Attention: Restrictions on use of AUA, AUAER, and UCF content in third party applications, including artificial intelligence technologies, such as large language models and generative AI.
You are prohibited from using or uploading content you accessed through this website into external applications, bots, software, or websites, including those using artificial intelligence technologies and infrastructure, including deep learning, machine learning and large language models and generative AI.

ARTIFICIAL INTELLIGENCE Magnetic Resonance Fingerprinting Technology for Noninvasive Quantification of Prostate Cancer

By: Eduardo Thadeu de Oliveira Correia, MD, PhD, University Hospitals Cleveland Medical Center, Ohio; Yong Chen, PhD, University Hospitals Cleveland Medical Center, Ohio, Case Western Reserve University, Cleveland, Ohio; Sree Harsha Tirumani, MD, University Hospitals Cleveland Medical Center, Ohio, Case Western Reserve University, Cleveland, Ohio; Dan Ma, PhD, Case Western Reserve University, Cleveland, Ohio, University Hospitals Cleveland Medical Center, Ohio; Mark A. Griswold, PhD, University Hospitals Cleveland Medical Center, Ohio, Case Western Reserve University, Cleveland, Ohio; Leonardo Kayat Bittencourt, MD, PhD, University Hospitals Cleveland Medical Center, Ohio, Case Western Reserve University, Cleveland, Ohio | Posted on: 05 Jan 2024

Introduction

Over the last decade, there has been a growing interest in the investigation of quantitative MRI techniques in the field of prostate imaging. The primary goal of quantitative imaging is to improve both intra- and inter-reader reproducibility by using objective tissue property values to diagnose and stage suspicious prostate cancer (PCa) lesions. Nonetheless, early conventional mapping techniques were often time-consuming and reproducibility across different MRI scanners is often a concern. Our team at Case Western Reserve University, in collaboration with Siemens Healthineers, developed a native quantitative MRI technique known as magnetic resonance fingerprinting (MRF).1 MRF allows for the simultaneous acquisition of T1 and T2 maps,1 as illustrated in the Figure, and more recently, diffusion maps,2 all within a clinically feasible time. Importantly, the multiple tissue properties maps acquired by MRF are incoherently coregistered, enabling direct multiparametric tissue analysis.

image

Figure. Magnetic resonance fingerprinting (MRF) mapping of the prostate in patients with and without a focal suspicious lesion. The top row illustrates a case of a Prostate Imaging Reporting and Data System (PI-RADS) 1 prostate MRI, with no focal suspicious lesions in the peripheral zone. The bottom row illustrates a case of a PI-RADS 4 MRI, with a focal suspicious lesion in the right peripheral zone (indicated by a white arrow). As demonstrated in previous studies,5,13 MRF-derived T1 and T2 values of the suspicious lesions are significantly lower than T1 and T2 values of the normal-appearing peripheral zone (contoured by a dotted line). ADC indicates apparent diffusion coefficient; T2WI, T2-weighted imaging.

To this date, within the field of abdominal radiology, particularly in genitourinary imaging, the primary focus of MRF has been to improve the noninvasive detection and characterization of PCa in both the transition zone (TZ) and peripheral zone (PZ).3-5 This paper provides an overview of prior developments in prostate MRF while highlighting emerging applications that hold potential for reshaping PCa management.

