18 The effect of gadolinium-based contrast agent administration on magnetic resonance fingerprinting-based T1 relaxometry in patients with prostate cancer Nikita Sushentsev1†, MD, Joshua D Kaggie1†, PhD, Guido Buonincontri2, PhD, Rolf F Schulte3, PhD, Martin J Graves1, PhD, Vincent J Gnanapragasam4,5,6, PhD, Tristan Barrett1,7*, MD 1Department of Radiology, Addenbrooke’s Hospital and University of Cambridge, Cambridge, UK 2IMAGO7 Foundation, Pisa, Italy 3GE Healthcare, Munich, Germany 4Department of Urology, Addenbrooke’s Hospital, Cambridge, UK 5Academic Urology Group, Department of Surgery & Oncology, University of Cambridge, Cambridge, UK 6Cambridge Urology Translational Research and Clinical Trials Office, University of Cambridge, Cambridge, UK 7CamPARI Prostate Cancer Group, Addenbrooke’s Hospital and University of Cambridge, Cambridge, UK † These authors contributed equally to this work *Corresponding author: Dr Tristan Barrett (e-mail: tb507@medschl.cam.ac.uk) Department of Radiology University of Cambridge School of Clinical Medicine Box 218, Cambridge Biomedical Campus Cambridge, CB2 0QQ, United Kingdom Magnetic resonance fingerprinting (MRF) is a rapidly developing fast quantitative mapping technique able to produce multiple property maps with reduced sensitivity to motion. MRF has shown promise in improving the diagnosis of clinically significant prostate cancer but requires further validation as part of a prostate multiparametric (mp) MRI protocol. mpMRI protocol mandates the inclusion of dynamic contrast enhanced (DCE) imaging, known for its significant T1 shortening effect. MRF could be used to measure both pre- and post-contrast T1 values, but its utility must be assessed. In this proof-of-concept study, we sought to evaluate the variation in MRF T1 measurements post gadolinium-based contrast agent (GBCA) injection and the utility of such T1 measurements to differentiate peripheral and transition zone tumours from normal prostatic tissue. We found that the T1 variation in all tissues increased considerably post-GBCA following the expected significant T1 shortening effect, compromising the ability of MRF T1 to identify transition zone lesions. We, therefore, recommend performing MRF T1 prior to DCE imaging to maintain its benefit for improving detection of both peripheral and transition zone lesions while reducing additional scanning time. Demonstrating the effect of GBCA on MRF T1 relaxometry in patients also paves the way for future clinical studies investigating the added value of post-GBCA MRF in PCa, including its dynamic analysis as in DCE-MRF. Introduction Prostate cancer (PCa) is the second commonest male malignancy worldwide with multiparametric (mp) MRI now recommended by major European and American guidelines as the first-line investigation for patients with suspected early stage disease1–4. The current Prostate Imaging Reporting and Data System (PI-RADS) guidelines only incorporate qualitative measures for interpretation, however, quantitative metrics have been suggested as a means of reducing the considerable interobserver variation of PI-RADS evaluation, shortening the learning curve of mpMRI, and improving diagnostic performance5–8. Magnetic resonance fingerprinting (MRF) is a quantitative technique able to simultaneously generate multiple inherently spatially registered property maps (e.g. T1, T2, apparent proton-density). Quantitative mapping provided by MRF has demonstrated high reproducibility between centers over standard T1 or T2 mapping techniques that can have a system dependence9,10. These maps can be obtained in the presence of motion while being acquired within imaging times comparable to or faster than conventional mapping techniques11,12. The described features of MRF present a particular interest for cancer imaging, where fast and robust quantitative characterization of tissue biology can add value to routinely used qualitative measures for image assessment12. MRF has shown promise for identifying both peripheral zone (PZ) and transition zone (TZ) prostate lesions, demonstrating added value to standard mpMRI sequences for differentiating between indolent and clinically significant disease13–15. Further prospective validation of MRF requires additional evaluation of how it can be incorporated into a standard clinical prostate mpMRI protocol, which includes dynamic contrast enhanced (DCE) imaging3. Gadolinium is known to have a preferential T1 shortening effect at low doses, which was confirmed for MRF in pre-clinical studies involving a murine glioblastoma model16,17. However, no attempts have been made to investigate the impact of gadolinium-based contrast agents (GBCA) on MRF-based T1 relaxometry within patients in the clinical setting. For prostate imaging, understanding the potential added value and robustness of post-GBCA MRF is of considerable practical interest for two major reasons. Firstly, should MRF be incorporated into the clinical mpMRI protocol, the decision on its running order in relation to DCE-MRI should be evidenced and balanced against the increasing trend towards reducing scanning time18–21. Secondly, post-GBCA MRF T1 mapping may improve the performance of DCE-MRI in TZ lesions, where it is currently of limited use in the context of adequate T2-weighted imaging and diffusion-weighted imaging22,23. Finally, investigating the robustness of MRF T1 relaxometry post GBCA in the clinical setting would also be relevant to cancers located in other anatomical regions where MRF has also shown promising results and the use of GBCA is routine. Therefore, in this proof-of-concept study we sought to evaluate the variation in MRF T1 measurements post GBCA administration and evaluate its impact on the technique’s ability to differentiate between tumor and normal tissue in patients with biopsy-proven PZ and TZ prostate lesions. To ensure the robustness of the MRF technique used, we also validated it against “gold standard” quantitative mapping techniques as part of a phantom study. Results Phantom results Figure 1a-b shows the mean MRF T1 values obtained from the ISMRM/NIST phantom plotted against T1 values obtained using the “gold standard” inversion recovery fast spin echo (IR-FSE) imaging and 3D variable flip angle (VFA) T1 mapping. The results show a strong linear correlation between MRF and IR-FSE (R2 = 0.996, p = < 0.0001, slope = 0.991) and a slightly weaker correlation between MRF and VFA (R2 = 0.975, p = < 0.0001, slope = 0.732). The comparison of slopes of the linear fits (presented alongside y-intercepts in Figure 1a-b) suggests better performance of MRF compared to VFA in the phantom setting. Figure 1c demonstrates the mean phantom MRF T2 values plotted against the values obtained using the “gold standard” multiple spin echo (MSE) T2 mapping; slopes and y-intercepts are also shown. Although still significant, the correlation between the values is considerably lower than for both T1 method comparisons presenting as parabolic rather than linear relationship (R2 = 0.9129, p = <0.0001), suggesting lower reliability of MRF T2 mapping used in this study. Bland-Altman plots were constructed to evaluate the agreement between the aforementioned techniques and are presented in Figure 1d-f. The mean bias and the 95% limits of agreement (LOA) for T1 values obtained using IR-FSE and MRF and VFA and MRF are presented in Figures 1d and 1e, respectively, while the same parameters for T2 values obtained using MSE and MRF are shown in Figure 1f. One data point with the longest T1 was outside the LOA for IR-FSE vs MRF. Patient characteristics The study included 14 patients with biopsy-proven PCa with mean age 70 years (IQR, 67.3-73.5 years), mean PSA 6.29 ng/mL (IQR, 3.8-8.7 ng/mL), with mean time since last biopsy being 16 months (range 4-48 months). A total of 19 MR-visible prostate lesions were included in the analysis, 10 of which were located in the peripheral zone (PZ) and 9 in the transition zone (TZ). Three lesions exhibited intermediate-grade Gleason score 3+4=7 disease (grade group 2) while other lesions harboured low-grade disease with Gleason score of 3+3=6 (grade group 1). In vivo agreement between MRF-, VFA- and MSE-based T1 and T2 relaxometry Summary pre- and post-GBCA MRF- and pre-GBCA VFA- and MSE-based T1 and T2 along with ADC values obtained from all prostate lesions combined (n = 19), PZ (n = 10) and TZ (n = 9) lesions, pooled nPZ and nTZ, internal obturator muscle and subcutaneous fat (n = 14 for all) are presented in Table 1. Bland-Altman analysis showed lower agreement between pre-GBCA MRF- and VFA-based T1 relaxation times compared to the phantom experiment, which is to be expected given the physiological motion in vivo, with the mean bias being 