MRI Physics: Scanner Setup and Tuning

1) Safety checks; sample in the coil, coil in the magnet.
2) Tune & match the loaded coil.
3) Find the water frequency (FID → spectrum).
4) Calibrate RF power (which power gives 90°).
5) Localizer images.
6) Shim the region of interest, then re-centre the frequency.
7) Set receiver gain.
8) Run the protocol and check a first image.

Each step needs the previous one to be right.
Coil coupling must work before any MR signal is reliable; the pulse must be on resonance before you can calibrate its flip angle; the localizer tells you where to shim; shimming can shift the frequency, so you re-centre after. (Practical 1 did localizer/shim before the RF calibration; automatic routines also reorder, but the dependencies stay the same.)

Paramagnetic Cu2+ ions shorten water's T1 (and T2), so Mz recovers quickly and scans can be repeated fast without saturation.
Pure water has a T1 of seconds, which would force very long waits between repetitions.

Its B1 is uniform over the whole cylinder, so every region gets the same flip angle and similar sensitivity.
A surface coil would give strong signal only near the loop and fading signal deeper in.

Most likely a proton-free fluorinated fluid (e.g., Fomblin): it gives no 1H signal (black background), keeps the tissue from drying, and reduces the susceptibility jump an air boundary would cause.
Ordinary oil contains hydrogen and would show up.

First, with the sample already in the coil in its final position, before any MR measurement.
A tuned and matched coil turns RF power into B1 efficiently (correct flip angles) and passes the weak MR voltage to the receiver (good SNR).

x-axis: frequency around the target (≈ 600 MHz for 1H at 14.1 T); y-axis: reflected power.
Goal: a deep dip exactly at the target frequency. Practical criterion: < ~10% reflected (0% is ideal but hard to hold).

Tuning. It changes a capacitor, which shifts the coil's resonance f_coil = 1/(2π√(LC)) and slides the dip sideways onto the target.

Matching. It brings the coil's impedance to 50 Ω so less power bounces back, which deepens the dip.
Then re-check tuning: the two knobs interact, so you iterate.

Sample loading: the sample's size, conductivity and position change the coil's effective L, C and losses, so its resonance and impedance shift.
Re-tune after repositioning or changing samples. A disconnected cable gives total reflection: the dip disappears.

With gradients off: pulse, record the FID, Fourier-transform it, find the water peak, and set f_RF and f_ref onto it.
(γ/2π)·B0 is only the starting guess; drift and the local field move the true resonance slightly.

Raise f_ref by 1.5 kHz, so Δf = f_L − f_ref → 0.
The peak moves to 0 and the FID stops oscillating (smooth decay). The width stays the same: centring is bookkeeping, not a field correction. (Peak 30 Hz below zero → lower f_ref by 30 Hz.)

Keep power fixed and step the pulse duration (or fix the duration and step the power); record the initial FID amplitude each time.
First maximum = 90°; back to ≈ 0 at 180°; maximum again at 270°. Signal ∝ |sin θ|. Why: the console must know what setting gives the flip angle you ask for in this coil + sample.

Recovery time (full Mz each time: long TR or a short-T1 sample), receiver gain, and the delay before the amplitude is read.
Otherwise a change in amplitude could come from saturation or scaling instead of the flip angle.

120 µs at 10 W, or 40 W at 60 µs.
θ ∝ duration, so double the time; or θ ∝ B1 ∝ √P, so double B1 = 4× power. (Same coil, sample and pulse shape.) The practical sweep started at ~10 µs and found the first maximum near 60 µs.

B1 is not perfectly uniform (and some spins are off resonance), so parts of the sample get more or less than 180° and keep some Mxy.
For the same reason the calibration minimum is shallow and the maxima are less sharp: the total signal mixes regions with different flip angles.

The power that gives a 90° flip with a standard reference pulse (here noted as 1 ms) in this coil + sample. Other pulses are scaled from it by angle, duration and shape.
Lower reference power = more efficient transmit for that set-up. For a transmiteceive coil that usually also means better receive sensitivity (reciprocity), but it is a transmit calibration, not an SNR measurement. If shown as attenuation in dB, more dB = less power.

