Magnetic resonance imaging physics: Bloch equations, relaxation, and the Lauterbur-Mansfield Nobel revolution

Intuition Beginner

Magnetic resonance imaging (MRI) makes detailed three-dimensional pictures of the inside of your body without using X-rays. It works because your body is about 60 percent water, and the hydrogen nuclei in water (single protons) act like tiny magnets. In a strong magnetic field, the protons line up. A brief radio pulse tips them out of alignment; as they relax back, they emit radio signals that the scanner detects.

Different tissues have different relaxation times, which is why MRI can tell them apart. Gray matter, white matter, spinal fluid, fat, and muscle each return to equilibrium at different speeds. By tuning the scanner to weight the signal toward one relaxation time or another, radiologists get the contrast they need.

Paul Lauterbur (1973, Nature) was the first to add magnetic-field gradients and produce a two-dimensional image. Peter Mansfield (1977, echo-planar imaging) developed a single-shot technique that made fast imaging possible — the basis of functional MRI. They shared the 2003 Nobel Prize in Medicine. About 50 million MRI scans are performed annually; MRI has no ionizing radiation and is essential for neurology, oncology, and musculoskeletal diagnosis.

Visual Beginner

The picture shows hydrogen protons as small spinning arrows inside a magnetic field, precessing around the field direction at the Larmor frequency. A radio-frequency pulse tips the arrows sideways; as they relax back, two clocks run at once — T1 (the recovery of the longitudinal component) and T2 (the decay of the transverse component). Magnetic-field gradients then encode spatial position into the radio signal, and a computer reconstructs the image.

The bottom of the picture shows two clinical images side by side: a T1-weighted scan, where fat is bright and water is dark, and a T2-weighted scan, where water is bright — useful for spotting edema, tumors, and inflammation.

Worked example Beginner

In 1992, Kwong and colleagues at Massachusetts General Hospital scanned a volunteer's visual cortex during rest and during a flashing-light stimulus. They used T2-star-weighted imaging, which is sensitive to the oxygenation state of hemoglobin.

Step 1: deoxyhemoglobin is paramagnetic, meaning it disturbs the local magnetic field. Oxyhemoglobin is not. So the local T2-star signal depends on how much deoxyhemoglobin sits in the nearby blood vessels.

Step 2: when the visual cortex activated, blood flow increased faster than oxygen extraction — so the local deoxyhemoglobin actually decreased, the local field became more uniform, and the T2-star signal lengthened. The activated cortex paradoxically appeared brighter on the image.

Step 3: subtracting the rest image from the stimulation image produced a bright spot exactly over primary visual cortex. This is the BOLD (Blood Oxygen Level Dependent) contrast mechanism, discovered by Ogawa in 1990.

What this tells us: a 6-percent BOLD signal change, riding on a slow 4-to-6-second hemodynamic lag, is enough to map human brain activity without opening the skull — launching functional MRI, now used in over 100,000 neuroimaging studies.

Check your understanding Beginner

Formal definition Intermediate+

Definition (nuclear magnetic resonance and the magnetisation vector). A nucleus with non-zero spin $I$ carries a magnetic moment $\mathbf{\mu }=\gamma \mathbf{I}$, where $\gamma$ is the gyromagnetic ratio. For the hydrogen-1 nucleus (the dominant source of clinical MRI signal), $\gamma /2\pi \approx 42.58MHz/T$. A sample of such nuclei in a static field ${\mathbf{B}}_0=B_0\hat{\mathbf{z}}$ carries a bulk magnetisation vector $\mathbf{M}(t)=(M_x(t),M_y(t),M_z(t))$ defined as the nuclear-spin magnetic moment per unit volume. The classical equation of motion for $\mathbf{M}$ in a field $\mathbf{B}$ is $d\mathbf{M}/dt=\gamma \mathbf{M}\times \mathbf{B}$, describing precession about $\mathbf{B}$ at the Larmor frequency ${\omega }_0=\gamma B_0$.

Definition (T1 and T2 relaxation). Following a radio-frequency pulse that tips $\mathbf{M}$ away from $\hat{\mathbf{z}}$, two independent relaxation processes return the system to equilibrium $M_0\hat{\mathbf{z}}$:

• T1 (spin-lattice, longitudinal) relaxation governs recovery of $M_z$ towards $M_0$: $M_z(t)=M_0(1-e^{-t/T_1})$ after a 90-degree pulse, where $T_1$ is the time constant for energy exchange with the molecular lattice.
• T2 (spin-spin, transverse) relaxation governs decay of $M_{xy}=(M_x^2+M_y^2{)}^{1/2}$: $M_{xy}(t)=M_{xy}(0)e^{-t/T_2}$, where $T_2$ is the time constant for loss of phase coherence among precessing spins.