410.5 ms and the 95% LOA ranging between -1171.0 ms and 1192.0 ms as demonstrated in Figure 2a. A similar trend was revealed when the Bland-Altman analysis was used to evaluate the agreement between pre-GBCA MRF- and MSE-based T2 relaxation times, with the mean bias being -281.2 ms and 95% LOA ranging between -727.5 ms and 165.1 ms (Figure 2b). MRF T1 variation post GBCA Table 2 presents coefficients of variation (CVs) calculated for all acquired values to compare their variation in different tissues. Pre-GBCA MRF T1 demonstrated low variation in all tissues except for nTZ, where it reached 25.4%. (Figure 3a) Post-GBCA, MRF T1 variation was above 25% in all tissue types except nPZ, muscle and fat, where CVs remained in the same category as pre-GBCA. A particularly marked, almost five-fold increase in data heterogeneity was observed for MRF T1 values obtained from TZ lesions and, to a lesser extent nTZ, whereas PZ lesions demonstrated only a two-fold increase in CV and only a marginal change in variation was noted in nPZ. (Figure 3b and Table 2) MRF T2 variation post GBCA In line with the outcomes of the Bland-Altman analysis indicating lower robustness of MRF T2 compared to “gold standard” T2 mapping, pre-GBCA MRF T2 values were considerably more variable compared to pre-GBCA MRF T1 values in all tissues except fat, further suggesting low reliability of in vivo MRF T2 mapping used in this study. The variation of MRF T2 also increased considerably post-GBCA reaching 113.7% in nTZ (Table 2). MRF-based T1 relaxometry for differentiating tumour and normal tissue Prior to GBCA administration, a paired t-test revealed significantly shorter MRF T1 values for both peripheral and transition zone lesions when compared to corresponding nPZ and nTZ in the same patients (1640.0 ms ± 368.1 ms vs 2200.0 ms ± 776.5 ms for PZ and 1696.0 ms ± 200.5 ms vs 1966.0 ms ± 315.1 ms for TZ; p = 0.03 and 0.013, respectively) (Figure 4a). In pooled nPZ, MRF T1 relaxation time was significantly longer than in nTZ (2521.0 ms ± 405.9 ms vs 1753.0 ms ± 444.7 ms; p < 0.0001). Post-GBCA MRF T1 remained significantly shorter within peripheral zone lesions compared to the corresponding normal PZ (678.4 ms ± 287.9 ms vs 1317.0 ms ± 219.6 ms; p < 0.0001), however, there was no longer a significant difference in TZ tumours compared to corresponding nTZ (723.8 ms ± 407.3 ms vs 966.4 ms ± 635.5 ms, p = 0.207) (Figure 4b). Pooled nTZ T1 relaxation time was again significantly shorter than those of nPZ (1270 ms ± 224.6 ms vs 723.8 ms ± 407.3 ms; p < 0.0001). Paired t-test showed a significant MRF T1 shortening effect of GBCA in all tissues (Figure 5). The information about the diagnostic utility of MRF- and MSE-based T2, VFA-based T1 and ADC mapping is provided in the Supplementary Information. Discussion This prospective, proof-of-concept study demonstrates the effect of gadolinium-based contrast agent administration on MRF-based T1 relaxometry in the clinical setting. We have shown that GBCA considerably increases MRF T1 variation in both normal and malignant prostate tissues and compromises its diagnostic utility in the transition zone. To our knowledge, this is the first study reporting both pre- and post-GBCA MRF T1 and T2 values as well as a combination of MRF-, VFA- and MSE-based T1 and T2 values obtained from the same patients with prostate cancer. These results will help inform future studies, when MRF may be incorporated into prostate mpMRI protocols as an additional sequence. The observed MRF T1 shortening effect post-GBCA is expected, as gadolinium facilitates both longitudinal and transverse magnetic relaxation, thereby shortening both T1 and T2 of tissues24,25. Other authors also observed a similar trend when measuring gadolinium and dysprosium concentrations in murine glioma models using dual contrast-MRF. 16,26 The observed GBCA-induced increase in MRF T1 heterogeneity varied between tissues. Normal TZ exhibited the greatest variation pre-GBCA, which is expected in this age group, given the high prevalence of benign prostatic hyperplasia (BPH). Marked hypervascularity within BPH nodules may additionally explain the more marked (two-fold) increase in MRF T1 heterogeneity in nTZ tissue following GBCA administration22. The finding of increased MRF T1 variation was even more marked for TZ lesions (almost five-fold), which likely reflects previously reported large variation in their microvascular parameters and explains the inability of MRF T1 to identify TZ lesions post-GBCA27. Although a study by Panda et al. has shown the added value of pre-GBCA MRF T2WI, which is considered the primary sequence for assessment of the TZ28–30, the reduced performance of post-GBCA MRF T1 in assessing TZ lesions should be considered when deciding on its timing in relation to DCE. Conversely, normal PZ, fat and muscle, which can be considered as “control” type tissues given their relatively low vascularity and morphological homogeneity in this age group, maintained low MRF T1 heterogeneity post-GBCA. Hence, understanding the rationale for acquiring MRF T1 prior to GBCA administration would not only ensure its optimal performance for both TZ and PZ assessment but also enable evidenced planning of the overall scanning time, which is critical due to the growing demand on imaging services. We also demonstrated that pre-GBCA MRF T1 relaxation times were significantly lower in cancers compared to normal tissue in both the TZ and PZ of the prostate. These findings align well with previous studies where a combination of MRF and ADC maps worked best for identifying PZ and TZ lesions; our MRF T1 and ADC absolute values are similar to these reported values13–15. Our pre-GBCA T1 values obtained from PZ and TZ lesions are comparable to those reported previously in clinically significant PCa (1628.0 ms ± 344.0 ms vs 2247.0 ms ± 450 ms for PZ lesions and nPZ; 1450 ms ± 110 ms vs 1800 ms ± 150 ms for TZ lesions and nTZ)13,14 and consistent with those reported by Yu et al.15 for low-grade tumours that dominated the sample size in our study (1679.0 ms ± 422 ms). Although we showed low variation of ADC values, significant motion and susceptibility artefacts were not observed in our cohort, which is less representative of real-life clinical practice where motion often hinders assessment of the peripheral zone31–33, for which DWI is the key sequence and where 75-80% of clinically significant lesions are located34,35. Acceptable pre-GBCA variation of MRF T1 coupled with the technique’s intrinsic robustness to motion therefore further support the need for investigating the added value of MRF in prostate imaging, particularly when DWI fails due to artefact. As expected, MRF had considerably shorter maximum scanning time of 3min 40s compared to standard T1 and T2 mapping at 4min 20s and almost 6 min, respectively. The additional benefit of MRF for prostate imaging is that it can be sensitive to a wider range of T1 values (in this work, 1000-2000 ms) than VFA, owing to MRF’s extended range of measured flip angles. Conventional T1 mapping techniques, in turn, can be confounded by T2 effects, the choice of flip angles, and field (B1+) non-uniformity36–38. This reduction of T1 sensitivity to the aforementioned confounders is illustrated by our lower measured heterogeneity for all tissue types when compared with VFA T1 measurements, which may explain superior performance of pre-GBCA MRF T1 in vivo for differentiating tumour versus normal prostatic tissue in both PZ and TZ; better performance of MRF T1 was also demonstrated in the phantom study. Lower heterogeneity was noted for MSE T2 mapping, which is more well-studied in prostate cancer39. This study has several limitations. Firstly, the small sample size may have artificially increased data variation leading to the inability of post-GBCA MRF T1 and pre-GBCA VFA T1 and MSE T2 to identify TZ and PZ tumours, respectively, and did not allow us to quantify grade-dependent variation of MRF T1 post GBCA. Secondly, only patients with low- and intermediate-grade disease were included in this study, which may also have had an impact on the diagnostic utility of both MRF and conventional mapping techniques. However, the inability of post-GBCA MRF T1 to detect any TZ lesions regardless of their Gleason grade indicates a reduction in robustness of the technique, while the added value of pre-GBCA MRF in detecting both TZ and PZ lesions has been demonstrated previously. VFA- and MSE-based T1 and T2 maps were not acquired post-GBCA, however, a head-to-head comparison of post-GBCA MRF and standard mapping techniques was not the purpose of this study. In vivo MRF-derived T2 values were unreliable in this study showing parabolic relationship with