Quantitative MRI mOOCT1 Mapping · Exercise 4
From human to mouse: is the IR protocol still good?
Imagine you’re a graduate student in a lab that recently transitioned from human imaging to animal imaging (specifically, mouse models of multiple sclerosis). Because there has never been MRI data acquired in your lab in mouse, your advisor asks you to acquire T1 maps of some healthy mice to get the an estimate of T1 values in white matter so that you can use this value to optimize other pulse sequence protocols (eg protocol parameters for T1-weighted images, optimal flip angle for spoiled gradient angle SNR, etc). You do some literature search and find a paper that measured T1 in mice (Kuo et al. 2005), but using a different pulse sequence.
- a. Using qMRLab, determine if the IR imaging protocol (TR = 1000 ms , TI = [100, 300, 600, 900] ms) you used for your previous human study at 1.5 T (where WM has a T1 of ~600ms) would yield good results if applied to mice (at 9.4 T, (Kuo et al. 2005) reports a WM T1 of ~1700 ms). Assume an SNR of 100. Explain how you’ve reached your conclusion.
- b. If the human T1 mapping 1.5T protocol is not appropriate to image mice at 9.4T, explain via written arguments why this might be the case. Use simulation figures to support your arguments.
- c. Propose a new protocol that might yield more precise and/or accurate T1 maps. Why did you make those changes? How did you determine the improvement?
a. The human protocol on human and mouse WM
| Human WM T1 (ms) | at 1.5 T |
| Mouse WM T1 (ms) | at 9.4 T |
| TR (ms) | |
| TIs (ms) | |
| SNR | |
| Noise trials |
The mean is still close to the truth, but the spread is 2.3× larger for mouse WM (CV 11% vs 5%): the protocol is not good enough for mice.
b. Why the human protocol fails in mice
b. Explain why the human 1.5 T protocol is not appropriate to image mice at 9.4 T. Use simulation figures to support your arguments.
- TR ≪ T1: TR = 1000 ms is 0.6 T1, far from the 2–5 T1 of conventional IR, so the magnetization barely recovers between inversions.
- Small signal range across the four TIs, so noise matters more.
- Long-TR fitting would be badly wrong (see the table).
c. A protocol for mouse WM
| Mouse WM T1 (ms) | |
| TR (ms) | |
| TIs (ms) | |
| SNR |
TR ≈ 3 T1 lets the magnetization recover, and TIs out to ≈ 2.4 T1 sample the whole curve: more precise, even per unit of scan time.
Exercise 4: answers
| Part | Answer |
|---|---|
| a | At SNR 100, the human protocol gives WM T1 with CV ≈ 5% in humans but ≈ 11% in mice: not good enough |
| b | TR = 1000 ms ≈ 0.6 T1: little recovery, a compressed signal range and a broken long-TR assumption |
| c | TR 5000 ms, TI 100 / 1000 / 2000 / 4000 ms: CV ≈ 2.3%, about twice as precise for the same scan time |
All simulations and fits on these slides: qmrust (Rust/WebAssembly port of qMRLab), running in the browser.