Current Applications of Prostate MRF

One pioneering application of prostate MRF was introduced by Yu et al, who proposed a novel prostate MRI examination that combines standard apparent diffusion coefficient (ADC) maps with MRF-derived T1 and T2 maps.3 In this paper, Yu et al demonstrated that T1 and T2 relaxation times of areas of known PCa were markedly lower than those of the normal-appearing PZ. Furthermore, integrating ADC, T1, and T2 maps yielded an area under the curve (AUC) of 0.83 for distinguishing low-grade from intermediate- and high-grade PCa lesions.3 Similar findings with the same methodology were shown by Panda et al, albeit focusing on the TZ.4 In this investigation, Panda et al showed that the combination of MRF-derived T1 maps and standard ADC mapping could allow for the differentiation of PCa lesions from the normal TZ, with an AUC of 0.94.4 Subsequently, Shiradkar et al assessed the likely biological basis behind variations in T1 and T2 relaxation times using histopathology specimens from radical prostatectomy.6 Their study showed that parameters of tissue composition ratio (percentage of epithelium, stroma, and lumen) differed between areas of normal prostatic tissue, prostatitis, and PCa.6 Remarkably, T1 and T2 relaxation times were correlated with parameters of tissue composition ratio.6 Beyond localized PCa, Choi et al also demonstrated that prostate MRF–derived T1 and T2 maps could differentiate PCa bone metastasis from normal bone, expanding the horizon of potential prostate MRF applications.7

Following these groundbreaking applications, several investigations were published looking at more technical aspects of prostate MRF. For instance, Sushentsev et al explored the feasibility of performing prostate MRF examinations at 1.5 T (tesla).8 Though promising, multicenter studies with higher sample sizes are still required to validate the acquisition of prostate MRF maps using lower field scanners. Another study by Sushentsev et al raised questions about the impact of contrast administration on T1 relaxation times, which could potentially affect the ability of T1 maps to distinguish TZ lesions from normal TZ.9 Nevertheless, a subsequent study from Lee et al, involving a larger cohort, showed that T1 and T2 relaxation times remained significantly lower for PCa compared with normal areas of the PZ and TZ after contrast administration.10 This finding may pave the way for the development of truly quantitative delayed contrast-enhanced sequences for prostate imaging using MRF. Han et al also contributed to the advancement of prostate MRF by showing the feasibility of a 3D acquisition of prostate MRF maps.11

In preparation for larger-scale implementation of prostate MRF examinations, Lo et al conducted a study across 5 different scanners from 3 institutions in the United States and Brazil. Their findings demonstrated minimal intrascanner and interscanner variations in MRF-derived T1 and T2 relaxation times.12 This underscores the repeatability and reproducibility of prostate MRF acquisitions across various scanners and centers. More recently, Correia et al provided the first reference values for T1 and T2 relaxation times of the whole normal-appearing PZ across patients in different Prostate Imaging Reporting and Data System (PI-RADS) categories.13 A comprehensive overview of key studies on prostate MRF is available in the Table.

