Resolution of MRI is three different jobs that share one word: spatial (how small a structure stays two structures), temporal (how short a time you can sample), contrast (how two tissues stay two greys). It is not how to read a head MRI. It is not a cine loop. It is not CT resolution, and it is not a Hounsfield unit.
If you meant sequences, a search pattern, stroke / mass / MS / blood → how to read head MRI scans. If you meant a movie of a heartbeat → cine MRI. If you meant the same three pillars on CT → resolution of a CT scan. If you meant HU / window → Hounsfield units.
427 already points here for “Tesla / spatial-temporal-contrast.” This page is that physics. PYCAD builds custom web DICOM viewers and imaging models. It does not ship an MRI scanner or a “resolution optimizer.”
Three words, three knobs
| Word | What it is | What sets it | What it is not |
|---|---|---|---|
| Spatial | Smallest distance at which two points stay two points | In-plane: FOV ÷ matrix. Through-plane: slice thickness (+ gap) | Tesla. A 3 T magnet is not a smaller voxel by itself |
| Temporal | Time between samples of the same plane | TR, views-per-segment, whether you gate | A camcorder. Cardiac cine is rebuilt across beats — cine MRI |
| Contrast | Two tissues stay two greys | The sequence (T1 / T2 / FLAIR / DWI / bSSFP), not the matrix | A read. How those greys become a report is 427 |
You trade them. A smaller voxel is less signal in that voxel. You pay with more averages, a longer hold, more noise, or a thicker slice. A protocol that is “high-res” on a still brain is the wrong protocol on a heart.
Tesla is SNR, not a pixel
Field strength (1.5 T, 3 T, 7 T) is how strongly the protons line up. More field → more signal → you can afford a smaller voxel or a shorter scan. It does not rewrite the FOV/matrix formula.
- 1.5 T. The workhorse. Fewer susceptibility artifacts around air / metal. Fine for most body work and many brains.
- 3 T. Roughly twice the field, more SNR. The usual ask when the question is a small nerve, a tiny pituitary, or a tighter matrix on the same hold. Also more dielectric shading and more metal artifact.
- 7 T and up. Research and a few named clinical niches. Not a department default. Safety, implants, and banding are the reason.
A 3 T scan with a coarse matrix is still a coarse scan. A 1.5 T scan with a tight matrix and enough averages can be the sharper one. Read the voxel, not the magnet brochure.
The spatial formula
In-plane voxel size ≈ field of view ÷ matrix size, per axis. A 240 mm FOV on a 256 matrix is about 0.94 mm. The same FOV on 512 is about 0.47 mm and needs more signal or more time. Slice thickness is the third edge of the voxel. A 0.5 mm in-plane grid with a 5 mm slice is a pancake, not a cube.
Partial-volume is what that pancake does to a thin structure: the voxel averages the thing you care about with whatever else is in the slice. That is a spatial-resolution problem, not a contrast-resolution problem.
Interpolation in the viewer is not acquisition. A 256 matrix blown up on a 4K monitor is still 256. Do not call zoom “we scanned at higher resolution.”
Sequences change contrast, not the grid
Spin-echo vs gradient-echo, T1 vs T2 vs FLAIR vs DWI — those pick which tissue is bright. They do not change FOV/matrix. A FLAIR that makes a plaque visible is contrast resolution doing its job. How you then read the set is 427.
Parallel imaging and compressed sensing buy time. They can make a tighter matrix fit a breath-hold. They also fold artifact when the coil map is wrong. AI reconstruction is the same deal: it can denoise a small-voxel scan; it is not a new Tesla rating. None of that is a PYCAD recon product.
What this page is not
- Not CT. Detector pitch, reconstruction kernel, and HU live on resolution of a CT scan and Hounsfield units.
- Not cine as a product. Temporal resolution of a cardiac loop is cine MRI.
- Not a search pattern. 427 already said so.
- Not a market CAGR, a Tesla buyer’s guide, or a patient-prep FAQ. Dropped.
If the missing piece is a web viewer that can hang the sequences and honour the voxel size in the tags, that is the imaging piece. Case studies.