MRI from Zero

From the principles of T1 and T2 to fat–muscle separation, 23Na and 31P

2026-10-25

Goals of this lecture

  • Explain what T1 and T2 mean, with a picture
  • Explain how TR and TE relate to T1- and T2-weighted images
  • Explain how fat and muscle are separated (STIR, Dixon, PDFF)
  • Explain what 23Na (TSC, ICwS, TQF) and 31P (PCr, pH) can tell us
  • Know how NMR, MRS and MRI differ, and when to use analysis software and Python

Created by generative AI (Claude). Bibliographic details checked against Crossref. Demos and Python code: Book version (https://lab.nghmst.com/study/en/mri_hands_on/)

MRI in one sentence

Inside a strong magnet, send radio waves to atomic nuclei and listen to the radio waves that come back

  1. In the static field B0, nuclear spins line up and precess (wobble like a top)
  2. A radio wave at that frequency (RF pulse) tips the magnetization over
  3. The tipped magnetization rotates and produces a signal
  4. How it recovers (T1) and fades (T2) differs between tissues → contrast
  5. Gradient fields add position information → image

Sources: Bloch (1946); Purcell et al. (1946); Lauterbur (1973)

1. Nuclei are tiny magnets

Larmor frequency

\[f_0 = \frac{\gamma}{2\pi} B_0\]

Nucleus γ/2π (MHz/T) Frequency at 3 T Sensitivity relative to 1H
1H 42.58 127.7 MHz 1
31P 17.24 51.7 MHz 0.066
23Na 11.26 33.8 MHz 0.093
13C 10.71 32.1 MHz 0.016

Each nucleus has its own frequency → a dedicated coil is needed

Relative sensitivity per equal number of nuclei. Real signal = this × concentration in the body

Why are 23Na images so coarse?

  • Hydrogen in water: about 110 mol/L
  • Sodium in tissue: tens of mmol/L
  • Sensitivity 0.093 × concentration difference → signal about 1/10,000 of 1H

→ Larger voxels, longer scans, many averages

Source: Madelin and Regatte (2013)

2. The NMR signal, T1 and T2

RF pulses and the FID

  • 90° pulse: tips the longitudinal magnetization fully into the transverse plane → maximum signal
  • 180° pulse: flips it upside down (inversion) → used in STIR and IR methods
  • The signal induced in the coil by the rotating magnetization = FID (free induction decay)

T1 and T2

T1 (longitudinal relaxation)

Time for the longitudinal magnetization to recover

\(M_z = M_0(1 - e^{-t/T_1})\)

Short for fat, long for water

T2 (transverse relaxation)

Time for the transverse magnetization to dephase and fade

\(M_{xy} = M_0 e^{-t/T_2}\)

Long for water, short for muscle, longer with edema

Relaxation curves

  • Fat: recovers fast → signal even with short TR → bright on T1-weighted
  • Water: transverse signal lasts → signal even at long TE → bright on T2-weighted

Typical values (3 T, approximate)

Tissue T1 (ms) T2 (ms)
Skeletal muscle ~1400 ~40–50
Fat ~400 ~100
Water (e.g., CSF) ~4000 ~2000
  • In vivo T2 of lower-leg muscle: 41–43 ms
  • Values depend on scanner, sequence and analysis → measure under your own conditions

Sources: Stanisz et al. (2005); Gerhalter et al. (2017); Höger et al. (2023). Fat and water values are approximate

T2* and the spin echo

  • Field inhomogeneity also makes spins drift apart → apparent decay T2* (≤ T2)
  • A 180° pulse reverses the order → fast spins catch up → echo
  • Cancels the inhomogeneity effect so the true T2 can be measured

Source: Hahn (1950)

3. Making an image

Gradients: position → frequency

\[f(x) = \frac{\gamma}{2\pi}(B_0 + G_x x)\]

Role In short
Slice selection Which slice to image
Frequency encoding Horizontal position → frequency
Phase encoding Vertical position → phase (repeated once per row)

Scan time ≈ TR × number of phase encodes × number of averages

Source: Lauterbur (1973)

k-space

  • Center = contrast, periphery = edges and detail
  • Low-resolution 23Na images are close to “center only” → beware partial volume effects