MSE-based T2 values, the reliability of which was also undermined by the longest TE being four-fold shorter than the calculated T2. However, gadolinium is known to have a more prominent T1 shortening effect at low doses, which underpinned the focus of this work on MRF T1 relaxometry. Moreover, as evidenced by the comparison with the “gold standard” IR-FSE, MRF T1 mapping used in this study was considered robust. In future work investigating T2 with MRF, we would give strong consideration for methods that increase the signal-to-noise ratio by increasing the slice thickness, voxel sizes, or by performing averaging with additional acquisitions. While averaging for MRF T2 mapping would remain sensitive to motion, we hypothesize that this effect would be reduced due to the pattern matching algorithm of MRF, therefore, supporting the clinical feasibility of both MRF T1 and T2 mapping. Acquiring MSE-based T2 maps with longer TEs to match the known T2 values observed in the prostate could also be helpful to improve their reliability and address the reported parabolic relationship with MRF, which may, however, represent a consistent trend worth further investigation. In conclusion, GBCA administration leads to a considerable increase in MRF T1 variation following the expected significant T1 shortening effect and compromises its ability to detect TZ lesions. Therefore, performing MRF T1 prior to DCE imaging as part of a prostate mpMRI protocol should be considered as a preferred option to retain the technique’s added value for both PZ and TZ lesions and reduce the additional scanning time. Methods Phantom study To evaluate the accuracy of T1 and T2 measurements, MRF and standard relaxation mapping data were obtained from the ISMRM/NIST phantom40. Phantom data were obtained on a 3T MR750 scanner (GE Healthcare, Waukesha, WI, USA) using a 32-channel receiver coil. Regions-of-interest (ROI) were created from the spheres in either the T1 or T2 layer of the phantom. Conventional T1 maps were obtained with an inversion recovery (IR) and variable flip angle (VFA) techniques. A conventional T2 map was obtained with multiple spin echo (MSE) measurements. The field-of-view (FOV) = 260x260 mm2, matrix = 256x256 and slice thickness = 3 mm matched in all techniques. Inversion recovery fast spin echo (IR-FSE) T1 maps from the T1 layer were obtained with inversion times (TI) = 50 ms, 100 ms, 200 ms, 400 ms, 800 ms, 1600 ms, 2400 ms, repetition time (TR) = 8000 ms and echo time (TE) = 13 ms. Multiple 3D GRE sequences were used for T1 mapping using the VFA method with flip angles of 2°, 5°, 8°, 12°, 15°, 18°, 22°, 26°, TR = 10 ms, TE = 1.6 ms. MSE T2 maps were obtained with TR = 600 ms and TEs = 8.2, 16.3, 24.5, 32.6, 40.8, 49.0, 57.1, 65.3 ms. Fitting was performed using a non-linear least squares fit to the IR, VFA and MSE signal equations in Python. MR fingerprinting was also performed as further described in the MR fingerprinting protocol subsection. Patient study All elements of this prospective study were carried out in accordance with the Declaration of Helsinki and were approved by the institutional ethics board (NRES Committee East of England, UK), with written informed consent obtained from all participants. All methods were performed in accordance with the relevant guidelines and regulations. Patients on active surveillance, with MR-visible, biopsy-proven prostate cancer were included in this study. Exclusion criteria included prostate biopsy within the preceding 3 months, presence of pelvic metalwork, or any previous treatment for PCa. Biopsy technique Depending on clinical recommendation, biopsy was performed by either a transrectal or transperineal approach, using MRI/ultrasound fusion. All biopsy procedures were performed by experienced urologists and included 12-24 systematic cores, with 2-4 separate target cores acquired from the MRI defined lesion/s. All targets were defined by radiologists pre-procedure using T2-weighted imaging as the primary and diffusion-weighted imaging as the secondary source images, using the DynaCAD system (InVivo Corp, Orlando, FL) for transrectal and Biopsee software (Oncology Systems Limited, Shrewsbury, UK) for transperineal approaches as previously described41. Multiparametric MRI Patients underwent prostate MRI on a 3T MR750 scanner (GE Healthcare, Waukesha, WI) using a 32-channel receiver coil. Intravenous injection of hyoscine butylbromide (Buscopan, 20 mg/mL; Boehringer, Ingelheim am Rhein, Germany) was administered prior to imaging to reduce peristaltic movement, unless clinically contraindicated. Multiparametric MRI protocol included Axial T1 and multiplanar high-resolution T2-weighted 2D fast recovery FSE (field of view (FOV) 18x18 cm; voxel size 0.35x0.35 mm2; slice thickness 3 mm; gap 0 mm). Diffusion-weighted imaging (DWI) was performed using a spin-echo echo-planar imaging pulse sequence (FOV 28cm; slice thickness 3mm; gap 0mm; b-values: b-150, b-750, and b-1,400 s/mm2) and an additional small FOV (24 cm) b-2,000s/mm2 DWI sequence; apparent diffusion coefficient (ADC) maps were calculated automatically. T1 and T2 mapping were performed with VFA, MSE and MRF prior to dynamic contrast enhancement (DCE). DCE was performed using a standard sequence (FOV 24cm; slice thickness and gap 3mm and 0mm, respectively; temporal resolution 7 seconds) following a bolus of Gadobutrol (Gadovist, 0.1 mmol/kg, Bayer) at 28 seconds via a power injector, at a rate of 3 ml/s (dose 0.1 mmol/kg). Post-GBCA MRF was performed immediately after DCE in all patients. In vivo variable flip angle (VFA) T1 mapping Multiple 3D GRE sequences were used for T1 mapping using the variable flip angle method with flip angles of 2°, 5°, 12°, 20°, and 32°. Each flip angle was acquired in 52 seconds, for a total duration of 4 minutes 20 seconds. Other parameters included: FOV = 36 cm, matrix = 256x256, slices = 52, slice thickness = 3 mm, echo time (TE) = 2.0ms, repetition time (TR) = 15 ms, with 70% sampling. In vivo multiple spin echo (MSE) T2 mapping Multiple echo 2D FSE images were acquired for T2 mapping; FOV 36 cm, matrix 256x256, voxel size 1.4x1.4 mm2, slice thickness 2.5mm, TR 2.6 s, TEs = 8.5 ms, 16.9 ms, 25.4 ms, 42.3 ms, 50.8 ms, 59.2 ms, 67.7 ms, averages = 0.5 (partial k-space), acquisition time 359 s (5 minutes 59 seconds). MR fingerprinting protocol A 2D steady-state-free-precession (SSFP) MRF sequence with inversion preparation was used for T1 and T2 mapping with 979 under-sampled, interleaved spirals for k-space sampling. The maximum gradient strength per spiral was 28 mT/m and the maximum slew rate was 108 T/m/s. The imaging parameters were: FOV = 260x260 mm2, matrix = 256x256, slices = 15-22, slice thickness = 3.0 mm, spacing 1.0 mm, sampling bandwidth = ±250 kHz, slice dephasing = 8π, echo time (TE) = 2.5 ms, repetition time (TR) = 10ms, acquisition time = 9.79 s/slice (maximum scanning time 3 minutes 40 seconds). We used similar flip angle lists to those used in previous works in order to more accurately assess the utility of MRF with standard lists36,42. Axial images were acquired to match the standard acquisition planes for other prostate image assessments. MRF image reconstruction, dictionary simulation and pattern matching parameters are listed in the Supplementary information. Image analysis MRF T1 values for MR-visible lesions (PI-RADS scores 4/5)6, normal peripheral zone (nPZ), normal transition zone (nTZ), subcutaneous abdominal fat and normal internal obturator muscle were calculated from ROIs originally drawn on anatomical T2-weighted images with reference to ADC maps (as presented in Supplementary Figure 1) by a single fellowship-trained uro-radiologist with 12 years’ experience of reporting prostate MRI using the open-source segmentation software ITK-SNAP43. ROIs were then transposed on to the MRF T1 and T2, VFA- and MSE-based T1 and T2 and ADC maps with their size and location being matched to the appropriate FOV parameters and anatomical position of the outlined structures using in-house software developed within Python using the PyQtGraph and PyDicom libraries44. nPZ and nTZ ROIs were drawn in regions with biopsy-confirmed healthy tissue. Statistics In the phantom study, simple linear regression was used to evaluate the relationship between T1 and T2 values obtained using IR-FSE, VFA, MSE and MRF mapping techniques with their agreement assessed using the Bland-Altman analysis. In the patient study, the Shapiro-Wilk test was applied to assess the normality of imaging values with their intergroup comparison performed using either paired or unpaired t-test as appropriate. Planned independent paired comparisons between MRF T1 and T2 values obtained from PZ and TZ lesions versus corresponding nPZ and nTZ were not adjusted for multiplicity; all other post hoc comparisons were adjusted using the Holm-Šidak method with alpha set at 0.05 as advised by an expert biostatistician. The variation of imaging values was evaluated using coefficient of variation (CV); CV of less than 25% indicated acceptable heterogeneity. Agreement between the conventional VFA- and MRF-based T1 relaxation times, treated as single paired measurements, was assessed using the Bland-Altman analysis. All plots and figures were created in Prism 8 (GraphPad Software, San Diego, CA). References: 1. Rawla, P. & Rawla, P. Epidemiology of Prostate Cancer. World J. Oncol. 