Tissue-dependent relaxation times generate image contrast: CSF has long $T_1/T_2\approx 4000/2000\mathrm{ms}$; fat has short $T_1/T_2\approx 250/80\mathrm{ms}$; gray matter $\approx 1000/100\mathrm{ms}$; white matter $\approx 800/80\mathrm{ms}$ (values at 1.5 T).

Definition (spatial encoding via gradients). A linear magnetic-field gradient $\mathbf{G}(t)=(G_x,G_y,G_z)$ superimposed on ${\mathbf{B}}_0$ makes the Larmor frequency depend on position: $\omega (\mathbf{r},t)=\gamma (B_0+\mathbf{G}(t)\cdot \mathbf{r})$. Three orthogonal gradients encode the three spatial dimensions: slice selection ($G_z$ applied during the RF pulse selects a thin slab of resonating protons); frequency encoding ($G_x$ applied during signal readout makes precession frequency linear in $x$); phase encoding ($G_y$ applied briefly between excitation and readout encodes $y$-position into the accumulated phase ${\varphi }_y=\gamma G_{y}y\tau$).

Definition (k-space and Fourier reconstruction). The raw MRI signal $s(t)$ equals the 2D Fourier transform of the transverse magnetisation sampled along a trajectory in k-space, where $\mathbf{k}(t)=\gamma {\int }_0^t\mathbf{G}(t^{\prime })d{t}^{\prime }$. The reconstructed image is the inverse 2D Fourier transform of $s(\mathbf{k})$. Different pulse sequences trace different k-space trajectories — Lauterbur's 1973 radial back-projection, Edelstein-Hutchison 1980 rectilinear spin-warp, Mansfield 1977 zig-zag echo-planar — but all reconstruct by Fourier inversion.

Counterexamples to common slips Intermediate+

• "MRI uses radiation." No. The scanner transmits RF photons at megahertz frequencies — about $10^{-7}$ times the energy of X-ray photons — and uses static magnetic fields. There is no ionizing radiation.
• "T1 and T2 are the same thing." No. T1 measures recovery of the longitudinal magnetisation $M_z$ (energy return to the lattice); T2 measures decay of the transverse magnetisation $M_{xy}$ (loss of phase coherence among spins). Always $T_2\le T_1$, and in biological tissues typically $T_2\approx T_1/5$ to $T_1/10$.
• "MRI is unsafe for everyone with any metal implant." Mostly false. The absolute contraindications are ferromagnetic implants — older pacemakers, certain aneurysm clips, cochlear implants, metal fragments in the eyes. Modern MR-conditional pacemakers and most orthopaedic hardware (titanium) are safe at 1.5 T.
• "fMRI directly measures neural activity." No. fMRI measures the BOLD hemodynamic response, which lags neural activity by 4 to 6 seconds and reflects a complex coupling of blood flow, blood volume, and oxygen metabolism. It is an indirect surrogate for neural activity.
• "Higher field strength always improves MRI." Field strength improves signal-to-noise (roughly linearly), but also increases susceptibility artifacts, RF heating (specific absorption rate), and cost. 1.5 T is the clinical workhorse; 3 T is common in academic centres; 7 T is the practical clinical ceiling as of 2026.

Key result: the Bloch equations and T1/T2 relaxation Intermediate+

Theorem (Bloch 1946 — the equations of motion for nuclear magnetisation). For a sample of spin-$1/2$ nuclei with gyromagnetic ratio $\gamma$ in a magnetic field $\mathbf{B}(\mathbf{r},t)={\mathbf{B}}_0+\mathbf{G}(t)\cdot \mathbf{r}+{\mathbf{B}}_1(t)$, the bulk magnetisation vector $\mathbf{M}(\mathbf{r},t)$ satisfies the Bloch equations

$$\frac{d\mathbf{M}}{dt}=\gamma \mathbf{M}\times \mathbf{B}-\frac{M_x\hat{\mathbf{x}}+M_y\hat{\mathbf{y}}}{T_2}-\frac{(M_z-M_0)\hat{\mathbf{z}}}{T_1},$$

where ${\mathbf{B}}_1(t)$ is the applied RF field (rotating at the Larmor frequency in the transverse plane), and $T_1,T_2>0$ are the longitudinal and transverse relaxation times. In the rotating frame at frequency ${\omega }_0=\gamma B_0$, with ${\mathbf{B}}_1(t)=B_1(t)(\cos \omega t,-\sin \omega t,0)$, the on-resonance solutions following a 90-degree pulse are $M_z(t)=M_0(1-e^{-t/T_1})$ and $M_{xy}(t)=M_0{e}^{-t/T_2}$.