4. Contrast

TR and TE

\[S = PD \cdot (1 - e^{-TR/T_1}) \cdot e^{-TE/T_2}\]

TR TE Bright
T1-weighted short (~500 ms) short (~10 ms) fat
T2-weighted long (~4000 ms) long (~100 ms) water, edema
PD-weighted long short hydrogen-rich tissue

Same phantom, different appearance

TR is the knob for T1 differences, TE for T2 differences

Quantitative maps

  • The brightness of weighted images is relative
  • Research estimates the values themselves, pixel by pixel
    • T2 map: fit \(S_0 e^{-TE/T_2}\) to multi-echo data
    • T1 map: inversion recovery, variable flip angle
    • MR fingerprinting: T1, T2 and fat fraction at once

Sources: Ma et al. (2013); Höger et al. (2023)

5. Separating fat and muscle

Two clues

  1. T1 difference: fat has a short T1 → bright on T1-weighted, can be nulled with STIR
  2. Chemical shift: fat protons precess about 3.4 ppm slower than water protons

\[\Delta f = 3.4 \times 10^{-6} \times 127.7\ \text{MHz} \approx 434\ \text{Hz}\ (3\ \text{T})\]

Fat suppression and STIR

  • Fat Sat (CHESS): excite only the fat frequency first and spoil it. Sensitive to field inhomogeneity
  • STIR: 180° inversion → fat crosses zero after TI

\[M_z(TI) = M_0(1 - 2e^{-TI/T_1}) \Rightarrow TI_{null} = T_1 \ln 2\]

  • Fat T1 ≈ 400 ms → TI ≈ 280 ms (example: TIRM with TI = 260 ms)

The same idea applied to 23Na is the IR (intracellular-weighted) method

Source: Höger et al. (2023)

The Dixon method

  • In-phase: \(S_{IP} = W + F\), opposed-phase: \(S_{OP} = W - F\)
  • \(W = (S_{IP}+S_{OP})/2\), \(F = (S_{IP}-S_{OP})/2\), \(FF = F/(W+F)\)
  • At 3 T: OP ≈ 1.15 ms, IP ≈ 2.3 ms
  • Vendor Dixon methods use phase correction and multiple echoes to avoid water–fat swaps

Source: Dixon (1984)

Dixon example (phantom)

  • The fat-infiltrated area (FF 15%) is darker on opposed-phase and visible on the fat-fraction map

PDFF and which method to use

  • PDFF: fat fraction corrected for multiple fat peaks and T2*. The standard measure from multi-echo Dixon
Purpose Method
Muscle/fat cross-sectional area, volume T1-weighted or Dixon → segmentation
Fat inside muscle PDFF from multi-echo Dixon
Edema, inflammation STIR, fat-suppressed T2, water T2 map

Sources: Reeder et al. (2012); Hu and Kan (2013)

6. Multinuclear: 23Na and 31P

Features of 23Na

  • Spin-3/2 quadrupolar nucleus → very short T2
    • short 0.8 ms, long 12.4 ms (lower-leg muscle)
  • Short T1 too (~34 ms) → short TR, many averages
  • Needs ultrashort TE (UTE, 0.3 ms) 3D radial acquisition
Concentration Motion
Extracellular fluid ~140 mmol/L free
Intracellular fluid ~10–15 mmol/L restricted

Sources: Gerhalter et al. (2017); Höger et al. (2023); Marton et al. (2026)

TSC and the two-compartment model

  • TSC: average concentration of all Na in a voxel (mmol/L tissue)
  • Signal converted to concentration with reference phantoms (20 and 40 mmol/L)

\[TSC \approx f_{ECW} \times 140 + f_{ICW} \times 10\]

Extracellular Na is 14× more concentrated → a small rise in extracellular fluid raises TSC a lot

  • Muscle TSC ranges 10–25 mmol/L across studies → compare absolute values with care

Sources: Höger et al. (2023); Marton et al. (2026)