10, 63–89 (2019). 2. Kasivisvanathan, V. et al. MRI-Targeted or Standard Biopsy for Prostate-Cancer Diagnosis. N. Engl. J. Med. 378, 1767–1777 (2018). 3. EAU Guidelines: Prostate Cancer | Uroweb. https://uroweb.org/guideline/prostate-cancer/. 4. Standard Operating Procedure for Multiparametric Magnetic Resonance Imaging in the Diagnosis, Staging and Management of Prostate Cancer - American Urological Association. https://www.auanet.org/guidelines/mri-of-the-prostate-sop. 5. Gaziev, G. et al. Defining the learning curve for multiparametric magnetic resonance imaging (MRI) of the prostate using MRI-transrectal ultrasonography (TRUS) fusion-guided transperineal prostate biopsies as a validation tool. BJU Int. 117, 80–86 (2016). 6. Turkbey, B. et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. European Urology vol. 76 340–351 (2019). 7. Greer, M. D. et al. Interreader variability of prostate imaging reporting and data system version 2 in detecting and assessing prostate cancer lesions at prostate MRI. Am. J. Roentgenol. 212, 1197–1205 (2019). 8. de Rooij, M. et al. ESUR/ESUI consensus statements on multi-parametric MRI for the detection of clinically significant prostate cancer: quality requirements for image acquisition, interpretation and radiologists’ training. Eur. Radiol. (2020) doi:10.1007/s00330-020-06929-z. 9. Bipin Mehta, B. et al. Magnetic resonance fingerprinting: a technical review. Magn. Reson. Med. 81, 25–46 (2019). 10. McGivney, D. F. et al. Magnetic resonance fingerprinting review part 2: Technique and directions. J. Magn. Reson. Imaging 51, 993–1007 (2020). 11. Ma, D. et al. Magnetic resonance fingerprinting. Nature 495, 187–192 (2013). 12. Poorman, M. E. et al. Magnetic resonance fingerprinting Part 1: Potential uses, current challenges, and recommendations. Journal of Magnetic Resonance Imaging vol. 51 675–692 (2020). 13. Panda, A. et al. Targeted Biopsy Validation of Peripheral Zone Prostate Cancer Characterization With Magnetic Resonance Fingerprinting and Diffusion Mapping. Invest. Radiol. 54, 485–493 (2019). 14. Panda, A. et al. MR fingerprinting and ADC mapping for characterization of lesions in the transition zone of the prostate gland. Radiology 292, 685–694 (2019). 15. Yu, A. C. et al. Development of a Combined MR Fingerprinting and Diffusion Examination for Prostate Cancer. Radiology 283, 729–738 (2017). 16. Gu, Y. et al. Fast magnetic resonance fingerprinting for dynamic contrast-enhanced studies in mice. Magn. Reson. Med. 80, 2681–2690 (2018). 17. Xiao, Y. D. et al. MRI contrast agents: Classification and application (Review). International Journal of Molecular Medicine vol. 38 1319–1326 (2016). 18. Sushentsev, N. et al. The effect of capped biparametric magnetic resonance imaging slots on weekly prostate cancer imaging workload. Br. J. Radiol. 93, 20190929 (2020). 19. Kuhl, C. K. et al. Abbreviated biparametric prostate MR imaging in men with elevated prostate-specific antigen. Radiology 285, 493–505 (2017). 20. van der Leest, M. et al. High Diagnostic Performance of Short Magnetic Resonance Imaging Protocols for Prostate Cancer Detection in Biopsy-naïve Men: The Next Step in Magnetic Resonance Imaging Accessibility. Eur. Urol. 76, 574–581 (2019). 21. Zawaideh, J. P. et al. Diagnostic accuracy of biparametric versus multiparametric prostate MRI: assessment of contrast benefit in clinical practice. Eur. Radiol. 30, (2020). 22. Rosenkrantz, A. B. et al. Transition Zone Prostate Cancer: Revisiting the Role of Multiparametric MRI at 3 T. Am. J. Roentgenol. 204, W266–W272 (2015). 23. Barrett, T., Rajesh, A., Rosenkrantz, A. B., Choyke, P. L. & Turkbey, B. PI-RADS version 2.1: one small step for prostate MRI. Clinical Radiology vol. 74 841–852 (2019). 24. Caravan, P., Ellison, J. J., McMurry, T. J. & Lauffer, R. B. Gadolinium(III) chelates as MRI contrast agents: Structure, dynamics, and applications. Chem. Rev. 99, 2293–2352 (1999). 25. De Leõn-Rodríguez, L. M., Martins, A. F., Pinho, M. C., Rofsky, N. M. & Sherry, A. D. Basic MR relaxation mechanisms and contrast agent design. J. Magn. Reson. Imaging 42, 545–565 (2015). 26. Anderson, C. E. et al. Dynamic, Simultaneous Concentration Mapping of Multiple MRI Contrast Agents with Dual Contrast - Magnetic Resonance Fingerprinting. Sci. Rep. 9, (2019). 27. Van Niekerk, C. G., Witjes, J. A., Barentsz, J. O., Van Der Laak, J. A. W. M. & Hulsbergen-Van De Kaa, C. A. Microvascularity in transition zone prostate tumors resembles normal prostatic tissue. Prostate 73, 467–475 (2013). 28. Padhani, A. R. et al. Dynamic contrast enhanced MRI of prostate cancer: Correlation with morphology and tumour stage, histological grade and PSA. Clin. Radiol. 55, 99–109 (2000). 29. Hansford, B. G. et al. Dynamic contrast-enhanced MR imaging curve-type analysis: Is it helpful in the differentiation of prostate cancer from healthy peripheral zone? Radiology 275, 448–457 (2015). 30. Greer, M. D. et al. Validation of the dominant sequence paradigm and role of dynamic contrast-enhanced imaging in Pi-RADS version 2. Radiology 285, 859–869 (2017). 31. Mazaheri, Y., Vargas, H. A., Nyman, G., Akin, O. & Hricak, H. Image artifacts on prostate diffusion-weighted magnetic resonance imaging: Trade-offs at 1.5 tesla and 3.0 tesla. Acad. Radiol. 20, 1041–1047 (2013). 32. Caglic, I., Hansen, N. L., Slough, R. A., Patterson, A. J. & Barrett, T. Evaluating the effect of rectal distension on prostate multiparametric MRI image quality. Eur. J. Radiol. 90, 174–180 (2017). 33. Slough, R. A., Caglic, I., Hansen, N. L., Patterson, A. J. & Barrett, T. Effect of hyoscine butylbromide on prostate multiparametric MRI anatomical and functional image quality. Clin. Radiol. 73, 216.e9-216.e14 (2018). 34. McNeal, J. E., Redwine, E. A., Freiha, F. S. & Stamey, T. A. Zonal distribution of prostatic adenocarcinoma. Correlation with histologic pattern and direction of spread. Am. J. Surg. Pathol. 12, 897–906 (1988). 35. McNeal, J. E. Cancer volume and site of origin of adenocarcinoma in the prostate: Relationship to local and distant spread. Hum. Pathol. 23, 258–266 (1992). 36. Buonincontri, G. et al. Multi-site repeatability and reproducibility of MR fingerprinting of the healthy brain at 1.5 and 3.0 T. Neuroimage 195, 362–372 (2019). 37. Heule, R., Ganter, C. & Bieri, O. Variable flip angle T1 mapping in the human brain with reduced t2 sensitivity using fast radiofrequency-spoiled gradient echo imaging. Magn. Reson. Med. 75, 1413–1422 (2016). 38. Cheng, H. L. M. & Wright, G. A. Rapid high-rsolution T1 mapping by variable flip angles: Accurate and precise measurements in the presence of radiofrequency field inhomogeneity. Magn. Reson. Med. 55, 566–574 (2006). 39. Hoang Dinh, A. et al. Characterization of prostate cancer using T2 mapping at 3 T: A multi-scanner study. Diagn. Interv. Imaging 96, 365–372 (2015). 40. Jiang, Y. et al. Repeatability of magnetic resonance fingerprinting T1 and T2 estimates assessed using the ISMRM/NIST MRI system phantom. Magn. Reson. Med. 78, 1452–1457 (2017). 41. Barrett, T. et al. Three-year experience of a dedicated prostate mpMRI pre-biopsy programme and effect on timed cancer diagnostic pathways. Clin. Radiol. 74, (2019). 42. Jiang, Y., Ma, D., Seiberlich, N., Gulani, V. & Griswold, M. A. MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magn. Reson. Med. 74, 1621–1631 (2015). 43. Yushkevich, P. A. et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 31, 1116–1128 (2006). 