Proof.

Step 1 — the bare precession equation. From classical electromagnetism, a magnetic moment $\mathbf{\mu }$ in a field $\mathbf{B}$ experiences the torque $\mathbf{\tau }=\mathbf{\mu }\times \mathbf{B}$. Equating torque to the rate of change of angular momentum, $d\mathbf{I}/dt=\mathbf{\tau }$, and using $\mathbf{\mu }=\gamma \mathbf{I}$, gives $d\mathbf{\mu }/dt=\gamma \mathbf{\mu }\times \mathbf{B}$. Summing over the ensemble of nuclei per unit volume yields $d\mathbf{M}/dt=\gamma \mathbf{M}\times \mathbf{B}$, the Bloch equation without relaxation. Quantum-mechanically, this classical equation reproduces the Heisenberg equation $d\hat{\mathbf{M}}/dt=(i/\hslash )[\hat{H},\hat{\mathbf{M}}]$ for the spin operator expectation value under the Zeeman Hamiltonian $\hat{H}=-\gamma \mathbf{B}\cdot \hat{\mathbf{I}}$; the classical form is exact for spin-$1/2$ expectation values.

Step 2 — adding T1 recovery. After the RF pulse, the $z$-component of $\mathbf{M}$ must return to its thermal-equilibrium value $M_0=N{\gamma }^2{\hslash }^{2}I(I+1)B_0/(3k_{B}T)$. By time-dependent perturbation theory and the fluctuation-dissipation relation, the relaxation is exponential with a time constant $T_1$ set by the spectral density of molecular motion at the Larmor frequency (Bloembergen-Purcell-Pound 1948). The phenomenological term $-(M_z-M_0)\hat{\mathbf{z}}/T_1$ captures this energy exchange with the lattice.

Step 3 — adding T2 dephasing. The transverse components $M_x,M_y$ decay for two reasons: (i) irreversible loss of phase coherence from random spin-spin interactions (true $T_2$), and (ii) reversible dephasing from microscopic field inhomogeneities (called $T_2^{\prime }$). Together they produce the observed $T_2^{\ast }$ with $1/T_2^{\ast }=1/T_2+1/T_2^{\prime }$. The phenomenological terms $-M_x\hat{\mathbf{x}}/T_2-M_y\hat{\mathbf{y}}/T_2$ capture the irreversible part.

Step 4 — solution in the rotating frame. Transform to the frame rotating at ${\omega }_0$ about $\hat{\mathbf{z}}$. On resonance, the effective field in the rotating frame is ${\mathbf{B}}_{\text{eff}}=B_1(t){\hat{\mathbf{x}}}^{\prime }$, so a pulse of duration $\tau =\pi /(2\gamma B_1)$ rotates $\mathbf{M}$ from $\hat{\mathbf{z}}$ into the transverse plane. Setting ${\mathbf{B}}_1=0$ after the pulse, the equations decouple: $d{M}_z/dt=-(M_z-M_0)/T_1$ integrates to $M_z(t)=M_0(1-e^{-t/T_1})$, and $d{M}_{xy}/dt=-M_{xy}/T_2$ integrates to $M_{xy}(t)=M_{xy}(0)e^{-t/T_2}$. The detected signal — the free induction decay — is the transverse component oscillating at ${\omega }_0$ with envelope $e^{-t/T_2^{\ast }}$. $\square$

Bridge. The Bloch equations are the foundational reason every clinical MRI pulse sequence works: a 90-degree pulse followed by a short echo time $TE\ll T_1$ produces $T_2$-weighted contrast, while a long repetition time $TR\gg T_1$ followed by a short $TE$ produces $T_1$-weighted contrast, and the tissue-dependent relaxation times identified above with the Bloch solutions are exactly the source of all clinical image contrast. This builds toward the k-space reconstruction framework that appears again in 35.09.01 medical diagnostics survey, where the spatial-encoding argument is given its clinical-imaging context, and the central insight is that Lauterbur's gradient innovation identifies position in the sample with a frequency in the detected signal — the bridge is between Bloch's nuclear-induction ODE and the modern radiological image, generalising from a one-dimensional free-induction decay to a full three-dimensional image via gradient-controlled k-space traversal.