ICwS and TQF

  • ICwS (IR method): image at the TI where free Na (longer T1) crosses zero
    • TI = 34 ms → nulls the component with \(T_1 = 34/\ln 2 \approx 49\) ms
    • Not a complete separation → “intracellular-weighted”
  • TQF: uses quadrupolar interaction to select only Na bound to macromolecules (includes some interstitial Na)

Sources: Nagel et al. (2011); Höger et al. (2023); Gerhalter et al. (2017)

What 23Na has shown (muscle)

Study Finding
Eccentric exercise, DOMS TSC and ICwS ↑ right after, back by 48 h. Water T1/T2 ↑ at 48 h
Manipulating vascular volume 23Na measures more sensitive to extracellular fluid changes than 1H T2
Hypertension, aldosteronism Na stored in muscle, reduced by KCl supplementation

Sources: Höger et al. (2023); Gerhalter et al. (2017); Marton et al. (2026)

31P MRS

  • Peaks: Pi (~+4.9), PCr (0), γ-, α-, β-ATP (−2.5, −7.5, −16 ppm)
  • Resting muscle: PCr ~30 mmol/L, ATP ~8 mmol/L
  • pH: calculated from the Pi–PCr shift δ
  • PCr recovery time constant τ after exercise: shorter = higher mitochondrial capacity

\[pH = 6.75 + \log_{10}\frac{\delta - 3.27}{5.63 - \delta}\]

Sources: Kemp et al. (2007); Moon and Richards (1973); Meyerspeer et al. (2020)

31P spectrum and PCr recovery (synthetic data)

7. NMR, MRS and MRI

The difference: how much position information is used

Position What you learn
NMR none (whole sample) which molecules, in what state
MRS selected region metabolites at that location
MRI per pixel where, and how much

The more position you give up, the stronger, faster and more detailed the signal

  • Example: a whole-calf 23Na spectrum gives many measures in 14 minutes

Source: Gerhalter et al. (2017)

8. Analysis

Data and software

  • DICOM (scanner) → dcm2niix → NIfTI (research). Beware personal data in DICOM
Purpose Software
ROI, segmentation 3D Slicer, ITK-SNAP
Brain imaging FSL, FreeSurfer, SPM, ANTs
MRS LCModel, jMRUI, OXSA, FSL-MRS
Custom analysis Python (nibabel, pydicom, SimpleITK, scipy), MATLAB, R

Sources: Fedorov et al. (2012); Yushkevich et al. (2006); Jenkinson et al. (2012)

Python: T2 map and TSC calibration

  • T2: log-linear or nonlinear fit (per pixel)
  • TSC: calibration line from phantoms (0, 20, 40 mmol/L)

Full code in Chapter 8 of the Book version

Summary

  • T1 = recovery time, T2 = fading time. TR and TE choose which one you see
  • Fat and muscle are separated by T1 differences and chemical shift (STIR, Dixon, PDFF)
  • 23Na: TSC, ICwS and TQF with UTE. TSC is sensitive to extracellular fluid
  • 31P: metabolism and pH from PCr, ATP and Pi; mitochondrial capacity from PCr recovery
  • The more NMR/MRS gives up position, the higher the sensitivity