44. Mason, D. SU-E-T-33: Pydicom: An Open Source DICOM Library. Med. Phys. 38, 3493–3493 (2011). Acknowledgments. The authors acknowledge research support from National Institute of Health Research Cambridge Biomedical Research Centre, Cancer Research UK (Cambridge Imaging Centre grant number C197/A16465), Cancer Research UK and the Engineering and Physical Sciences Research Council Imaging Centre in Cambridge and Manchester and the Cambridge Experimental Cancer Medicine Centre. Author contributions statement. NS, JK, and TB formulated the research idea, developed the study design and methodology. NS performed data collection, statistical and computational analysis, wrote the initial draft of the article and coordinated its subsequent revision. JK set up and validated the imaging techniques used in this study, implemented the computing code and supporting algorithms and oversaw research data production and curation. NS and JK jointly conducted the phantom study. GB and RFS provided technical support to JK in setting up the imaging techniques at the core institution thereby serving as external mentors for the core team. MJG facilitated the data collection and provided the required instrumentation and computing resources for both phantom and clinical components of this study. TB supervised and coordinated both the phantom and clinical components of this study and was instrumental in revising the manuscript. VJG curated patient inclusion in this study and coordinated communication with the clinical urology team. All authors contributed considerably to reviewing the presentation of the submitted work. Additional information. Competing interests. The authors declare no competing interests. Data availability. The primary research data is available at https://data.mendeley.com/datasets/g3k7xwjpd3/draft?a=da4b46df-b5b3-4dac-aa8d-84b8c75ee526 Figure legends. Figure 1. Linear regression plots (a-c) and Bland-Altman plots (d-f) comparing MRF T1 values with those obtained using IR-FSE (a, d) and VFA (b, e) mapping techniques and MRF T2 values with those obtained using MSE technique (c, f). Figures (a-c) include captions representing slopes of the linear fits, y-intercepts and coefficients of determination (R2). On figures (d-f), dotted lines represent upper and lower 95% limits of agreement and bold lines represent the mean biases with appropriate captions included. MRF = magnetic resonance fingerprinting, IR-FSE = inversion recovery fast spin echo, VFA = variable flip angle, MSE = multiple spin echo, SD = standard deviation. Figure 2. Bland-Altman plots comparing pre-GBCA in vivo MRF and VFA T1 values (a) and MRF and MSE T2 values (b) obtained from all tissues included in the analysis. Dotted lines represent upper and lower 95% limits of agreement and bold lines represent the mean biases with appropriate captions included. MRF = magnetic resonance fingerprinting, VFA = variable flip angle, MSE = multiple spin echo. Figure 3. Coefficients of variation of pre-gadolinium (a) and post-gadolinium (b) MRF T1 relaxation times obtained from prostate lesions, pooled normal PZ and TZ, internal obturator muscle and subcutaneous abdominal fat. PZ = peripheral zone, TZ = transition zone, MRF = magnetic resonance fingerprinting. Figure 4. Box-and-whisker plots comparing MRF T1 relaxation times obtained from prostate lesions and corresponding normal PZ and TZ before (a) and after (b) gadolinium-based contrast agent administration. Top and bottom of boxes represent 25th and 75th percentiles of data, respectively; line in boxes represents the median value and bars represent minimum and maximum values. MRF = magnetic resonance fingerprinting, PZ = peripheral zone, TZ = transition zone. Figure 5. Box-and-whisker plots comparing MRF T1 relaxation times obtained before and after gadolinium-based contrast agent administration in prostate lesions, pooled normal PZ and TZ, internal obturator muscle and subcutaneous abdominal fat. MRF = magnetic resonance fingerprinting, PZ = peripheral zone, TZ = transition zone. **** p < 0.0001, *** p = 0.0002. Tables. Tissue MRF T1 (pre-Gd), ms MRF T1 (post-Gd), ms MRF T2 (pre-Gd), ms MRF T2 (post-Gd), ms VFA T1 (pre-Gd), ms MSE T2 (pre-Gd), ms ADC, (pre-Gd), mm2/s All lesions 1666.0 ± 294.0 717.8 ± 346.0 443.6 ± 259.0 252.9 ± 167.5 1990.0 ± 522.7 76.6 ± 27.8 0.94 ± 0.17 PZ lesions 1640.0 ± 368.1 678.4 ± 287.9 507.8 ± 292.7 273.4 ± 160.7 1986.0 ± 629.5 89.2 ± 21.8 0.93 ± 0.16 TZ lesions 1696.0 ± 200.5 761.5 ± 414.8 372.2 ± 209.0 230.1 ± 181.5 2002.0 ± 413.7 69.3 ± 9.8 0.90 ± 0.14 Normal PZ 2521.0 ± 405.9 1270 ± 224.6 546.7 ± 294.0 326.5 ± 255.3 2188.0 ± 813.9 139.4 ± 79.12 1.61 ± 0.22 Normal TZ 1753.0 ± 444.7 723.8 ± 407.3 451.0 ± 228.0 237.9 ± 270.5 2118.0 ± 732.1 88.56 ± 11.67 1.27 ± 0.14 Muscle 1542 ± 211.4 1214 ± 149.4 232.2 ± 156.3 180.6 ± 142.4 1813.0 ± 334.9 41.6 ± 2.7 1.17 ± 0.04 Fat 414.3 ± 67.1 327.4 ± 65.4 240.4 ± 41.1 213.7 ± 41.5 1682.0 ± 491.3 115.7 ± 6.6 0.08 ± 0.07 Table 1. Summary values derived from different tissue types from MRF T1 and T2 maps obtained both before and after gadolinium-based contrast agent administration as well as those derived from pre-contrast VFA T1, MSE T2 and ADC maps. MRF = magnetic resonance fingerprinting, VFA = variable flip angle, MSE = multiple spin echo, ADC = apparent diffusion coefficient, PZ = peripheral zone, TZ = transition zone. Tissue MRF T1 (pre-Gd), % MRF T1 (post-Gd), % MRF T2 (pre-Gd), % MRF T2 (post-Gd), % VFA T1 (pre-Gd), % MSE T2 (pre-Gd), % ADC (pre-Gd), % All lesions 17.6 48.2 58.4 66.2 26.3 24.0 17.6 PZ lesions 22.5 42.4 57.6 58.8 31.7 24.5 17.5 TZ lesions 11.8 54.5 56.2 78.9 20.7 14.2 15.6 Normal PZ 16.1 17.7 53.8 78.2 37.2 56.7 13.9 Normal TZ 25.4 56.3 50.6 113.7 34.6 13.2 10.9 Muscle 13.7 12.3 67.3 78.9 18.5 6.6 3.7 Fat 16.2 20.0 17.1 19.4 29.2 5.7 8.0 Table 2. Coefficients of variations of pre-contrast and post-contrast MRF T1 and T2 maps as well as pre-contrast VFA T1, MSE T2 and ADC maps derived from different tissue types. MRF = magnetic resonance fingerprinting, VFA = variable flip angle, MSE = multiple spin echo, ADC = apparent diffusion coefficient, PZ = peripheral zone, TZ = transition zone. Supplemental Data 17 image1.emf image2.emf image3.emf 1 The effect of gadolinium -based contrast agent administration on MR fingerprinting -based T1 relaxometry in patients with prostate cancer Background: Magnetic resonance fingerprinting (MRF) is a novel, fast quantitative mapping technique able to generate multiple property maps that are insensitive to motion. Having shown promise in improving diagnosis of clinically significant prostate cancer (PCa), MR F requires further validation as part of a prostate multiparametric MRI protocol, which includes DCE imaging. Purpose: To evaluate the effect of gadolinium on MRF -based T1 relaxometry to identify the timing of MRF in relation to DCE imaging as part of a prostate multiparametric protocol. Study type: Prospective. Population: 14 patients with biopsy-proven, low/intermediate risk PCa enrolled on active surveillance. Sequence: 3T MRI, MRF, varying flip angle T1 and T2 mapping. Assessment: Regions-of-interest were outlined for prostate lesions, normal peripheral zone (nPZ), normal transition zone (nTZ), subcutaneous abdominal fat and internal obturator muscle with reference to T2WI/ADC. Microsoft_Word_Document.docx 16 The effect of gadolinium-based contrast agent administration on MR fingerprinting-based T1 relaxometry in patients with prostate cancer Background: Magnetic resonance fingerprinting (MRF) is a novel, fast quantitative mapping technique able to generate multiple property maps that are insensitive to motion. Having shown promise in improving diagnosis of clinically significant prostate cancer (PCa), MRF requires further validation as part of a prostate multiparametric MRI protocol, which includes DCE imaging. Purpose: To evaluate the effect of gadolinium on MRF-based T1 relaxometry to identify the timing of MRF in relation to DCE imaging as part of a prostate multiparametric protocol. Study type: Prospective. Population: 14 patients with biopsy-proven, low/intermediate risk PCa enrolled on active surveillance. Sequence: 3T MRI, MRF, varying flip angle T1 and T2 mapping. Assessment: Regions-of-interest were outlined for prostate lesions, normal peripheral zone (nPZ), normal transition zone (nTZ), subcutaneous abdominal fat and internal obturator muscle with reference to T2WI/ADC. Statistics: Paired/unpaired t-test and coefficient of variation (CV). The data are presented as mean ± SD. Results: Post-contrast MRF T1 values were significantly shorter compared to pre-contrast MRF T1 values in all tissues, including all prostate lesions combined (0.72s ± 0.35s vs 1.67s ± 0.29s, respectively), PZ lesions (0.68s ± 0.29s vs 1.64s ± 0.37s), TZ lesions (0.76s ± 0.42s vs 1.70s ± 0.20s), nPZ (1.27s ± 0.23s vs 2.52s ± 0.41s), nTZ (0.73s ± 0.41s vs 1.76s ± 0.45s), muscle (1.21s ± 0.15s vs 1.54s ± 0.21s) and fat (0.33s ± 0.07s vs 0.41s ± 0.07s); P < 0.0005 for all. Pre-contrast MRF T1 values were significantly shorter in both peripheral and transition zone lesions compared to corresponding nPZ and nTZ (1.64s ± 0.37s vs 2.20s ± 0.78s for PZ; 1.70s ± 0.20s vs 1.97s ± 0.32s for TZ; P < 0.05 for both). Post-contrast MRF T1 relaxation times were also significantly shorter in PZ lesions compared to nPZ (0.68s ± 0.29s vs 1.32s ± 