Magnetic resonance imaging (MRI) of post-mortem tissue is typically carried out with samples chilled or at room temperature. However, the magnetic susceptibility and relaxivities of different tissue components can have different temperature dependencies. This needs to be taken into account when comparing post mortem results to in vivo data.

Ferritin is the main storage molecule for non-haeme iron in many organisms, which is required for respiration and various other metabolic processes [1]. It is one of a few sources of iron known to occur in sufficient concentrations to affect MRI [2,3]. The ability to map iron distributions is clinically useful in brain, liver and heart iron overload diseases, but to validate in vivo methods we need to be able to compare in vivo and post mortem data.

The aim of this study was to characterise the effect of temperature on the NMR behaviour of ferritin at varying concentrations.

Four cylinders of agar were doped with ferritin at concentrations ranging from 0.5 to 3mg/ml. These cylinders were embedded in a spherical agar phantom of 12cm diameter, in a parallel two by two arrangement. The phantom was then placed inside a tank (Figure 1), around which cold water was pumped, using a Tamson Instruments water bath, to achieve a temperature of 5⁰C. The water round the phantom was then warmed in 5⁰C increments to 35⁰C, allowing the temperature to stabilize after each temperature increment, which took approximately 30 minutes each time. The phantom and surrounding tank were scanned at each temperature in a Philips 7T MRI scanner with the cylinders oriented parallel to the field, using a varying inversion time, inversion recovery fast field echo sequence (1.25x1.125x1.25mm³ voxel, FOV=200x180 x72.5mm³, 7 TI values in the range 192-2652ms) and a multi-echo gradient echo sequence (0.94x0.94x1.7mm³ voxel, FOV=210x210x42.5mm³, TE1= 4ms, ΔTE=4ms, 10 echoes).

R1 and R2* values were calculated from an ROI in the centre of each tube (Figure 2). The signal magnitude from the multi GE sequence was linearly fitted to measure R2* and the polarity restored magnitude data from the inversion recovery was fitted to obtain R1. The phase evolution with TE from an ROI inside each tube was compared to that in an annulus around the tube to calculate the field offset inside the tube which is given by $$\triangle B_{int}=\frac{B_0 \triangle \chi}{2}(\cos^2\theta-\frac{1}{3})$$ where B0 is the scanner field and θ is the angle of the tube to B0. This was used to calculate the susceptibility difference, Δχ.

The susceptibility of each cylinder increased linearly with increasing ferritin concentration with a constant of proportionality of 0.88ppm per mg Fe per gram agar at 293K and decreased with increasing temperature. The gradient of the susceptibility variation with concentration, increased with inverse of the temperature (Figure 3) as expected.

Figure 4 shows that R2* decreased for higher concentrations and increased for lower concentrations with increasing temperature so the R2* values showed less variation with concentration. The R2* of the undoped agar around the tubes was found to increase with temperature, with a similar increase found over multiple ROIs. The gradient of R2* with ferritin concentration was used to calculate the R2* relaxivity which increased significantly with the inverse of the temperature (changing by nearly a factor of 2 over the temperature range 5-35⁰C).

R1 for the ferritin samples decreased with increasing temperature, as did R1 for the undoped agar as shown in Figure 5. Ferritin-induced R1 relaxivity also increased with the inverse of the temperature.

Phantoms and support platforms for post-mortem samples are usually made from agar and this study confirms previous work that showed that the T2 of agar decreases with temperature [4], and showed that agar T1 increases with temperature.

In ferritin-doped agar gels R1 and R2* are both sensitive to the magnetic susceptibility effects of iron. Ferritin contains a superparamagnetic core whose average magnetization has a linear dependence on inverse temperature over the temperature range studied here [1] as confirmed by the results in Figure 3. The more significant increase in R2* with inverse temperature may be explained by the additional temperature sensitivity of water diffusion rates.