Table. Key Studies on Prostate Magnetic Resonance Fingerprinting

Study, y Study design Patient/volunteer population N Aim Main conclusions
Correia et al 202313 Retrospective, single center Patients with suspected PCa that had PI-RADS 1-5 MRIs and MRF maps available 124 Patients Comprehensively assess the distribution of MRF-derived T1 and T2 relaxation times of the whole normal-appearing PZ The mean T1 relaxation time was 1941 ms, while the mean T2 relaxation time was 88 ms
Lo et al 202212 Prospective, multicenter (5 different scanners across 3 institutions) Patients with suspected PCa that underwent a prostate MRF 24 Patients Investigate the multicenter reproducibility and repeatability of T1 and T2 relaxation times Intrascanner variation was about 2% for T1 and 4.7% for T2. Interscanner variation between institutions was about 4.9% for T1 and 8.1% for T2
Lee et al 202210 Retrospective, single center Patients without a previous history of PCa that underwent a TRUS biopsy and had a prebiopsy contrast-enhanced prostate MRI with MRF acquisitions 57 Patients Assess MRF-derived T1 and T2 relaxation times of noncontrast-enhanced and contrast-enhanced MRF for both the normal PZ and TZ and also for PCa Median nonenhanced and contrast-enhanced T1 values were 1906 and 880 ms for the normal PZ, 1624 and 542 ms for the normal TZ, and 1510 and 605 ms for PCa. Median nonenhanced and contrast-enhanced T2 values were 180 and 186 ms for the normal PZ, 101 and 91 ms for the normal TZ, and 81 and 73 ms for PCa
Choi et al 20217 Retrospective, single center Patients with suspected PCa that had pelvic bone metastasis on MRI 30 Patients Assess the feasibility of using MRF to evaluate PCa bone metastasis ROIs of bone metastasis had significantly higher nonenhanced and contrast-enhanced T1 relaxation times and significantly lower nonenhanced and contrast-enhanced T2 relaxation times compared with benign bone
Han et al 202111 Retrospective, single center Patients with suspected PCa that underwent prostate MRF 90 Patients Assess the feasibility of a 3D prostate MRF acquisition T1 and T2 relaxation times obtained from a 3D prostate MRF acquisition had an excellent correlation with relaxation times obtained in the phantom study
Sushentsev et al 20218 Prospective, single center Volunteers without previous diagnosis or treatment for PCa 10 Healthy volunteers Assess the reproducibility of MRF-derived T1 relaxation times from phantoms and healthy volunteers at both 1.5 T and 3 T field strengths Mean T1 relaxation time was significantly higher at 3 T than at 1.5 T. There was a strong linear correlation between 1.5 T and 3 T T1 relaxation times. Interscanner agreement was acceptable for both 1.5 T and 3 T T1 mapping
Shiradkar et al 20216 Retrospective, single center Patients with suspected PCa that underwent a prostate MRF and were subsequently submitted to radical prostatectomy 14 Patients Investigate the histopathological basis that justifies prostate MRF measurements for characterizing prostatitis and PCa Areas of normal PZ, prostatitis, and PCa had different tissue composition ratios. There were significant correlations between different parameters of tissue composition ratio and T1 and T2 relaxation times
Sushentsev et al 20209 Prospective, single center Patients with MRI-visible biopsy-proven PCa on active surveillance 14 Patients Evaluate the variation of T1 relaxation time after contrast administration Mean T1 relaxation times before and after contrast administration were 2521 and 1270 ms for the normal PZ, 1753 and 724 ms for the normal TZ, and 1666 and 718 ms for PCa lesions. Contrast administration impaired the ability of T1 to differentiate TZ lesions from areas of normal TZ
Panda et al 20194 Retrospective, single center Patients with suspected PCa submitted to targeted biopsy that had a prebiopsy MRF 67 Patients Investigate the role of MRF combined with clinical ADC mapping to characterize TZ lesions The combination of MRF-derived T1 and standard ADC maps could differentiate TZ lesions from the normal TZ with an AUC of 0.94
Panda et al 20195 Retrospective, single center Patients with suspected PCa that were submitted to a targeted biopsy 85 Patients Investigate if MRF-based T1 and T2 relaxation times in addition to conventional ADC mapping are able to distinguish PZ lesions from areas of benign prostatic tissue MRF-derived T1 and conventional ADC maps could differentiate between PCa and negative biopsies with an AUC of 0.83
Yu et al 20173 Retrospective, single center Patients with suspected PCa that underwent systematic or targeted biopsies and had a prebiopsy MRF 131 Patients Assess the role of MRF-derived T1 and T2 relaxation times combined with clinically available ADC maps to characterize prostatic diseases Standard ADC combined with MRF-derived T2 maps had the highest AUC (0.83) to differentiate high- or intermediate-grade tumors from low-grade PCa. Mean T1, T2, and ADC values of PCa were significantly lower than those from the normal PZ
Abbreviations: ADC, apparent diffusion coefficient; AUC, area under the curve; MRF, magnetic resonance fingerprinting; MRI, magnetic resonance imaging; N, number; PCa, prostate cancer; PI-RADS, Prostate Imaging Reporting and Data System; PZ, peripheral zone; ROI, region of interest; T, tesla; TRUS, transrectal ultrasound; TZ, transition zone.