0.22s, respectively; P < 0.0001), however, no difference was observed between MRF T1 of TZ lesions and nTZ (0.76s ± 0.42s vs 0.97s, respectively ± 0.64s, P = 0.207). Contrast administration also led to a considerable increase in data heterogeneity in nTZ, PZ and TZ lesions, in which CV reached 54.5%. Data conclusion: Gadolinium significantly shortens MRF T1 in all tissues, increasing its dispersion and compromising its ability to identify TZ lesions. Therefore, MRF should be run prior to DCE imaging as part of a prostate multiparametric protocol to achieve optimal results. Keywords: magnetic resonance fingerprinting, MRI, prostate cancer, gadolinium, T1 relaxometry, multiparametric protocol Introduction Prostate cancer (PCa) is the second commonest male cancer and the fifth leading cause of cancer mortality worldwide. (1) Several prospective trials, including PROMIS and PRECISION, have triggered changes in major European and American guidelines, which now recommend multiparametric (mp) MRI as the first-line investigation for patients with suspected early stage PCa. (2-5) Despite having advantages over the traditional systematic biopsy pathway, mpMRI has limitations including a known learning curve and considerable interobserver variation for image assessment. (6-8) The current Prostate Imaging Reporting and Data System (PI-RADS) guidelines only incorporate qualitative measures for interpretation and quantitative metrics have been suggested as a means of reducing variability. (9,10) Magnetic resonance fingerprinting (MRF) is a quantitative technique able to generate multiple property maps (e.g., T1, T2, apparent proton-density). These maps are inherently spatially registered and insensitive to motion, and can be acquired within imaging times comparable to or faster than conventional mapping techniques. (11,12) Instead of varying a single acquisition parameter at a time, such as imaging with a single flip-angle variation per full k-space acquisition, MRF simultaneously varies multiple sequence parameters using a pseudorandom manner to generate fingerprint-like signal evolutions for combinations of desired tissue properties. These signal evolutions are then matched with a collection of simulated fingerprints, based on a range of T1 and T2 values, thus generating multiple property maps within a single acquisition. (13,14) In PCa, MRF has shown promise for identifying both peripheral and transition zone lesions while also differentiating between indolent and clinically significant disease. (15-17) Prospective validation of MRF results in the clinical setting likely requires its addition to the standard clinical mpMRI sequences, which include dynamic contrast enhanced (DCE) imaging as a mandatory component. (7) Understanding the impact of gadolinium on the diagnostic utility of MRF-based T1 relaxometry due to preferential T1 shortening effect of gadolinium at low doses, is critical for making an informed decision on the running order of MRF in relation to DCE as part of a prostate multiparametric protocol.(18) MRF may also provide a method for fast, accurate T1 mapping in the presence of gadolinium-based relaxation time shortening to provide. If MRF proves more reliable than other T1 mapping methods, MRF may present an opportunity to reduce gadolinium-based contrast agent doses. Therefore, in the present study we sought to evaluate the effect of gadolinium-based contrast agent administration on MRF T1 relaxation time in patients with biopsy-proven peripheral and transition zone prostate lesions. Materials and Methods A local Institutional Review Board and ethics committee granted approval for this prospective study (ethics reference: anonymised), with written informed consent obtained from all participants. Patients with MR-visible, biopsy-proven PCa were included in this study. Exclusion criteria included prostate biopsy within the preceding 3 months, presence of pelvic metalwork, or any previous treatment for PCa. Comment by Пользователь: @Vincent – could you please provide the TAPS study ethics reference number? Multiparametric MRI Patients underwent prostate MRI on a 3T HDx Discovery MR750 scanner (GE Healthcare, Waukesha, WI) using a 32 channel phased array coil. Intravenous injection of hyoscine butylbromide (Buscopan, 20 mg/mL; Boehringer, Germany) was administered prior to imaging to reduce peristaltic movement, unless clinically contraindicated. Multiparametric MRI protocol included Axial T1 and multiplanar high-resolution T2-weighted 2D fast recovery FSE (field of view (FOV) 18x18cm; slice thickness 3mm; gap 0mm). Diffusion-weighted imaging (DWI) was performed using a spin-echo echo-planar imaging pulse sequence (slice thickness 3mm; gap 0mm; b-values: b-150, b-750, and b-1,400 s/mm2) and an additional small FOV (24 cm) b-2,000s/mm2 DWI sequence; apparent diffusion coefficient (ADC) maps were calculated automatically. Dynamic contrast enhancement (DCE) was performed using a standard sequence (FOV 24cm; slice thickness and gap 3mm and 0mm, respectively; temporal resolution 7 seconds) following a bolus of Gadobutrol (Gadovist, 0.1 mmol/kg, Bayer) at 28 seconds. Add T1 and T2 mapping sequences. Comment by Пользователь: @Josh could you please insert these as discussed – maybe as a separate section? Full acquisition time would be important as we are stressing in the Discussion that MRF is indeed not only more reliable but also faster than conventional mapping MR Fingerprinting protocol A 2D steady-state-free-precession (SSFP) MRF sequence with inversion preparation was used for obtaining slice. (11,19) 979 under-sampled, interleaved spirals were used for k-space sampling. The maximum gradient strength per spiral was 28 mT/m and the maximum slew rate was 108 T/m/s. The imaging parameters were: FOV = 260x260 mm2, matrix = 256x256, slices = 15-22, slice thickness = 3.0 mm, spacing 1.0 mm, sampling bandwidth = ±250 kHz, slice dephasing = 8π, echo time (TE) = 2.5 ms, repetition time (TR) = 10ms, acquisition time = 9.79 seconds/slice. We used similar flip angle lists to those used in previous works in order to more accurately assess the utility of MRF with a standard lists.(20,21) Axial images were acquired to match the standard acquisition planes for other prostate image assessments. Image reconstruction Comment by Пользователь: @Josh from here down the MRF methodology becomes quite detailed… should this be made supplemental data? Each under-sampled spiral was reconstructed, re-gridded and reconstructed to image space with 48 parallel CPUs. After reconstruction, each coil channel was combined using adaptive coil combination based on weights determined from the average of the time frames.(22) The under-sampled images were reduced from 979 to 16 images using the singular value decomposition (SVD) decomposition weights determined during dictionary compression.(23) Comment by Пользователь: @Josh could you please expand acronym? MRF dictionary simulation Dictionary simulations were performed using the extended phase graph formalism, and included the slice profile.(24) The ranges and incremental (step-size) changes of the T1 and T2 values that were simulated in the dictionary were T1 = [0.01:0.005:1; 1:0.04:6] seconds ([minimum: step-size: maximum]), and for T2 = [0.005:0.001:0.1; 0.1:0.01:4; 4:0.04:6] seconds (where the semi-colons indicate concatenated lists). The dictionary size was compressed to 16 singular vectors (rank) with SVD to reduce the size for long term storage and faster dictionary matching.(23) MRF pattern matching T1 and T2 values are were assigned to each voxel after pattern matching. MRF used inner product pattern matching of the signals with the simulated dictionary to obtain the best T1 and T2 match with the acquired reconstructed data. The inner products between the normalized measured signal evolution of each voxel and each normalised dictionary entry are calculated. The dictionary entry returning the maximum value for the inner product is taken as the best representation of the acquired signal evolution. Image analysis MRF T1 and T2 values for lesions, normal peripheral zone (nPZ), normal transition zone (nTZ), subcutaneous abdominal fat and normal internal obturator muscle were calculated from ROIs drawn by a single fellowship-trained uro-radiologist using the open-source segmentation software ITK-SNAP.(25) ROIs were then transposed into MRI-based ADC, T1 and T2 maps with their size and location being matched to the appropriate FOV parameters and anatomical position of the outlined structures using in-house software developed within Python using the PyQtGraph and PyDicom libraries.