 0.994 for T2, indicating good correspondence between MRF and the reference methods. The MRF test-retest experiment furthermore indicates stability of the MRF measurement at the selected sequence parameters in the phantom experiment.
      Fig. 1
      Fig. 1Validation of the MRF-Dixon sequence in a phantom experiment. The phantom contained 8 vials containing different amounts of Primovist in gelatin, surrounded by pig fat. (a) and (b) show the comparison of the MRF (water) T1 and T2 values to gold standard reference quantitative methods (i.e., inversion recovery (IR) for T1 mapping and multi-echo spin echo (MESE) for T2 mapping. (c) and (d) show the results of the test-retest experiment.

      3.3 Pre-contrast T1 and T2 values of cancerous vs. normal fibroglandular tissue

      The T1 mean value of cancerous tissue vs. that of normal fibroglandular tissue is 1146.9 ± 129.1 ms vs. 1051.6 ± 95.4 ms (p = 0.007). The T2 mean value of cancerous tissue vs. that of normal fibroglandular is 83.8 ± 14.4 ms vs. 72.7 ± 12.4 ms (p = 0.03) (Table 2).

      3.4 Post-contrast T1 and T2 values of cancerous vs. normal fibroglandular tissue

      After contrast injection, the T1 mean value of cancerous tissue vs. that of normal fibroglandular tissue is 367.3 ± 121.5 ms vs. 690.3 ± 200.3 ms (p = 0.0005). The post contrast T2 mean value of cancerous tissue vs. that of normal fibroglandular tissue is 76.9 ± 11.5 ms vs. 69.8 ± 15.2 ms (p = 0.12). All mean values and standard deviations of T1- and T2-MRF-maps before and after contrast injection are provided in Table 2, Table 3.
      Table 3Differences between the T1 mean values and T2 mean values of cancerous tissue vs. normal tissue.
      MRF pre-contrastMRF post-contrast
      Mean of T1/ms mean value ± standard deviationMean of T2/ms mean value ± standard deviationMean of T1/ms mean value ± standard deviationMean of T2/ms mean value ± standard deviation
      BC1146.89 ± 129.1282.77 ± 14.35367.3 ± 121.576.86 ± 11.49
      FGT1051.64 ± 95.3572.69 ± 12.39690.3 ± 200.369.76 ± 15.20
      P-value< 0.01< 0.05< 0.01> 0.05

      3.5 Correlation of the decrease of T1 relaxation time and the increase of signal intensity in DCE-MRI

      T1 relaxation time of cancer in MRF shows a mean decrease of 68 ± 10 % in cancers and a mean decrease of 34 ± 19 % in normal fibroglandular tissue. In DCE-MRI the mean increase in the enhancement rate in the last image post-contrast is 105 ± 34 % for cancers and 21 ± 14 % for FGT.
      The calculated Pearson´s correlation coefficient between the decrease of T1 relaxation time and the increase of the enhancement rate was a positive value both for FGT (r = 0.7) and for cancer (r = 0.6).