Emerging Applications

Our institution has developed an MRF technique for the simultaneous acquisition of relaxometry and diffusion maps of the brain.2 We are currently exploring the adaptation of this sequence for use in prostate imaging to generate repeatable, reproducible, and motion-robust ADC maps in addition to the already validated T1 and T2 maps. Our group is also examining the use of rigid coregistration between prostatectomy specimens and optimized prostate MRF acquisitions to improve the correlation between whole-mount histopathology and MRF maps, striving to get nearer to the concept of virtual pathology. Another promising application under investigation by our group involves the integration of prostate MRF to optimize existing biopsy decision-making workflows. Moreover, MRF holds the potential to improve the characterization of PI-RADS 3 lesions, which represent uncertainty and equivocal findings in which the current standard multiparametric MRI does not add value to biopsy decisions. MRF-derived T1 and T2 assessment can provide quantitative information to aid biopsy decision-making and potentially improve the management of the PI-RADS 3 subgroup. Additionally, leveraging radiomics to assess multiparametric voxel-wise quantitative data obtained with prostate MRF may unveil new promising applications in clinical practice.

Conclusions

Prostate MRF stands as a valuable quantitative MRI technique with proven clinical applications for noninvasive PCa detection and characterization. However, robust evidence supporting its widespread adoption in clinical practice is still needed. This gap may be attributed to the absence of large-scale, multi-institutional clinical studies supporting the benefit of integrating prostate MRF into clinical workflows. Future research should focus on the optimization and prospective large-scale validation of prostate MRF to facilitate its broader implementation in clinical practice.

  1. Ma D, Gulani V, Seiberlich N, et al. Magnetic resonance fingerprinting. Nature. 2013;495(7440):187-192.
  2. Afzali M, Mueller L, Sakaie K, et al. MR fingerprinting with b-tensor encoding for simultaneous quantification of relaxation and diffusion in a single scan. Magn Reson Med. 2022;88(5):2043-2057.
  3. 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.
  4. 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.
  5. 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.
  6. Shiradkar R, Panda A, Leo P, et al. T1 and T2 MR fingerprinting measurements of prostate cancer and prostatitis correlate with deep learning-derived estimates of epithelium, lumen, and stromal composition on corresponding whole mount histopathology. Eur Radiol. 2021;31(3):1336-1346.
  7. Choi MH, Lee S-W, Kim HG, et al. 3D MR fingerprinting (MRF) for simultaneous T1 and T2 quantification of the bone metastasis: initial validation in prostate cancer patients. Eur J Radiol. 2021;144:109990.
  8. Sushentsev N, Kaggie JD, Slough RA, Carmo B, Barrett T. Reproducibility of magnetic resonance fingerprinting-based T1 mapping of the healthy prostate at 1.5 and 3.0 T: a proof-of-concept study. PLoS One. 2021;16(1):e0245970.
  9. Sushentsev N, Kaggie JD, Buonincontri G, et al. The effect of gadolinium-based contrast agent administration on magnetic resonance fingerprinting-based T1 relaxometry in patients with prostate cancer. Sci Rep. 2020;10(1):20475.
  10. Lee YS, Choi MH, Lee YJ, Han D, Kim DH. Magnetic resonance fingerprinting in prostate cancer before and after contrast enhancement. Br J Radiol. 2022;95(1131):20210479.
  11. Han D, Choi MH, Lee YJ, Kim DH. Feasibility of novel three-dimensional magnetic resonance fingerprinting of the prostate gland: phantom and clinical studies. Korean J Radiol. 2021;22(8):1332-1340.
  12. Lo W, Bittencourt LK, Panda A, et al. Multicenter repeatability and reproducibility of MR fingerprinting in phantoms and in prostatic tissue. Magn Reson Med. 2022;88(4):1818-1827.
  13. de Oliveira Correia ET, Qiao PL, Griswold MA, Chen Y, Bittencourt LK. Magnetic resonance fingerprinting based comprehensive quantification of T1 and T2 values of the background prostatic peripheral zone: correlation with clinical and demographic features. Eur J Radiol. 2023;164:110883.

advertisement

advertisement