Quick, low-resolution GRE images in three orthogonal planes to check the sample is centred in the magnet and coil, and to plan slices, voxels and the shim region.
It is for planning; its short TR/TE/small flip are chosen for speed, not for a particular contrast.

Where the three planes intersect, the same spins are excited by more than one plane's pulses, so they have less time to recover → less Mz → darker bands.
It is saturation, not missing sample or bad shimming. Stronger with long T1 or short spacing between pulses; fades when T1 is short.

1) B0 inhomogeneity → extra dephasing (GRE especially; improves after shimming).
2) B1 non-uniformity in transmit or receive.
3) Air bubbles: no water plus a susceptibility jump, so the dark area is bigger than the bubble.
4) Saturation at slice intersections.

After the localizer has defined the region, with gradients off. Adjust shim currents until the water peak is narrow and smooth and the FID lasts longer.
Physics: a more uniform B0 → more similar Larmor frequencies → slower dephasing (longer T2*). Re-centre the frequency afterwards.

Linear (x, y, z) first, then second order (z^2, xy, …); go back and forth because the patterns overlap.
Judge by linewidth (FWHM) and FID persistence, not by peak height, which also depends on gain and flip angle.

Narrow but off-centre → move f_ref (frequency centring). Broad but centred → shim (the field is non-uniform).
One fixes the electronic reference, the other fixes the actual field.

B0 map: where the field/frequency is off → guides shimming. B1 map: where the achieved flip angle is off → diagnoses RF coil non-uniformity.
Uneven brightness can come from either, so they are not interchangeable checks.

It amplifies the received voltage so the strongest signal fills the ADC range without clipping.
Too high → clipping; too low → digitization noise matters. It does not change B1, the flip angle or the spins, and cannot raise the sample's intrinsic SNR.

1) Tuning → coil resonance (dip position).
2) Matching → coil impedance / power transfer (dip depth).
3) Frequency centring → electronic reference (peak position; spins unchanged).
4) Shimming → the actual B0 field (peak width, FID length).
5) RF power → B1 amplitude (flip angle).
6) Receiver gain → amplification (signal scale, clipping).
Trap: tuning moves the coil's resonance, not the protons' frequency; centring moves the peak, not its width.

Across the diameter: a rounded profile, because central slabs cut through a wider chord of the circle → more water. Along the length: nearly rectangular, because every slab has the same circular cross-section.
The concentration is uniform; the shape shows how much water sits at each position.

D = Δf / ((γ/2π)·G) = 8516 / (42.58 × 10^6 × 0.020) = 0.010 m = 10 mm.
Across the tube this is the diameter; along it, the length.

Stronger G → the profile spans more Hz, but converting back with the larger G gives the same size: the object did not change.
Reversed polarity → mirror image (a symmetric cylinder looks the same).

Both fall with TE, but GRE falls faster: it decays with T2* (field spread not refocused), SE with T2 (static spread refocused by the 180°).
Fit exponentials to the echo-centre amplitudes; T2* ≤ T2, and better shimming brings GRE closer to SE.

The signal rising as the spins rephase after the 180°, peaking at TE, then falling again as the static field differences dephase them once more.
So one echo's shape shows T2*-type dephasing on both sides, while echo-centre heights across different TEs give T2. Recording could only start after the dead time (coil ring-down, T/R switching).

Little T2 contrast (short TE) and little T1 contrast (long TR) → mostly proton-density weighting.
Ex vivo or fixed tissue has different T1/T2 from living tissue, so the familiar in vivo brightness pattern may not appear: "short" and "long" are relative to the sample.

Frequency, shim and reference power are reused while the set-up is unchanged; redo them after a new sample or coil, repositioning, a new target region or drift. Receiver gain and method-specific adjustments run for each scan.
Before committing to a long scan, check a short one for wrap, clipping, noise and dark bands, plus time and SARradient limits.