Exercises Intermediate+

Advanced results Master

Result 1 (Bloch 1946 — the Bloch equations). Felix Bloch's 1946 Physical Review 70:460 paper "Nuclear Induction" ^[Bloch1946] introduced the phenomenological equations of motion for the bulk magnetisation vector that bear his name. Bloch derived the equations from classical torque arguments, verified them with the first observation of nuclear induction in water, and identified the Larmor frequency, $T_1$, and $T_2$ as the load-bearing parameters. Bloch shared the 1952 Nobel Prize in Physics with Edward Purcell, who independently observed nuclear magnetic resonance in condensed matter (Purcell-Torrey-Pound 1946 Phys. Rev. 69:37 ^{[Purcell1946]}).

Result 2 (Lauterbur 1973 — the first MR image). Paul Lauterbur's 1973 Nature 242:190 paper "Image Formation by Induced Local Interactions" ^{[Lauterbur1973]} showed that applying a magnetic-field gradient makes the NMR signal spatially dependent, and that back-projection of the gradient-encoded signal reconstructs a two-dimensional image. Lauterbur's first published image was two 1-mm capillary tubes of heavy water in ordinary water. This is the foundational imaging paper of the entire MRI field; Lauterbur shared the 2003 Nobel Prize in Physiology or Medicine.

Result 3 (Mansfield 1977 — echo-planar imaging). Peter Mansfield's 1977 J. Phys. C 10 paper "Multi-planar image formation using NMR spin echoes" ^{[Mansfield1977]} introduced echo-planar imaging (EPI): a single-shot acquisition that rapidly alternates the readout gradient while blipping the phase-encode gradient, tracing the entire k-space zig-zag trajectory in one TR (about 50 ms). EPI is the technical enabler of functional MRI, real-time cardiac MRI, and diffusion tensor imaging. Mansfield shared the 2003 Nobel Prize with Lauterbur.

Result 4 (Edelstein-Hutchison 1980 — spin-warp imaging). Edelstein, Hutchison and colleagues (1980, Phys. Med. Biol. 25:751) introduced the spin-warp method, in which the phase-encode gradient is stepped through a regular grid of values while the readout gradient stays fixed. Spin-warp is the basis of nearly all clinical MRI today: it is robust, has well-understood artifacts, and reconstructs cleanly via 2D inverse Fourier transform. The combination of spin-warp for structural imaging and EPI for functional imaging covers the great majority of clinical pulse sequences.

Result 5 (Le Bihan 1986 — diffusion MRI). Le Bihan's 1986 introduction of diffusion-sensitised MRI (Radiology 161:401) applied strong bipolar gradients (the Stejskal-Tanner 1965 pulse pair) to make the MR signal sensitive to the Brownian motion of water. Diffusion-weighted imaging (DWI) detects acute stroke within minutes of onset, because cytotoxic edema restricts water diffusion. Diffusion-tensor imaging (DTI), introduced by Basser-Mattiello-LeBihan 1994, fits a 3x3 diffusion tensor at each voxel and derives fractional anisotropy and principal-diffusion-direction maps, enabling in-vivo tracing of white-matter tracts (tractography).

Result 6 (Ogawa 1990 and Kwong 1992 — BOLD fMRI). Ogawa and colleagues' 1990 PNAS 87:9868 paper ^[Ogawa1990] demonstrated that the MR signal in rat brain at high field depended on blood oxygenation, because deoxyhemoglobin is paramagnetic and dephases surrounding tissue water — the BOLD (Blood Oxygen Level Dependent) contrast mechanism. Kwong and colleagues' 1992 PNAS 89:5675 paper ^[Kwong1992] produced the first human functional brain maps from the BOLD signal, showing visual-cortex activation during a flashing-light stimulus. The combination launched functional MRI as the dominant tool of cognitive neuroscience, with over 100,000 published studies as of 2026.

Result 7 (modern extensions). 7-Tesla human MRI (UHF), approved by the FDA in 2017, exploits the linear SNR-field scaling to provide sub-millimetre cortical-layer imaging. Hyperpolarised 13C MRI (Ardenkjaer-Larson 2003) dissolves hyperpolarised carbon-13 substrates into a patient, where the 10,000-fold signal enhancement enables real-time metabolic imaging of prostate and liver tumours. MR elastography (Muthupillai 1995) images shear-wave propagation induced by an external driver, producing quantitative maps of tissue stiffness — the clinical standard for staging liver fibrosis non-invasively. Real-time MRI (Uecker 2010) uses radial golden-angle k-space sampling and compressed-sensing reconstruction to produce 20-to-50-frame-per-second cardiac and interventional MRI.

