MRI from Zero

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

Published

October 25, 2026

Introduction

English version — machine-assisted translation (by generative AI, Anthropic Claude) of the original Japanese hands-on. 日本語版はこちら.

ImportantThis material was created by AI

This material was created by generative AI (Claude), and the bibliographic details of the cited references (authors, journal, volume, pages, DOI) were checked automatically against Crossref. It has not been reviewed by an expert. Many of the numbers are “typical values / rough guides” and vary with the scanner, imaging parameters, and source. If you find an error, please let the author know.

MRI is a machine that “sends radio waves to atomic nuclei inside the body while it sits in a strong magnet, and then listens to the radio waves that come back.” This material is designed so that even people who are not confident with physics or equations can understand the principles of MRI by trying things out hands-on.

  • What exactly are T1 and T2?
  • Why is fat bright on T1-weighted images and water bright on T2-weighted images?
  • How are fat and muscle separated? (fat suppression, the Dixon method)
  • How does 23Na (sodium) or 31P (phosphorus) MRI differ from ordinary MRI?
  • What is the difference between NMR, MRS, and MRI?
  • What software and Python code are used in real analyses?

We will explain these questions one by one.

Who this is for

  • Researchers, graduate students, and engineers who use (or plan to use) MRI data
  • People in fields such as body composition, muscle, exercise, and nutrition who want to understand the MRI terms that appear in papers
  • High-school physics (magnets, waves, frequency) is all the background you need

How to use this material

Demos that run in the browser

When you move a slider, the signal or graph is recalculated on the spot (Observable JS). No installation needed.

Python code

The code in each chapter runs as-is when copied (numpy / scipy / matplotlib). The figures in the text were made by running this very code.

Links to real papers

In the 23Na chapter, we explain what the measures used in skeletal muscle papers (Gerhalter et al. 2017; Höger et al. 2023; Marton et al. 2026) (TSC, ICwS, TQF, etc.) actually measure.

Chapter overview

Chapter Content Approx. time
1 Nuclei are tiny magnets: spin, precession, Larmor frequency 15 min
2 The NMR signal: RF pulses, FID, T1 and T2 20 min
3 Making an image: gradients, k-space, Fourier transform 20 min
4 Contrast: TR, TE, and T1/T2 weighting 20 min
5 Separating fat and muscle: fat suppression, the Dixon method, fat fraction 20 min
6 Multinuclear MRI: 23Na and 31P 25 min
7 How NMR, MRS, and MRI differ 10 min
8 Quantitative analysis, software, and Python 25 min
9 Review quiz 15 min
NoteSlides

Summary slides for the lecture are here (press F for full screen, Esc for overview).

The big picture

flowchart LR
  A["Strong static field B0<br/>(1.5T / 3T / 7T)"] --> B["Nuclear spins align<br/>and precess (Ch. 1)"]
  B --> C["Tip them with an RF pulse<br/>→ signal (FID) (Ch. 2)"]
  C --> D["T1 recovery, T2 decay<br/>= differ by tissue (Ch. 2)"]
  D --> E["Gradients encode position<br/>→ k-space → image (Ch. 3)"]
  E --> F["Choose contrast<br/>with TR and TE (Ch. 4)"]
  F --> G["Separate fat and water (Ch. 5)"]
  C --> H["Nuclei other than 1H<br/>23Na, 31P (Ch. 6)"]

Setting up Python

The code up to Chapter 8 runs with only the following libraries.

pip install numpy scipy matplotlib

In the second half of Chapter 8, where we read real data (DICOM, NIfTI), we additionally use pydicom and nibabel.

pip install pydicom nibabel