(26) Statistics The Shapiro-Wilk test was applied to assess the distribution of imaging values with their intragroup heterogeneity evaluated using coefficient of variation (CV) and intergroup comparison performed using either paired t-test as appropriate. Agreement between MRI- and MRF-based T1 and T2 relaxation times, treated as single paired measurements, was tested using the Bland-Altman analysis with difference plots created in Prism 8 (GraphPad Software, San Diego, CA). Results The study included 14 patients with biopsy-proven PCa with mean age 70 years (IQR, 67.3-73.5 years), mean PSA 6.29 ng/mL (IQR, 3.8-8.7 ng/mL), with mean time since last biopsy being 16 months (range 4-48 months). A total of 19 MR-visible prostate lesions were included in the analysis, 10 of which were located in the peripheral zone (PZ) and 9 in the transition zone (TZ). Three lesions exhibited intermediate-grade Gleason score 3+4=7 disease (grade group 2) while other lesions harboured low-grade disease with Gleason score of 3+3=6 (grade group 1). Pre-gadolinium MRF and MRI-based mapping agreement and heterogeneity Summary pre-gadolinium MRF-based T1 and T2 along with MRI-based T1, T2 and ADC values obtained from all prostate lesions combined (n = 19), PZ (n = 10) and TZ (n = 9) lesions, pooled nPZ and nTZ, internal obturator muscle and subcutaneous fat (n = 14 for all) are presented in Table 1. Bland-Altman plots showed acceptable agreement between MRF and MRI-based T1 relaxation times, with 95% limits of agreement ranging between -1.99 to 1.17 (bias, SD -0.41 ± 0.81) (Figure 1) Coefficients of variation (CVs) were calculated for all acquired values to compare their dispersion in different tissues and are presented in Table 2. MRF T1 demonstrated low heterogeneity in all tissues except for nTZ, where it reached 25.4%. (Figure 2a) MRI-based T1 relaxation times showed proportionately higher heterogeneity compared to MRF T1. The relationship for T2 mapping was inverse, with MRF-derived T2 values demonstrated considerably higher dispersion compared to both MRI-based T2 relaxation times and MRF-derived T1 values. (Table 2) Post-gadolinium MRF-based T1 relaxometry Post-gadolinium MRF T1 and T2 obtained from prostate lesions, pooled nPZ, nTZ, internal obturator muscle and subcutaneous fat are presented in Table 1. Paired t-test showed a significant MRF T1 shortening effect of gadolinium in all tissues (Figure 3). MRF T1 heterogeneity increased from low to high in all tissue types except nPZ, muscle and fat, where CVs remained in the same category as pre-gadolinium. A particularly marked, almost five-fold increase in data heterogeneity was observed for MRF T1 values obtained from TZ lesions and, to a lesser extent nTZ, whereas PZ lesions demonstrated only a two-fold increase in CV and only a marginal change in CV was noted in nPZ. (Figure 2b and Table 2) MRF-based T1 relaxometry for differentiating tumour and normal tissue Prior to gadolinium-based contrast agent administration, a paired t-test revealed significantly shorter MRF T1 values for both peripheral and transition zone lesions when compared to corresponding nPZ and nTZ in the same patients (1.64s ± 0.37s vs 2.20s ± 0.78s for PZ and 1.70s ± 0.20s vs 1.97s ± 0.32s for TZ; P = 0.03 and 0.013, respectively). (Figure 4a) In pooled nPZ, MRF T1 relaxation time was significantly longer than in nTZ (2.52 ± 0.41 vs 1.76 ± 0.45; P < 0.0001). Post-contrast MRF T1 remained significantly shorter within peripheral zone lesions compared to the normal PZ (0.68 ± 0.29 vs 1.32 ± 0.22; P < 0.0001), however, it there was no longer a significant difference in TZ tumours compared to nTZ (0.76 ± 0.42 vs 0.97 ± 0.64, P = 0.207). (Figures 4b and 5) Pooled nTZ T1 relaxation time was again significantly shorter than those of nPZ (1.27 ± 0.23 vs 0.73 ± 0.41; P < 0.0001). MRF-based T2 relaxometry and MRI-based T1, T2 and ADC mapping for differentiating tumour and normal tissue Pre-contrast MRF T2 relaxation times were shorter in tumours compared to corresponding nPZ and nTZ (0.51 ± 0.29 vs 0.53 ± 0.26 for PZ and 0.37 ± 0.21 vs 0.53 ± 0.25 for TZ; P = 0.811 and 0.165, respectively), however, the results lacked statistical significance. No difference was also observed between MRF-based T2 of pooled nPZ and nTZ (0.55 ± 0.29 vs 0.45 ± 0.23; P = 0.413). Post-contrast MRF T2 values were slightly shorter in lesions compared to corresponding normal zones, but this difference was not significant (0.27 ± 0.16 vs 0.33 ± 0.27 for PZ and 0.23 ± 0.18 vs 0.11 ± 0.04 for TZ; P = 0.799 and 0.071, respectively). There was also no difference between T2 relaxation times of pooled nPZ and nTZ (0.33 ± 0.26 vs 0.24 ± 0.27, P = 0.419). MRI-based T1 and T2 relaxation times did not differ between PZ lesions and corresponding nPZ (1.99 ± 0.63 vs 2.17 ± 0.78 for T1 and 0.09 ± 0.02 vs 0.14 ± 0.09; P = 0.145 and 0.099, respectively), however, in the TZ, both T1 and T2 were significantly shorter in tumour compared to normal tissue (2.00 ± 0.41 vs 2.34 ± 0.53 for T1 and 0.07 ± 0.01 vs 0.09 ± 0.01 for T2; P = 0.02 and 0.003, respectively). T1 showed no difference between pooled nPZ and nTZ (2.19 ± 0.81 vs 2.12 ± 0.73; P = 0.475), while T2 values of nPZ were significantly longer than those of nTZ (0.14 ± 0.08 vs 0.09 ± 0.01; P = 0.045). ADC values were significantly lower in both PZ and TZ lesions compared to corresponding nPZ and nTZ (0.93 ± 0.16mm2/s vs 1.61 ± 0.22mm2/s for PZ and 0.90 ± 0.14mm2/s vs 1.35 ± 0.15 mm2/s for TZ; P = 0.0001 for both). Pooled nTZ had lower ADC values than nPZ (1.35 ± 0.15 vs 1.61 ± 0.22; P = 0.0004). Discussion This prospective, proof-of-concept study demonstrates the effect of gadolinium-based contrast agent administration on MRF-based T1 relaxometry in the clinical setting. We have shown that gadolinium significantly shortens MRF T1, considerably increases its heterogeneity in both normal and malignant prostate tissue, and compromises its diagnostic utility in the transition zone. These results will help inform future studies, when MRF may be incorporated into standard-of-care prostate mpMRI protocols. The observed MRF T1 shortening effect post-contrast is expected, as gadolinium facilitates both longitudinal and transverse magnetic relaxation, thereby shortening both T1 and T2 of tissues. (27,28) Anderson et al. also observed a similar trend when measuring gadolinium and dysprosium concentrations in mouse glioma model using dual contrast-MRF. (29) We also show that pre-gadolinium MRF-derived T1 relaxation times and ADC values were significantly lower in cancers compared to normal tissue in both the TZ and PZ of the prostate. These findings align well with previous studies where a combination of MRF-derived T1 and standard ADC worked best for identifying PZ and TZ lesions with our absolute values being similar to those reported in this study.(15-17) Low dispersion of ADC values is, however, less representative of real-life clinical practice as diffusion-related parameters are highly sensitive to motion, which may hinder assessment of the peripheral zone, where 75-80% of clinically significant lesions are located. (30-34) Acceptable pre-contrast heterogeneity of MRF T1 coupled with the technique’s intrinsic insensitivity to motion further support the need for investigating the added value of MRF in prostate imaging, particularly when DWI fails due to motion or susceptibility artefact. (Table 2) In addition to shortening MRF T1 values, gadolinium also led to a considerable increase in T1 heterogeneity, the degree of which varied among different tissues. Normal TZ exhibited the highest heterogeneity pre-contrast, which is not unexpected in this age group, given the high prevalence of benign prostatic hyperplasia (BPH). Marked hypervascularity of BPH nodules may in turn explain the two-fold increase in MRF T1 dispersion in nTZ following gadolinium-based contrast agent administration.(35) This trend was even more prominent in TZ lesions, in which the increase in MRF T1 heterogeneity was almost five-fold, which may explain the inability of MRF T1 to identify TZ lesions post-contrast. In fact, DCE-MRI is known to be of limited use for assessment of TZ lesions, which have been shown to have similar quantitative DCE-MRI metrics as normal TZ. (36-38) Conversely, nPZ, fat and muscle, which can be considered more as “control” type tissues given their relatively low vascularity and morphological homogeneity, maintained low MRF T1 heterogeneity post-contrast. To our knowledge, this is the first study reporting both MRF and conventional T1 and T2 values obtained from patients with prostate cancer. As expected, MRF had considerably shorter scanning time of ?min compared to conventional T1 and T2 mapping at ?min (?min and ? min, respectively) The ability of MRF to obtain a broader range of flip angles compared to conventional variable flip angle T1 mapping resulted in reduced heterogeneity for all tissue types, which is of particular importance for prostate imaging, where long T1 (1-2s) values are typical. High dispersion of conventional T1 mapping values in PZ lesions, which was observed in this study, may explain the technique’s limited diagnostic utility in the PZ; a similar trend was noted for conventional T2 mapping, which is more frequently used in prostate cancer.