      3.6 Visual evaluation of cancer on T1- and T2-MRF-maps before vs. after contrast injection

      Before contrast injection, X1, X2 and X3 detected 50 % (7/14), 50 % (7/14) and 57 % (8/14) of the cancers on the T1-MRF-map; on the T2-MRF map 36 % (5/14), 64 % (9/14) and 57 % (8/14) were respectively correctly recognized. After the injection of a contrast agent, X1, X2 and X3 detected 79 % (11/14), 86% (12/14) and 86 % (12/14) of cancers on the T1-MRF-map; on the T2-MRF-maps 36 % (5/14), 43 % (6/14) and 79 % (11/14) of cancers were correctly identified, respectively. Fig. 2 and 3 show the MRF maps pre and post contrast injection for two example cases. Information in details in Table 4.
      Fig. 2
      Fig. 2Patient 4 in the Table 3, Table 4: 63-year old postmenopausal patient with mammographically suspicious architectural distortion in the right breast (BIRADS 6). On the T2-weighted TSE image, a spiculated mass with architectural distortions is detectable in the right upper outer quadrant (a). The lesion showed a strong enhancement with bi-phasic dynamic curve in DCE-MRI subtraction images (b). In the pre-contrast T1- (c) and T2-MRF-map (d) the lesion was recognized by all the readers. Also in the post-contrast T1- (e) and T2-MRF-map (f) the cancer was detected by all the readers.
      Fig. 3
      Fig. 3Patient 7 in the Table 3, Table 4: 73-year old postmenopausal patient with mammographically suspicious architectural distortion in the right breast (BIRADS 6). On the T2-weighted TSE image, a spiculated mass with architectural distortions is detectable at the 12 o´clock position in the right breast (a). The lesion showed a strong enhancement with bi-phasic dynamic curve in DCE-MRI subtraction images (b). In the pre-contrast T1- (c) and T2-MRF-map (d) the lesion was not recognized by any of the readers. In the post-contrast T1-MRF-map (e) the cancer was detected by all the readers. In the post-contrast T2-MRF-map (f) the cancer was detected only by the reader with the longest experience.
      Table 4Visual assessment of each MRF-Maps for each reader (X1–X3): 1 was assigned if the cancer was detectable on the MRF image and − 1 if the cancer was not visible.
      MRF maps pre-contrastMRF maps post-contrast
      MRF-T1 mapMRF-T2 mapMRF-T1 mapMRF-T2 map
      PatientX1X2X3X1X2X3X1X2X3X1X2X3
      1111111111-1-1-1
      2-1-1-1-1-1-1111-1-1-1
      3111-111111-1-11
      4111111111111
      5-1-1-1-1-1-1111-1-11
      6-1-1-1-11-1111-1-11
      7-1-1-1-1-1-1111-1-11
      81-1-1-1-1-1-1-11-1-11
      9111111111111
      10-111-1-11111-111
      111-11111111111
      12-1-1-1-11-1-1-1-111-1
      13111111111111
      14-111-111-11-1-1-11

      4. Discussion

      In this study, MRF maps were acquired both pre and post contrast. Our MRF data on pre-contrast T1- and T2-MRF-maps showed significantly longer T1 and T2 values for breast cancer as compared to FGT. After injection of a clinical Gd-based contrast agent, the expected much stronger reduction of T1 relaxation time of breast cancer was observed, whereas, and again as expected, the contrast injection had no significant effect on the respective measured T2 relaxation times. The calculated correlation coefficient between the increase of signal intensity of cancers and FGT in DCE-MRI (last dynamic) and the decrease of T1-MRF relaxation time showed a linear correlation (r = 0.6 and r = 0.7, respectively). When it comes to post contrast relaxometry, timing aspects become important as the concentration of contrast agent in the tissue changes over time with characteristic enhancement kinetics [
      • Kuhl C.K.
      • Mielcareck P.
      • Klaschik S.
      • et al.
      Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions?.
      ]. We chose to compare the MRF values with the enhancement rates obtained from the last dynamic scan, as it was acquired just before the post contrast T1-MRF-map. It should be emphasized, that our post-contrast MRF acquisition observes the enhancement, as it was acquired after the dynamic series. Even during post contrast MRF acquisition, further wash-out of the contrast medium can take place. It is not surprising to find a higher correlation coefficient for FGT as for cancer, considering that the concentration of contrast agent in cancers might undergo more important changes than the one in FGT, which on the contrary reaches a more stable plateau. Generally speaking, a fast relaxometry technique is hence advantageous. In this regard, our MRF technique needs further technical development. While our post contrast MRF analysis was mainly intended for validation purposes, the assessment of lesion differentiation by absolute T1 changes could be an interesting topic for future research.
      As expected, cancers were best visible on the T1-MRF-map acquired after the injection of contrast material, where between 79 % (11/14) and 86 % (12/14) of cancers were recognized. Surprisingly, for each radiologist at least 50 % (7/14) of the cancers were detectable also on the pre contrast T1-MRF-map. The cancer detectability on the T2-MRF-map varies between 36 % (5/14) and 64 % (9/14), showing a greater, and probably experience dependent, inter-reader variability.
      An aspect that may affect the visibility of T1 and T2 differences in general is the choice of the color map [
      • Stoelzle M.
      • Stein L.
      Rainbow color map distorts and misleads research in hydrology–guidance for better visualizations and science communication.
      ]. Indeed, standardization of colormaps for qMRI is an active topic of debate and needs to be addressed by the community.
      Since the first description of magnetic resonance fingerprinting in 2013 by Ma et al. [
      • Ma D.
      • Gulani V.
      • Seiberlich N.
      • et al.
      Magnetic resonance fingerprinting.
      ], several studies focused on its possible clinical applications and possible technical improvements [

      M.E. Poorman, M.N. Martin, D. Ma, et al., Magnetic Resonance Fingerprinting Part 1: Potential Uses, Current Challenges, and Recommendations, 2019.