Synthesis. The 80-year arc from Bloch's 1946 nuclear-induction paper to modern 7-Tesla multimodal MRI is the foundational reason magnetic resonance imaging sits alongside X-ray imaging and ultrasound as a pillar of clinical radiology. The central insight — that a sample of nuclear spins obeys a classical ODE whose solutions carry tissue-identifying parameters $T_1$ and $T_2$ — appears again in every advanced result above, and the pattern generalises from Bloch's original one-dimensional free-induction decay to two-dimensional images (Lauterbur 1973, Edelstein 1980), to single-shot volume acquisition (Mansfield 1977), to microstructural contrast (Le Bihan 1986, diffusion), to hemodynamic functional imaging (Ogawa 1990, Kwong 1992, BOLD), to metabolic and mechanical tissue characterization (hyperpolarised 13C, MR elastography). Putting these together with the gradient-k-space-Fourier reconstruction pipeline identifies Bloch's phenomenological ODE with the modern radiological image, and this is exactly the bridge between nuclear-spin physics and clinical diagnosis; the bridge is also the reason a single 1.5-Tesla scanner can produce T1-weighted, T2-weighted, FLAIR, diffusion, perfusion, and functional images in one 30-minute session, simply by reprogramming the pulse sequence. The same mathematical structure builds toward 35.03.05 neurodegenerative disease, where MRI is the primary diagnostic imaging modality for Alzheimer's, Parkinson's, and multiple sclerosis, and the Bloch-equation relaxation model appears again in 29.03.04 Hubel-Wiesel visual cortex, where BOLD fMRI is the dominant method for modern visual-system neuroscience.

Full proof set Master

Proposition 1 (Larmor precession follows from the spin Zeeman Hamiltonian). For a spin-$1/2$ nucleus with gyromagnetic ratio $\gamma$ in a static field $B_0\hat{\mathbf{z}}$, the expectation values $⟨{\hat{M}}_x(t)⟩,⟨{\hat{M}}_y(t)⟩,⟨{\hat{M}}_z(t)⟩$ of the magnetic-moment operator $\hat{\mathbf{M}}=\gamma \hat{\mathbf{I}}$ obey $d⟨\hat{\mathbf{M}}⟩/dt=\gamma ⟨\hat{\mathbf{M}}⟩\times \mathbf{B}$, equivalent to the Bloch precession equation (without relaxation) at the Larmor frequency ${\omega }_0=\gamma B_0$.

Proof. The Zeeman Hamiltonian is $\hat{H}=-\hat{\mathbf{M}}\cdot {\mathbf{B}}_0=-\gamma B_0{\hat{I}}_z$. By the Heisenberg equation, $d⟨{\hat{M}}_i⟩/dt=(i/\hslash )⟨[\hat{H},{\hat{M}}_i]⟩$. Using the angular-momentum commutator $[{\hat{I}}_i,{\hat{I}}_j]=i\hslash {ϵ}_{ijk}{\hat{I}}_k$ and the linearity of $\hat{\mathbf{M}}$ in $\hat{\mathbf{I}}$, $[\hat{H},{\hat{M}}_x]=-\gamma B_0[{\hat{I}}_z,\gamma {\hat{I}}_x]=-i\hslash \gamma B_0{\hat{M}}_y$, giving $d⟨{\hat{M}}_x⟩/dt=\gamma B_0⟨{\hat{M}}_y⟩$. Cyclically, $d⟨{\hat{M}}_y⟩/dt=-\gamma B_0⟨{\hat{M}}_x⟩$ and $d⟨{\hat{M}}_z⟩/dt=0$. Combining: $d⟨{\hat{M}}_{xy}^{+}⟩/dt=-i\gamma B_0⟨{\hat{M}}_{xy}^{+}⟩$, so $⟨{\hat{M}}_x+i{\hat{M}}_y⟩(t)=(⟨{\hat{M}}_x⟩(0)+i⟨{\hat{M}}_y⟩(0))e^{-i{\omega }_{0}t}$ — precession at the Larmor frequency ${\omega }_0=\gamma B_0$, with $⟨{\hat{M}}_z⟩$ constant. In vector form this is $d⟨\hat{\mathbf{M}}⟩/dt=\gamma ⟨\hat{\mathbf{M}}⟩\times {\mathbf{B}}_0$, which is the Bloch equation without relaxation. $\square$