(39) Comment by Пользователь: @Josh could you please specify these? This study has several limitations. Firstly, the small sample size may have artificially increased data dispersion leading to the inability of MRI-derived T2 to identify PZ tumours, however, lesion differentiation was not the primary aim of the study. Secondly, only patients with low- and intermediate-grade disease were included in this study, which may also have had an impact on the diagnostic utility of both MRF and conventional mapping techniques. Conventional T1 and T2 maps were not acquired post-contrast, however, this became less relevant given the marked effect of gadolinium on MRF-based T1 relaxometry. MRF-derived T2 values were unreliable in this study, however, as gadolinium is known to have a more prominent T1 shortening effect at low doses, and T2 mapping, whether directly acquired or derived by MRF, can be susceptible to magnetic field non-uniformy, motion, and low signal-to-noise. In future work investigating T2 with MRF, we would give strong consideration for methods that increase the signal-to-noise ratio by increasing the slice thickness, voxel sizes, or by performing averaging with additional acquisitions. While averaging for MRF T2 mapping would remain sensitive to motion, we hypothesize that this effect would be reduced due to the pattern matching algorithm of MRF. In conclusion, MRF T1 should be run prior to DCE imaging as part of a prostate mpMRI protocol in order to ensure its diagnostic utility in both PZ and TZ lesions, which is compromised by a significant MRF T1 shortening effect of gadolinium. References: 1. Rawla P. Epidemiology of Prostate Cancer. World J Oncol 2019;10(2):63-89. 2. Kasivisvanathan V, Emberton M, Moore CM. MRI-Targeted Biopsy for Prostate-Cancer Diagnosis. N Engl J Med 2018;379(6):589-590. 3. Ahmed HU, El-Shater Bosaily A, Brown LC, et al. Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet 2017;389(10071):815-822. 4. Urology EAo. EAU Guidelines: Prostate Cancer | Uroweb. 2019. 5. American Urological Association SoAR. Standard Operating Procedure for Multiparametric Magnetic Resonance Imaging in the Diagnosis, Staging and Management of Prostate Cancer - American Urological Association. 2019. 6. Gaziev G, Wadhwa K, Barrett T, et al. Defining the learning curve for multiparametric magnetic resonance imaging (MRI) of the prostate using MRI-transrectal ultrasonography (TRUS) fusion-guided transperineal prostate biopsies as a validation tool. BJU Int 2016;117(1):80-86. 7. Turkbey B, Rosenkrantz AB, Haider MA, et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol 2019;76(3):340-351. 8. Greer MD, Shih JH, Lay N, et al. Interreader Variability of Prostate Imaging Reporting and Data System Version 2 in Detecting and Assessing Prostate Cancer Lesions at Prostate MRI. AJR Am J Roentgenol 2019:1-8. 9. Liu W, Turkbey B, Senegas J, et al. Accelerated T2 mapping for characterization of prostate cancer. Magn Reson Med 2011;65(5):1400-1406. 10. Metzger GJ, Kalavagunta C, Spilseth B, et al. Detection of Prostate Cancer: Quantitative Multiparametric MR Imaging Models Developed Using Registered Correlative Histopathology. Radiology 2016;279(3):805-816. 11. Ma D, Gulani V, Seiberlich N, et al. Magnetic Resonance Fingerprinting. Nature 2013;495(7440):187-192. 12. Poorman ME, Martin MN, Ma D, et al. Magnetic resonance fingerprinting Part 1: Potential uses, current challenges, and recommendations. J Magn Reson Imaging 2019. 13. Bipin Mehta B, Coppo S, Frances McGivney D, et al. Magnetic resonance fingerprinting: a technical review. Magn Reson Med 2019;81(1):25-46. 14. McGivney DF, Boyacioglu R, Jiang Y, et al. Magnetic resonance fingerprinting review part 2: Technique and directions. J Magn Reson Imaging 2019. 15. Yu AC, Badve C, Ponsky LE, et al. Development of a Combined MR Fingerprinting and Diffusion Examination for Prostate Cancer. Radiology 2017;283(3):729-738. 16. Panda A, O'Connor G, Lo WC, et al. Targeted Biopsy Validation of Peripheral Zone Prostate Cancer Characterization With Magnetic Resonance Fingerprinting and Diffusion Mapping. Invest Radiol 2019;54(8):485-493. 17. Panda A, Obmann VC, Lo WC, et al. MR Fingerprinting and ADC Mapping for Characterization of Lesions in the Transition Zone of the Prostate Gland. Radiology 2019;292(3):685-694. 18. Xiao YD, Paudel R, Liu J, Ma C, Zhang ZS, Zhou SK. MRI contrast agents: Classification and application (Review). Int J Mol Med 2016;38(5):1319-1326. 19. Kaggie J, Deen S, Kessler D, et al. Feasibility of Quantitative Magnetic Resonance Fingerprinting in Ovarian Tumors for T1 and T2 Mapping in a PET/MR Setting. IEEE Transactions on Radiation and Plasma Medical Sciences 2019;3(4):509-515. 20. Jiang Y, Ma D, Seiberlich N, Gulani V, Griswold MA. MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magn Reson Med 2015;74(6):1621-1631. 21. Buonincontri G, Biagi L, Retico A, et al. Multi-site repeatability and reproducibility of MR fingerprinting of the healthy brain at 1.5 and 3.0T. Neuroimage 2019;195:362-372. 22. Walsh DO, Gmitro AF, Marcellin MW. Adaptive reconstruction of phased array MR imagery. Magn Reson Med 2000;43(5):682-690. 23. McGivney DF, Pierre E, Ma D, et al. SVD compression for magnetic resonance fingerprinting in the time domain. IEEE Trans Med Imaging 2014;33(12):2311-2322. 24. Hennig J, Weigel M, Scheffler K. Calculation of flip angles for echo trains with predefined amplitudes with the extended phase graph (EPG)-algorithm: principles and applications to hyperecho and TRAPS sequences. Magn Reson Med 2004;51(1):68-80. 25. Yushkevich PA, Gerig G. ITK-SNAP: An Intractive Medical Image Segmentation Tool to Meet the Need for Expert-Guided Segmentation of Complex Medical Images. IEEE Pulse 2017;8(4):54-57. 26. Mason D. SU‐E‐T‐33: Pydicom: An Open Source DICOM Library. Volume 38. Med. Phys.; 2011. p. 3493-3493. 27. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB. Gadolinium(III) Chelates as MRI Contrast Agents: Structure, Dynamics, and Applications. Chem Rev 1999;99(9):2293-2352. 28. De Leon-Rodriguez LM, Martins AF, Pinho MC, Rofsky NM, Sherry AD. Basic MR relaxation mechanisms and contrast agent design. J Magn Reson Imaging 2015;42(3):545-565. 29. Anderson CE, Johansen M, Erokwu BO, et al. Dynamic, Simultaneous Concentration Mapping of Multiple MRI Contrast Agents with Dual Contrast - Magnetic Resonance Fingerprinting. Sci Rep 2019;9(1):19888. 30. McNeal JE, Redwine EA, Freiha FS, Stamey TA. Zonal distribution of prostatic adenocarcinoma. Correlation with histologic pattern and direction of spread. Am J Surg Pathol 1988;12(12):897-906. 31. McNeal JE. Cancer volume and site of origin of adenocarcinoma in the prostate: relationship to local and distant spread. Hum Pathol 1992;23(3):258-266. 32. Mazaheri Y, Vargas HA, Nyman G, Akin O, Hricak H. Image artifacts on prostate diffusion-weighted magnetic resonance imaging: trade-offs at 1.5 Tesla and 3.0 Tesla. Acad Radiol 2013;20(8):1041-1047. 33. Caglic I, Hansen NL, Slough RA, Patterson AJ, Barrett T. Evaluating the effect of rectal distension on prostate multiparametric MRI image quality. Eur J Radiol 2017;90:174-180. 34. Slough RA, Caglic I, Hansen NL, Patterson AJ, Barrett T. Effect of hyoscine butylbromide on prostate multiparametric MRI anatomical and functional image quality. Clin Radiol 2018;73(2):216.e219-216.e214. 35. Rosenkrantz AB, Kim S, Campbell N, Gaing B, Deng FM, Taneja SS. Transition zone prostate cancer: revisiting the role of multiparametric MRI at 3 T. AJR Am J Roentgenol 2015;204(3):W266-272. 36. Padhani AR, Gapinski CJ, Macvicar DA, et al. Dynamic contrast enhanced MRI of prostate cancer: correlation with morphology and tumour stage, histological grade and PSA. Clin Radiol 2000;55(2):99-109. 37. Hansford BG, Peng Y, Jiang Y, et al. Dynamic Contrast-enhanced MR Imaging Curve-type Analysis: Is It Helpful in the Differentiation of Prostate Cancer from Healthy Peripheral Zone? Radiology 2015;275(2):448-457. 38. Greer MD, Shih JH, Lay N, et al. Validation of the Dominant Sequence Paradigm and Role of Dynamic Contrast-enhanced Imaging in PI-RADS Version 2. Radiology 2017;285(3):859-869. 39. Hoang Dinh A, Souchon R, Melodelima C, et al. Characterization of prostate cancer using T2 mapping at 3T: a multi-scanner study. Diagn Interv Imaging 2015;96(4):365-372.