      ,

      D.F. Mcgivney, M.E. Poorman, K.E. Keenan, et al., Part 2: Technique and Directions, 2019, pp. 1–15.

      ]. However, MRF poses still technical challenges, especially if used for applications in the breast, where MRF with spiral readout is complicated by the high content of fat tissue that causes off-resonance blurring of the signal. This issue is even more important for spiral readout trajectories with longer acquisition times [
      • Panda A.
      • Chen Y.
      • Ropella‐Panagis K.
      • et al.
      Repeatability and reproducibility of 3D MR fingerprinting relaxometry measurements in normal breast tissue.
      ]. A possible solution has been proposed by Chen et al. [
      • Chen Y.
      • Panda A.
      • Pahwa S.
      • et al.
      Three-dimensional MR fingerprinting for quantitative breast imaging.
      ], with the application of a fat-suppression pulses to a multi-slice MRF sequence. Nevertheless, the main drawbacks of the suppression of fat [
      • Delfaut E.M.
      • Beltran J.
      • Johnson G.
      • et al.
      Fat suppression in MR imaging: techniques and pitfalls.
      ] signal are that other essential diagnostic information might be lost and that there is a reduced signal-to-noise ratio. By way of contrast, the technique used in our study separates water and fat signal based on a 3-point Dixon approach, which is robust to background field inhomogeneity and does, in principle, permit the analysis of both water and fat signals thus allowing the computation of images with less artifacts and better anatomical definition in comparison to other spectral-selective fat suppression techniques [
      • Lin C.
      • Rogers C.
      • Majidi S.
      Fat suppression techniques in breast magnetic resonance imaging: a critical comparison and state of the art.
      ].
      This MRF-Dixon technique has not been evaluated in the breast at 3 T yet; however, other MRF studies employing deblurring approaches suggest robustness of the technique at high field strength [

      T. Nolte, D. Truhn, N. Gross-Weege, et al., Undersampled spiral magnetic resonance fingerprinting with water and fat blurring correction, in: Proceedings of the 26th Annual Meeting of ISMRM Paris, France.