Proposition 2 (k-space signal equals 2D Fourier transform of the image). Under linear gradients $\mathbf{G}(t)$, the detected transverse magnetisation $s(t)$ in the rotating frame equals

$$s(\mathbf{k})={\int }_{{\mathbb{R}}^2}\rho (\mathbf{r})e^{-i2\pi \mathbf{k}\cdot \mathbf{r}}{d}^2\mathbf{r},$$

where $\mathbf{k}(t)=(\gamma /2\pi ){\int }_0^t\mathbf{G}(t^{\prime })d{t}^{\prime }$ and $\rho (\mathbf{r})$ is the spatial proton-density distribution. The reconstructed image is the inverse Fourier transform $\rho (\mathbf{r})=\int s(\mathbf{k})e^{+i2\pi \mathbf{k}\cdot \mathbf{r}}{d}^2\mathbf{k}$.

Proof. The local Larmor frequency under the gradient is $\omega (\mathbf{r},t)=\gamma (B_0+\mathbf{G}(t)\cdot \mathbf{r})$. In the rotating frame at ${\omega }_0=\gamma B_0$, the local transverse magnetisation at position $\mathbf{r}$ evolves as $M_{xy}(\mathbf{r},t)=M_{xy}(\mathbf{r},0)e^{-i\gamma {\int }_0^t\mathbf{G}(t^{\prime })\cdot \mathbf{r}d{t}^{\prime }}{e}^{-t/T_2^{\ast }}$. The detected signal is the spatial integral over the excited slice: $s(t)=\int \rho (\mathbf{r})M_{xy}(\mathbf{r},0)e^{-i\gamma {\int }_0^t\mathbf{G}(t^{\prime })\cdot \mathbf{r}d{t}^{\prime }}{e}^{-t/T_2^{\ast }}{d}^2\mathbf{r}$. Substituting $\mathbf{k}(t)=(\gamma /2\pi ){\int }_0^t\mathbf{G}(t^{\prime })d{t}^{\prime }$ and absorbing the $T_2^{\ast }$ envelope into a known weighting function: $s(\mathbf{k})=\int \rho (\mathbf{r})e^{-i2\pi \mathbf{k}\cdot \mathbf{r}}{d}^2\mathbf{r}$. By the Fourier inversion theorem, $\rho (\mathbf{r})=\int s(\mathbf{k})e^{+i2\pi \mathbf{k}\cdot \mathbf{r}}{d}^2\mathbf{k}$, completing the proof. The k-space trajectory is controlled entirely by the gradient waveforms $\mathbf{G}(t)$, which is why different pulse sequences trace different k-space paths yet all reconstruct by Fourier inversion. $\square$

Proposition 3 (Stejskal-Tanner diffusion weighting). Under a bipolar diffusion-sensitising gradient pair of amplitude $G$ and duration $\delta$ separated by interval $\Delta$, the diffusion-weighted signal is $S/S_0=e^{-bD}$ with $b={\gamma }^2{G}^2{\delta }^2(\Delta -\delta /3)$ and $D$ the apparent diffusion coefficient.

Proof. A bipolar pair of gradient lobes cancels phase accumulation for any stationary spin but not for spins that diffuse between the lobes. For a spin at position ${\mathbf{r}}_1$ during the first lobe and ${\mathbf{r}}_2$ during the second, the net phase is $\varphi =\gamma G\delta ({\mathbf{r}}_2-{\mathbf{r}}_1)\cdot \hat{\mathbf{G}}$. For freely diffusing water, $⟨({\mathbf{r}}_2-{\mathbf{r}}_1{)}^2⟩=6D(\Delta -\delta /3)$ by the Einstein diffusion relation, and the ensemble average of $e^{i\varphi }$ over Gaussian-distributed displacements gives the attenuation $S/S_0=\exp (-⟨{\varphi }^2⟩/2)=\exp (-{\gamma }^2{G}^2{\delta }^{2}D(\Delta -\delta /3))=\exp (-bD)$. This is the Stejskal-Tanner 1965 equation, the basis of all diffusion MRI. In acute stroke, intracellular water has $D\approx 0.4\times 10^{-3}\mathrm{m}{\mathrm{m}}^2/s$ (restricted), versus $D\approx 3\times 10^{-3}\mathrm{m}{\mathrm{m}}^2/s$ for free CSF; at $b=1000s/m{\mathrm{m}}^2$ this gives $S/S_0\approx 0.67$ for stroke tissue versus $0.05$ for CSF, making stroke tissue appear bright on DWI. $\square$