References

Bloch, F. 1946. “Nuclear Induction.” Physical Review 70: 460–74. https://doi.org/10.1103/PhysRev.70.460.
Dixon, W. T. 1984. “Simple Proton Spectroscopic Imaging.” Radiology 153: 189–94. https://doi.org/10.1148/radiology.153.1.6089263.
Fedorov, A., R. Beichel, J. Kalpathy-Cramer, et al. 2012. “3D Slicer as an Image Computing Platform for the Quantitative Imaging Network.” Magnetic Resonance Imaging 30: 1323–41. https://doi.org/10.1016/j.mri.2012.05.001.
Gerhalter, T., P. G. Carlier, and B. Marty. 2017. “Acute Changes in Extracellular Volume Fraction in Skeletal Muscle Monitored by 23Na NMR Spectroscopy.” Physiological Reports 5 (16): e13380. https://doi.org/10.14814/phy2.13380.
Hahn, E. L. 1950. “Spin Echoes.” Physical Review 80: 580–94. https://doi.org/10.1103/PhysRev.80.580.
Höger, S. A., L. V. Gast, B. Marty, et al. 2023. “Sodium and Quantitative Hydrogen Parameter Changes in Muscle Tissue After Eccentric Exercise and in Delayed-Onset Muscle Soreness Assessed with Magnetic Resonance Imaging.” NMR in Biomedicine 36 (2): e4840. https://doi.org/10.1002/nbm.4840.
Hu, H. H., and H. E. Kan. 2013. “Quantitative Proton MR Techniques for Measuring Fat.” NMR in Biomedicine 26: 1609–29. https://doi.org/10.1002/nbm.3025.
Jenkinson, M., C. F. Beckmann, T. E. J. Behrens, M. W. Woolrich, and S. M. Smith. 2012. “FSL.” NeuroImage 62: 782–90. https://doi.org/10.1016/j.neuroimage.2011.09.015.
Kemp, G. J., M. Meyerspeer, and E. Moser. 2007. “Absolute Quantification of Phosphorus Metabolite Concentrations in Human Muscle in Vivo by 31P MRS: A Quantitative Review.” NMR in Biomedicine 20: 555–65. https://doi.org/10.1002/nbm.1192.
Lauterbur, P. C. 1973. “Image Formation by Induced Local Interactions: Examples Employing Nuclear Magnetic Resonance.” Nature 242: 190–91. https://doi.org/10.1038/242190a0.
Ma, D., V. Gulani, N. Seiberlich, et al. 2013. “Magnetic Resonance Fingerprinting.” Nature 495: 187–92. https://doi.org/10.1038/nature11971.
Madelin, G., and R. R. Regatte. 2013. “Biomedical Applications of Sodium MRI in Vivo.” Journal of Magnetic Resonance Imaging 38: 511–29. https://doi.org/10.1002/jmri.24168.
Marton, A., H. T. N. Tran, N. Morisawa, et al. 2026. “Oral KCl Supplementation Safely Reduces Body Na+ Surplus and Blood Pressure.” Circulation Research 139: e329131. https://doi.org/10.1161/CIRCRESAHA.126.329131.
Meyerspeer, M. et al. 2020. “31P Magnetic Resonance Spectroscopy in Skeletal Muscle: Experts’ Consensus Recommendations.” NMR in Biomedicine 34: e4246. https://doi.org/10.1002/nbm.4246.
Moon, R. B., and J. H. Richards. 1973. “Determination of Intracellular pH by 31P Magnetic Resonance.” Journal of Biological Chemistry 248: 7276–78. https://doi.org/10.1016/S0021-9258(19)43389-9.
Nagel, A. M. et al. 2011. “3 Tesla Sodium Inversion Recovery Magnetic Resonance Imaging Allows for Improved Visualization of Intracellular Sodium Content Changes in Muscular Channelopathies.” Investigative Radiology 46: 759–66. https://doi.org/10.1097/RLI.0b013e31822836f6.
Purcell, E. M., H. C. Torrey, and R. V. Pound. 1946. “Resonance Absorption by Nuclear Magnetic Moments in a Solid.” Physical Review 69: 37–38. https://doi.org/10.1103/PhysRev.69.37.
Reeder, S. B., H. H. Hu, and C. B. Sirlin. 2012. “Proton Density Fat-Fraction: A Standardized MR-Based Biomarker of Tissue Fat Concentration.” Journal of Magnetic Resonance Imaging 36: 1011–14. https://doi.org/10.1002/jmri.23741.
Stanisz, G. J., E. E. Odrobina, J. Pun, et al. 2005. “T1, T2 Relaxation and Magnetization Transfer in Tissue at 3T.” Magnetic Resonance in Medicine 54: 507–12. https://doi.org/10.1002/mrm.20605.
Yushkevich, P. A., J. Piven, H. C. Hazlett, et al. 2006. “User-Guided 3D Active Contour Segmentation of Anatomical Structures: Significantly Improved Efficiency and Reliability.” NeuroImage 31: 1116–28. https://doi.org/10.1016/j.neuroimage.2006.01.015.