      ,
      • Koolstra K.
      • Webb A.G.
      • Veeger T.T.
      • et al.
      Water–fat separation in spiral magnetic resonance fingerprinting for high temporal resolution tissue relaxation time quantification in muscle.
      ,
      • Ostenson J.
      • Robison R.K.
      • Zwart N.R.
      • et al.
      Multi-frequency interpolation in spiral magnetic resonance fingerprinting for correction of off-resonance blurring.
      ]. Future technical developments should work towards an accelerated scan time per slice to enable whole breast coverage in clinically feasible scan times.
      In particular, our analysis showed that pre-contrast T1 and T2 relaxation values were significantly longer than those of FGT. Although this outcome is in contrast with the observed T2 hypointensity of most luminal carcinoma [
      • Ab Mumin N.
      • Hamid M.T.R.
      • Wong J.H.D.
      • et al.
      Magnetic resonance imaging phenotypes of breast cancer molecular subtypes: a systematic review.
      ] in normal DCE-MRI protocols, our results are concordant with the ones of Chen et al. [
      • Chen Y.
      • Panda A.
      • Pahwa S.
      • et al.
      Three-dimensional MR fingerprinting for quantitative breast imaging.
      ]. Overall, the T2 relaxation times observed in our cohort were longer than those reported by Chen et al. This can be explained by the influence of the non-ideal slice excitation profile that affects 2D (MRF) sequences [
      • Ma D.
      • Coppo S.
      • Chen Y.
      • et al.
      Slice profile and B1 corrections in 2D magnetic resonance fingerprinting.
      ], but not the 3D sequence used by Chen et al. The discrepancy between T2 hypointensity and longer pre-contrast T2 values measured by MRF of breast cancer with respect to FGT might be attributable to flow effects on the T2w fast spin echo sequence, hypothesis that however demands for further investigation.
      The idea behind MRF is tissue identification based on a possibly unique spectrum (i.e., fingerprint of MR-physical properties). In this study, we found significantly different changes in the T1 and T2 relaxation times of tumor and healthy FGT. The fact that at least 50 % of breast cancer could be visually detected on the T1 map, which are not normally visible on the native T1 images, point towards a potential of MRF as an additional sequence for the development of a MRI diagnostic protocol without the use of contrast material. However, before its use in clinical practice, future MRF studies should investigate differences in the relaxation times of malignant vs. benign lesions, ideally in large patient collectives. If the resulting relaxation times show potential to non-invasively classify different types of lesions, especially the enhancing ones, this could help in increasing the specificity of breast MRI and help reduce the number of false-positive findings in regular breast DCE.
      Nevertheless, this study has its limitations. First of all, due to its feasibility character, the number of patients is small. Especially with respect to differential diagnosis, a much larger patient cohort will need to be analyzed. Secondly, the use of MRF for screening purposes is unrealistic at this stage due to the long time needed to acquire the images of more than 100 s per slice. Indeed, the major drawback of our current approach is the fact that we had to pre-select the area where the MRF pulse sequence was obtained based on the pre-contrast images. In most MR images, the cancer can only be clearly seen after contrast injection. So in cases where the MRF sections were placed at an area that – based on the post-contrast images – included only the periphery of the cancer and the surrounding DCIS, the difference between the absolute values of the cancer vs. fibroglandular tissue was minimal, such as in patient 11, or even reversed, as in the patients 2, 7 and 12.
      Thirdly, the pre-selection of a single slice leads to the impossibility of analyzing the whole breast. Since breast cancers are often multifocal or even bilateral, this represents a major disadvantage of this procedure, which therefore has to be further developed before its application in the clinical practice.
      Concerning the visual analysis of the MRF-maps, it is worth to underline that the radiologists were aware that each patient had a cancer, which constitutes an important bias for the visual assessment of MRF-maps.

      5. Conclusion

      In conclusion, MRF is a promising method to distinguish between breast cancer and healthy fibroglandular tissue based on pre-contrast relaxation times. Pre- and post-contrast T1 changes can potentially give additional insight in the character of a lesion. However, the potential of MRF for differential diagnosis, especially to distinguish malignant from benign lesions, needs to be further investigated.

      CRediT authorship contribution statement

      Eloisa Zanderigo: Conceptualization, Methodology, Investigation, Data curation, Visualization, Formal analysis, Funding acquisition, Writing - original draft, Writing - review & editing. Luisa Huck: Investigation. Martina Distelmaier: Investigation. Ebba Dethlefsen: Investigation. Mirjiam Maywald: Data curation, Investigation, Methodology. Daniel Truhn: Software, Writing - review & editing. Timm Dirrichs: Investigation. Mariya Doneva: Software, Supervision. Volkmar Schulz: Conceptualization, Funding acquisition, Supervision, Writing - review & editing. Christiane K. Kuhl: Conceptualization, Methodology, Resources, Supervision, Project administration, Funding acquisition, Writing - review & editing. Teresa Nolte: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Supervision, Writing - review & editing.

      Grant Support

      This project has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under Grant agreement no. 667211 .
      We acknowledge support by Deutsche Forschungsgemeinschaft and Open Access Publishing Fund of University of Tübingen.

      Declaration of Competing Interest

      The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: I would like to mention the following conflict of interest: my co-author Mariya Doneva is employed by Philips Research Europe and my co-author Volkmar Schulz is a CEO of Hyperion Hybrid Imaging Systems GmbH.

      Acknowledgments

      N.a.

      Ethical statement

      Ethic number: Ethikvotum 205-17.
      Who issued the approval: The ethic commission (Ethikkommission) of RWTH.
      When: 15.08.2017.

