Introduction

Of the more than 32,000 MRI scanners around the world¹, very few are equipped with the latest hardware and software required for advanced imaging. Improving the speed and quality of images which can be collected on older scanners could help to provide state-of-the-art imaging to a larger fraction of the world’s population. The Magnetic Resonance Fingerprinting (MRF) framework presents an intriguing opportunity, as MRF is rapid and quantitative, and does not require the latest multi-channel receiver coils or rapid gradient systems. In this work, we explore the use of MRF to generate repeatable and reproducible T₁ and T₂ maps on older MRI scanners as a step towards replacing slow and qualitative imaging, and towards implementing MRF on low cost scanner platforms.

Methods

The ISMRM/NIST phantom² was scanned on a 1.5T Magnetom Symphony (Siemens Healthcare AG, Erlangen, Germany), which was delivered in 2002, and underwent a TIM gradient system upgrade in 2015. This scanner is equipped with a 4-channel head receive coil and a gradient system with a maximum strength of 22 mT/m and maximum slew rate of 100 mT/m/s. Maps were obtained using a FISP-based MRF sequence^3,4 with an in-plane spatial resolution of 1.6×1.6 mm², 5 mm slice thickness, and scan time of 26s. The spiral trajectory was optimized for the older gradient system by lengthening the readout to reduce the slew rate, and measured in an axial orientation prior to the first scan. Scans were repeated five times on this 16-year old scanner, and a test-retest experiment was also performed on two different days. To test the ability to image at arbitrary orientations, scans were also acquired with the phantom at a double oblique orientation. T₁ and T₂ maps for all scans were generated by pattern matching to a dictionary with T₁ ranging from 10 to 4500 ms and T₂ ranging from 2 to 1000 ms, corrected for slice profile imperfections⁵. Average T₁ and T₂ values within circular ROIs were recorded (T₂ up to 300ms²) for all five scans. Across the scans, average T₁ and T₂ values were compared to known reference values using linear correlation coefficients. Measurement variability was assessed by the coefficient of variation (CV). Reproducibility with the 16-year-old Symphony scanner was assessed by the correlation of measurements from the two different days and between the axial and double oblique orientations. Additionally, a proof-of-concept in vivo brain MRF scan was performed on the Symphony for a healthy volunteer.

Results

T₁ and T₂ maps from the NIST phantom in the axial and double oblique orientations are shown in Figure 1. Quantitatively, the average values measured in both orientations demonstrated a strong correlation with reference T₁ and T₂ values (R²≥0.99, Figure 2). Reproducibility of T₁ and T₂ measurements using the 16-year-old Symphony was good across the two different days and the two scan planes (R²≥0.99, Figure 3a-3b). The correlation of average T₁ and T₂ values measured on the Symphony to previously reported values measured on a modern 1.5T Aera scanner was also strong (R²≥0.99, Figure 3c). Average CV values for T₁ and T₂ were 2.0% and 3.1% (Figure 4). The proof-of-concept in vivo scan resulted in T₁ and T₂ maps where average white matter T₁ and T₂ values were 673 ms and 41.0 ms, respectively (Figure 5).

Discussion

This study demonstrates that T₁ and T₂ measurements with MRF are feasible on legacy MRI hardware. MRF maps were reproducible and comparable between the double oblique and axial geometries, suggesting that MRF can be performed across a range of orientations necessary for the clinical setting. The variability in T₁ and T₂ measurements was higher on the 16-year-old Symphony than expected on modern MRI scanners, but accurate parameter maps could be obtain in the NIST phantom with the rapid 26 second sequence. The proof-of-concept in vivo maps suggest that potential improvements exist: low-signal regions require compensation and the initially measured T₁ and T₂ values may need correction for possible T₂* or diffusion effects due to the limited gradient strength and long spiral readout required. While preliminary, these findings suggest that accelerated quantitative imaging strategies such as MRF could be applied to a large fraction of the install base of MRI scanners, and not just those built to the most modern specifications, potentially expanding the availability of these technologies to patients in both the developed and the developing world.

Conclusion

MRF is feasible on older scanners with weaker gradient systems and a relatively small number of receiver coils. This work suggests that newer data collection and processing approaches like MRF may enable older scanners to be used for state-of-the-art imaging, providing access to the benefits of these technologies to a larger population.

Acknowledgements

The authors acknowledge the assistance of Vladmir Nadtotchi, Ronald Collister, and Naren Nallapareddy in coordinating and performing the scans. This work was funded in part by the following sources: NIH R01HL094557, R01DK098503, R01EB016728, C06RR12463-01; NSF CBET 1553441, Siemens Healthineers (Erlangen, Germany). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.