Connections Master

• Medical diagnostics and imaging 35.09.01. This unit supplies the physics depth slice for the MRI entry in the chapter survey, building on the survey's account of imaging-modality selection across CT, MRI, ultrasound, PET, and plain radiography. The Bloch-equation mechanism derived here is the precise reason MRI excels at soft-tissue contrast while CT excels at bone and acute haemorrhage: MRI's signal depends on tissue-dependent proton relaxation parameters ($T_1$, $T_2$, $T_2^{\ast }$), whereas CT's signal depends only on electron density (linear attenuation coefficient). The survey's identification of pulse-sequence selection as the central clinical skill in MRI is given its physical content by the Bloch-equation analysis above.

• Neurodegenerative disease 35.03.05. MRI is the primary diagnostic imaging modality for Alzheimer's disease (hippocampal atrophy on T1; white-matter hyperintensities on FLAIR), Parkinson's disease (putaminal and substantia-nigra changes on susceptibility-weighted imaging and neuromelanin MRI), multiple sclerosis (demyelinating plaques on T2 and FLAIR), and the prion diseases (cortical ribboning on diffusion-weighted imaging). The relaxation-time dependence derived in the Bloch equations is the foundational reason each of these disease signatures is detectable on a specific pulse sequence. Conversely, the molecular-biology account of misfolded-protein propagation in 35.03.05 gives the pathological substrate that the MRI signal detects in vivo.

• Hubel-Wiesel visual cortex 29.03.04. The Hubel-Wiesel orientation and ocular-dominance columns described there are mapped in the living human brain primarily by BOLD fMRI, which relies on the Ogawa 1990 / Kwong 1992 contrast mechanism derived in the Advanced results above. The physiological hemodynamic coupling that links neural spiking to BOLD signal is the load-bearing assumption of every modern visual-cortex fMRI experiment, including retinotopic mapping, population receptive fields, and layer-dependent fMRI at 7 T. The connection from single-neuron electrophysiology (Hubel-Wiesel) to whole-brain human imaging (BOLD fMRI) is mediated entirely by the physics of the Bloch equation in the presence of paramagnetic deoxyhemoglobin.

• Neuroscience of consciousness 20.06.04. The Owen 2006 covert-awareness fMRI paradigm — in which a patient diagnosed as vegetative was instructed to imagine playing tennis and produced activation in motor-association cortex indistinguishable from healthy controls — relies entirely on the BOLD fMRI signal derived from the Bloch-equation analysis of T2-star relaxation. The disorders-of-consciousness literature has used fMRI-based covert-awareness detection as evidence that behavioural unresponsiveness can coexist with preserved higher cognition, and the underlying physics is exactly the gradient-encoded k-space reconstruction plus Ogawa-Kwong BOLD contrast detailed here.

Historical & philosophical context Master

Felix Bloch's 1946 paper "Nuclear Induction" (Physical Review 70:460) ^[Bloch1946] introduced the phenomenological equations of motion for the bulk magnetisation vector and verified them with the first observation of nuclear induction in a water sample. Edward Purcell, together with Torrey and Pound, independently observed nuclear magnetic resonance in condensed matter (Phys. Rev. 69:37, 1946) ^{[Purcell1946]}. Bloch and Purcell shared the 1952 Nobel Prize in Physics for these discoveries. The phenomenon remained a tool of physical chemistry for 25 years; in 1971 Raymond Damadian showed that rat tumour tissue had measurably longer T1 than normal tissue, raising the possibility of medical application.

The imaging revolution was launched by Paul Lauterbur's 1973 Nature paper ^{[Lauterbur1973]}, which demonstrated that a magnetic-field gradient encodes spatial position into the NMR signal and that back-projection reconstructs a two-dimensional image. Lauterbur's first published image was two capillary tubes of heavy water in ordinary water; his idea was initially rejected by journal reviewers as obvious, but the Nature editor published it anyway. Peter Mansfield's independent 1973 work (Mansfield-Grannell 1973, J. Phys. C 6) developed the complementary framework, and his 1977 echo-planar imaging paper ^{[Mansfield1977]} introduced single-shot acquisition, the technical enabler of functional MRI.