      References

        • Kuhl C.K.
        Current status of breast MR imaging: Part 2. Clinical applications.
        Radiology. 2007; 244: 672-691
        • Bennani-Baiti B.
        • Bennani-Baiti N.
        • Baltzer P.A.
        Diagnostic performance of breast magnetic resonance imaging in non-calcified equivocal breast findings: results from a systematic review and meta-analysis.
        PLoS One. 2016; 11e0160346
        • Mann R.M.
        • Cho N.
        • Moy L.
        Reviews and commentary • State of the art.
        Radiology. 2019; 292: 520-536
        • Sardanelli F.
        • Podo F.
        • Santoro F.
        • et al.
        Multicenter surveillance of women at high genetic breast cancer risk using mammography, ultrasonography, and contrast-enhanced magnetic resonance imaging (the high breast cancer risk Italian 1 study): final results.
        Invest. Radiol. 2011; 46: 94-105
        • Bakker M.F.
        • de Lange S. v
        • Pijnappel R.M.
        • et al.
        Supplemental MRI screening for women with extremely dense breast tissue.
        N. Engl. J. Med. 2019; 381: 2091-2102
        • Comstock C.E.
        • Gatsonis C.
        • Newstead G.M.
        • et al.
        Comparison of abbreviated breast MRI vs digital breast tomosynthesis for breast cancer detection among women with dense breasts undergoing screening.
        JAMA - J. Am. Med. Assoc. 2020; 323: 746-756
        • White N.S.
        • McDonald C.R.
        • Farid N.
        • et al.
        Diffusion-weighted imaging in cancer: physical foundations and applications of restriction spectrum imaging.
        Cancer Res. 2014; : 4638-4652
        • Charles-Edwards E.M.
        • de Souza N.M.
        Diffusion-weighted magnetic resonance imaging and its application to cancer.
        Cancer Imaging. 2006; 6: 135-143
        • Partridge S.C.
        • Amornsiripanitch N.
        DWI in the assessment of breast lesions.
        Top. Magn. Reson. Imaging. 2017; : 201-209
        • Chu W.
        • Jin W.
        • Liu D.
        • et al.
        Diffusion-weighted imaging in identifying breast cancer pathological response to neoadjuvant chemotherapy: a metaanalysis.
        Oncotarget. 2018; 9: 7088-7100
        • Partridge S.C.
        • Zhang Z.
        • Newitt D.C.
        • et al.
        Diffusion-weighted MRI findings predict pathologic response in neoadjuvant treatment of breast cancer: the ACRIN 6698 multicenter trial.
        Radiology. 2018; 289: 618-627
        • Baltzer P.
        • Mann R.M.
        • Iima M.
        • et al.
        Diffusion-weighted imaging of the breast—a consensus and mission statement from the EUSOBI International Breast Diffusion-Weighted Imaging working group.
        Eur. Radiol. 2020; 30: 1436-1450
      1. T.E. Merchant, G.R.P. Thelissen, P.W. de Graaf, Nieuwenhuizen CWEA: Acta Radiologica Application of a Mixed Imaging Sequence for MR Imaging Characterization of Human Breast Disease, 2010.

        • Seo M.
        • Ryu J.K.
        • Jahng G.H.
        • et al.
        Estimation of T2* relaxation time of breast cancer: correlation with clinical, imaging and pathological features.
        Korean J. Radiol. 2017; : 238-248
        • Liu L.
        • Yin B.
        • Shek K.
        • et al.
        Role of quantitative analysis of T2 relaxation time in differentiating benign from malignant breast lesions.
        J. Int. Med. Res. 2018; 46: 1928-1935
        • Manton D.J.
        • Chaturvedi A.
        • Hubbard A.
        • et al.
        Neoadjuvant chemotherapy in breast cancer: early response prediction with quantitative MR imaging and spectroscopy.
        Br. J. Cancer. 2006; 94: 427-435
        • Tan P.C.
        • Pickles M.D.
        • Lowry M.
        • Manton D.J.
        • Turnbull L.W.
        Lesion T2 relaxation times and volumes predict the response of malignant breast lesions to neoadjuvant chemotherapy.
        Magn. Reson. Imaging. 2008; 26: 26-34
        • Liu L.
        • Yin B.
        • Geng D.Y.
        • Lu Y.P.
        • Peng W.J.
        Changes of T2 relaxation time from neoadjuvant chemotherapy in breast cancer lesions.
        Iran. J. Radiol. 2016; : 13
        • Jung Y.
        • Gho S.-M.
        • Back S.N.
        • Ha T.
        • Kang D.K.
        • Kim T.H.
        The feasibility of synthetic MRI in breast cancer patients: comparison of T2 relaxation time with multiecho spin echo T2 mapping method.
        Br. J. Radiol. 2018; 92: 20180479
        • Meng T.
        • He N.
        • He H.
        • et al.
        The diagnostic performance of quantitative mapping in breast cancer patients: a preliminary study using synthetic MRI.
        Cancer Imaging. 2020; 20: 88
        • Ma D.
        • Gulani V.
        • Seiberlich N.
        • et al.
        Magnetic resonance fingerprinting.
        Nature. 2013; 495: 187-192
        • Chen Y.
        • Panda A.
        • Pahwa S.
        • et al.
        Three-dimensional MR fingerprinting for quantitative breast imaging.
        Radiology. 2019; 290: 33-40
        • Panda A.
        • Chen Y.
        • Ropella‐Panagis K.
        • et al.
        Repeatability and reproducibility of 3D MR fingerprinting relaxometry measurements in normal breast tissue.
        J. Magn. Reson. Imaging. 2019; 50: 1133-1143
        • Nolte T.
        • Gross-Weege N.
        • Doneva M.
        • et al.
        Spiral blurring correction with water–fat separation for magnetic resonance fingerprinting in the breast.
        Magn. Reson. Med. 2020; 83: 1192-1207
      2. K. Sommer, T. Amthor, P. Koken, et al., Determination of the optimum pattern length of MRF sequences, in: Proceedings of the 25th Annual Meeting of ISMRM, 2017.

