Introduction

There is a clinical need for quantitative measurements of MRI parameters in diagnostics¹. Reference phantoms containing aqueous solutions with calibrated parameters exist but lack properties like partial volume and structure that exist in-vivo. Current anthropomorphic phantoms aim to provide structure, but miss quantification and detailed anatomy². To shift toward more quantitative anatomy-mimicking solutions, we developed slice phantoms made of doped agar gels³. However, the agar gel process has several shortcomings; heating and cooling is sensitive to evaporation and changes in concentrations, resulting in variability. Managing hot, viscous fluids often leads to air bubble formation that results in susceptibility artifacts. Finally, it is difficult to achieve very long T2s (>200ms) to mimic cerebrospinal fluid (CSF) and blood⁴.
In this work, we propose a process for constructing and altering formulations of UV-curable hydrogels to achieve desired quantitative parameters. Our solutions containing acrylamide monomer, polyethylene glycol diacrylate crosslinker (PEGDA), and water-soluble TPO-Li (Diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide) photoinitiator⁵ are formed upon UV-exposure (Figure 2). We achieved a broad range of T1/T2 parameters by varying quantities of acrylamide, nickel chloride (NiCl2), manganese chloride (MnCl2), and demonstrate filling a stereolithography (SLA) 3D-printed slice phantom with anatomy-mimicking parameters. The phantom was derived from the brainweb⁶, designed in Rhino, optimized to avoid warping, and scaled down to 5" x 4.5" to fit in a Formlabs Form 3 printer (Figure 5a). By printing individual compartments representing white-matter, gray-matter, and CSF, we can control curing.

Methods

To map T1/T2 parameters of non-doped UV gels, we prepared large stock solutions and combined appropriate amounts to make 5mL solutions in aluminum-wrapped amber vials containing 0.007g TPO-Li, 0.0035g PEGDA, and linearly-varied concentrations of acrylamide from 0.15 to 0.6g/mL (Figure 3). To characterize T1/T2 parameters of paramagnetic ions, initial gel solutions were combined with linearly-varying concentrations of NiCl2 or MnCl2 in water (Figure 4c). 4mL of each sample were pipetted into 38mm diameter petri dishes (Figure 1). Samples were cured for 60 seconds in a 395nm UV-LED curing station operating at 101W/cm2 (Figure 1c-d), and sealed with parafilm to prevent evaporation.

Cured gels were scanned and mapped on a 3T GE MR750W scanner (Wakasha, Wisconsin) using inversion recovery T1 Spin-echo, and T2 Spin-echo sequences (TI: 50, 100, 150, 200, 250, 500, 750, 1000, 1250, 1500, 1750, 2000, 2250, 2500 ms; TE: 15, 25, 35, 55, 75, 95, 115, 135, 155, 175, 200, 250, 300, 350, 400 ms). Stickiness and compressibility were measured on a relative scale of 1-5 (maximum 5) by how much the gel adhered to a latex glove and could be mechanically compressed (Figure 3b-c). Stretch ratio (length of gel before breaking divided by length before stretching) was obtained through videos of stretching gels over a cm ruler (Figure 3d).

With scan data, R1=1/T1 and R2=1/T2 were estimated⁷ and fit linearly with dopant concentrations. Using the fit to find targets, a 3D-printed brain slice was fabricated with compartments representing different tissues, then each chamber was filled and cured (Figure 5b). The slice phantom weighted scan was imaged using 3D T1 BRAVO (TI: 225 ms) and T2 Spin-echo sequences (TE: 15ms).

Results

Figure 3 depicts an association between increased concentration of acrylamide and increased R1/R2 values, and between acrylamide concentration and gel properties. Figure 4 shows the (T1, T2) maps and linear fits to R1 and R2 derived from the data. The sample and mapping images (Figure 4a-b) are a subsample representation of the 24 doped samples created for the mapping. The scan results of the SLA-printed slice phantom constructed with doped UV gels hit target (T1, T2) tissue values of (668, 70) ms for white, (949, 102) ms for gray matter, and (1881, 1133) ms for CSF, as depicted in Figure 5.

Discussion

The (T1, T2) quantitative map of acrylamide in Figure 3a shows sharp increases in R1/R2 between 0.4 and 0.45g/mL. Observations of physical gel properties show a large increase in stretch ratio and an increase in stickiness around this concentration. The sharp increase in R1/R2 may be due to a structural state change of the gel between 0.4 and 0.45g/mL.
Success with T1/T2 targets in doped gels was found with 0.1g/mL acrylamide concentration. While we can achieve the T2s of white- and gray-matter via doping, the resulting T1s are shorter than those in tissue. More investigation is needed to modulate T2 while keeping T1 long.
The ability to precisely achieve target tissue parameters depends on repeatability of the gel fabrication process. Large stock solutions decrease error and achieve uniformly cured gels, which was previously difficult to achieve with heat-initiated agar gels. Additionally, the T2 of non-doped UV gels (>1000ms) is an order of magnitude longer than that of non-doped agar gels, allowing mimicking of CSF and blood². Future work includes developing an automated procedure to precisely dispense the gel solution, aiming to achieve boundaryless slice phantoms.

Conclusion

To the best of our knowledge, this is the first demonstration of fabricating MRI phantoms using UV-curable hydrogels. We demonstrated that we are capable of using a high-volume process of combining solutions to create UV-curable gels that can achieve desired quantitative T1/T2 parameters.

Acknowledgements

Authors ST and JH contributed equally. We thank the UC Berkeley Brain Imaging Center for the use of their facilities and our funding source: Siemens Medical Solutions USA, Inc..