The functional MRI revolution came two decades later. Ogawa's 1990 PNAS paper ^[Ogawa1990] demonstrated that the MR signal in rat brain at high field depended on blood oxygenation — the BOLD contrast mechanism — and Kwong's 1992 PNAS paper ^[Kwong1992] produced the first human functional brain maps from this signal. Lauterbur and Mansfield shared the 2003 Nobel Prize in Physiology or Medicine; Ogawa and Kwong's extension of the Lauterbur-Mansfield imaging framework to functional brain mapping has driven the bulk of cognitive neuroscience for three decades since.

Bibliography Master

@article{Bloch1946,
  author = {Bloch, F.},
  title  = {Nuclear Induction},
  journal = {Physical Review},
  volume = {70},
  number = {7-8},
  pages  = {460--474},
  year   = {1946},
  doi    = {10.1103/PhysRev.70.460},
}

@article{Purcell1946,
  author = {Purcell, E. M. and Torrey, H. C. and Pound, R. V.},
  title  = {Resonance Absorption by Nuclear Magnetic Moments in a Solid},
  journal = {Physical Review},
  volume = {69},
  number = {1-2},
  pages  = {37--38},
  year   = {1946},
  doi    = {10.1103/PhysRev.69.37},
}

@article{Lauterbur1973,
  author = {Lauterbur, P. C.},
  title  = {Image Formation by Induced Local Interactions:
            Examples Employing Nuclear Magnetic Resonance},
  journal = {Nature},
  volume = {242},
  number = {5394},
  pages  = {190--191},
  year   = {1973},
  doi    = {10.1038/242190a0},
}

@article{Mansfield1977,
  author = {Mansfield, P.},
  title  = {Multi-planar image formation using {NMR} spin echoes},
  journal = {Journal of Physics C: Solid State Physics},
  volume = {10},
  number = {3},
  pages  = {L55--L58},
  year   = {1977},
  doi    = {10.1088/0022-3719/10/3/004},
}

@article{Edelstein1980,
  author = {Edelstein, W. A. and Hutchison, J. M. S. and Johnson, G. and
            Redpath, T.},
  title  = {Spin warp {NMR} imaging and applications to human whole-body
            imaging},
  journal = {Physics in Medicine and Biology},
  volume = {25},
  number = {4},
  pages  = {751--756},
  year   = {1980},
  doi    = {10.1088/0031-9155/25/4/010},
}

@article{LeBihan1986,
  author = {Le Bihan, D. and Breton, E. and Lallemand, D. and
            Aubin, M.-L. and Vignaud, J. and Laval-Jeantet, M.},
  title  = {Separation of diffusion and perfusion in intravoxel
            incoherent motion {MR} imaging},
  journal = {Radiology},
  volume = {168},
  number = {2},
  pages  = {497--505},
  year   = {1988},
  doi    = {10.1148/radiology.168.2.3393671},
}

@article{Ogawa1990,
  author = {Ogawa, S. and Lee, T.-M. and Kay, A. R. and Tank, D. W.},
  title  = {Brain magnetic resonance imaging with contrast dependent on
            blood oxygenation},
  journal = {Proceedings of the National Academy of Sciences USA},
  volume = {87},
  number = {24},
  pages  = {9868--9872},
  year   = {1990},
  doi    = {10.1073/pnas.87.24.9868},
}

@article{Kwong1992,
  author = {Kwong, K. K. and Belliveau, J. W. and Chesler, D. A. and
            Goldberg, I. E. and Weisskoff, R. M. and Poncelet, B. P. and
            Kennedy, D. N. and Hoppel, B. E. and Cohen, M. S. and
            Turner, R. and Cheng, H.-M. and Brady, T. J. and Rosen, B. R.},
  title  = {Dynamic magnetic resonance imaging of human brain activity
            during primary sensory stimulation},
  journal = {Proceedings of the National Academy of Sciences USA},
  volume = {89},
  number = {12},
  pages  = {5675--5679},
  year   = {1992},
  doi    = {10.1073/pnas.89.12.5675},
}

@article{Stejskal1965,
  author = {Stejskal, E. O. and Tanner, J. E.},
  title  = {Spin Diffusion Measurements: Spin Echoes in the Presence of a
            Time-Dependent Field Gradient},
  journal = {Journal of Chemical Physics},
  volume = {42},
  number = {1},
  pages  = {288--292},
  year   = {1965},
  doi    = {10.1063/1.1695690},
}

@article{Damadian1971,
  author = {Damadian, R.},
  title  = {Tumor detection by nuclear magnetic resonance},
  journal = {Science},
  volume = {171},
  number = {3966},
  pages  = {1151--1153},
  year   = {1971},
  doi    = {10.1126/science.171.3976.1151},
}