        • Yarnykh V.L.
        Actual flip-angle imaging in the pulsed steady state: a method for rapid three-dimensional mapping of the transmitted radiofrequency field.
        Magn. Reson. Med. 2007; 57: 192-200
        • Kuhl C.K.
        • Kooijman H.
        • Gieseke J.
        • et al.
        Effect of B1 inhomogeneity on breast MR imaging at 3.0 T [1].
        Radiology. 2007; : 929-930
        • Kuhl C.K.
        • Mielcareck P.
        • Klaschik S.
        • et al.
        Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions?.
        Radiology. 1999; 211: S.101-110
        • Stoelzle M.
        • Stein L.
        Rainbow color map distorts and misleads research in hydrology–guidance for better visualizations and science communication.
        Hydrol. Earth Syst. Sci. 2021; 25: S. 4549-4565
      3. M.E. Poorman, M.N. Martin, D. Ma, et al., Magnetic Resonance Fingerprinting Part 1: Potential Uses, Current Challenges, and Recommendations, 2019.

      4. D.F. Mcgivney, M.E. Poorman, K.E. Keenan, et al., Part 2: Technique and Directions, 2019, pp. 1–15.

        • Delfaut E.M.
        • Beltran J.
        • Johnson G.
        • et al.
        Fat suppression in MR imaging: techniques and pitfalls.
        Radiographics. 1999; 19: 373-382
        • Lin C.
        • Rogers C.
        • Majidi S.
        Fat suppression techniques in breast magnetic resonance imaging: a critical comparison and state of the art.
        Rep. Med. Imaging. 2015; 8: 37
      5. T. Nolte, D. Truhn, N. Gross-Weege, et al., Undersampled spiral magnetic resonance fingerprinting with water and fat blurring correction, in: Proceedings of the 26th Annual Meeting of ISMRM Paris, France.

        • Koolstra K.
        • Webb A.G.
        • Veeger T.T.
        • et al.
        Water–fat separation in spiral magnetic resonance fingerprinting for high temporal resolution tissue relaxation time quantification in muscle.
        Magn. Reson. Med. 2020; 84: 646-662
        • Ostenson J.
        • Robison R.K.
        • Zwart N.R.
        • et al.
        Multi-frequency interpolation in spiral magnetic resonance fingerprinting for correction of off-resonance blurring.
        Magn. Reson. Imaging. 2017; 41: S.63-S.72
        • Ab Mumin N.
        • Hamid M.T.R.
        • Wong J.H.D.
        • et al.
        Magnetic resonance imaging phenotypes of breast cancer molecular subtypes: a systematic review.
        Acad. Radiol. 2021;
        • Ma D.
        • Coppo S.
        • Chen Y.
        • et al.
        Slice profile and B1 corrections in 2D magnetic resonance fingerprinting.
        Magn. Reson. Med. 2017; 78: 1781-1789