Case-Based Interpretation

From Physics to the Final Read

Every preceding lesson built a single instrument: an equation, a sequence, an artifact, a contrast mechanism. The radiologist plays the whole orchestra at once. A study lands on the workstation as a stack of co-registered volumes, each acquired with deliberately different timing, and the reader must convert that stack into a single sentence of impression. This capstone treats interpretation as an applied physics problem. The pixel intensity in front of you is the spin-echo or gradient-echo signal evaluated for a particular tissue at a particular TR, TE, TI and flip angle. If you know the physics, you can run the inference in reverse: from a pattern of signal across sequences back to the tissue that produced it.

We will assemble three things. First, a disciplined search pattern that guarantees every sequence is reviewed before any conclusion is drawn. Second, a cross-sequence signal grammar that maps the behavior of fat, fluid, blood at each stage of degradation, calcium, flowing spins, melanin and gadolinium onto a unique fingerprint. Third, a worked narrative on the live multi-contrast dataset in the viewer, ending in a synthesized impression and a decision about whether to recommend more imaging.

A Sequence-Aware Search Pattern

Misses in MRI are rarely failures of knowledge; they are failures of coverage. The single most reliable safeguard is the rule that no sequence is skipped and no conclusion is reached until every sequence has been seen. A lesion can be invisible on three sequences and obvious on the fourth, and the human eye anchors on the first abnormality it finds, suppressing the search for a second. A fixed routine defeats this.

1. Orient and check technique. Confirm the patient, the region, the plane, and that each labeled series matches its intended weighting. A mislabeled or motion-corrupted series invalidates downstream reasoning.
2. Survey each volume top to bottom in its own right. Scroll the entire T1 stack, then the entire T2 stack, then PD, then the angiographic data. Do not flip between sequences at a single slice yet; first build a mental model of each contrast alone.
3. Cross-reference at the level of any finding. Only after the independent surveys, return to a suspicious slice and compare the same anatomy across all sequences to read its signal fingerprint.
4. Apply an anatomic checklist. For brain: parenchyma, gray-white junction, ventricles and CSF spaces, vessels, sella and orbits, sinuses and mastoids, calvarium and scalp. The checklist forces you past the obvious.
5. Reconcile and synthesize. State what is abnormal, characterize it by tissue type, assign a differential ordered by signal logic, and decide whether the protocol is sufficient.

The Cross-Sequence Signal Grammar

Tissue characterization in MRI is pattern recognition over a vector of intensities. A voxel does not give you one number; it gives you one number per sequence. The diagnostic art is reading that vector. The underlying physics is the contrast equations you already know: longitudinal recovery governed by $T_1$, transverse decay governed by $T_2$, and modulation by proton density $\rho$. For a conventional spin echo the signal is

Read Eq. 18.1 as the engine behind the table below. Short $T_1$ tissues (fat, methemoglobin, gadolinium-shortened water, melanin, proteinaceous fluid) recover quickly and are bright when TR is short. Long $T_2$ tissues (simple fluid, edema) retain transverse signal and are bright when TE is long. Anything that destroys local phase coherence, such as the strong susceptibility of deoxyhemoglobin, hemosiderin or calcium, or rapidly moving spins washing out of the slice, drives signal toward zero, especially on gradient-echo and long-TE images.

| Tissue or substance | T1-weighted | T2-weighted | PD-weighted | Physical basis | | --- | --- | --- | --- | --- | | Fat | Bright | Intermediate to bright | Bright | Very short T1 (~250 to 300 ms at 1.5 T); fast-spin-echo keeps fat bright on T2 | | Simple fluid / CSF | Dark | Bright | Intermediate | Long T1 and very long T2 | | Proteinaceous / mucoid fluid | Bright | Variable | Bright | Macromolecules shorten T1 toward the fast-recovery regime | | Acute hemorrhage (deoxyhemoglobin) | Iso to dark | Dark | Iso | Paramagnetic, intracellular; strong T2 and T2* shortening | | Subacute hemorrhage (methemoglobin) | Bright | Dark then bright | Bright | Methemoglobin markedly shortens T1; T2 lengthens once cells lyse | | Chronic hemorrhage (hemosiderin) | Dark | Very dark | Dark | Superparamagnetic iron; profound T2* loss, bloom on gradient echo | | Calcium | Variable (often dark) | Usually dark | Dark | Low mobile proton density; susceptibility loss; rarely T1-bright if crystalline-bound | | Flowing blood (patent vessel) | Flow void (dark) on SE | Flow void on SE | Flow void | Spins leave the slice before refocusing; bright on flow-sensitive GRE / TOF | | Melanin | Bright | Dark | Variable | Stable free radicals are paramagnetic, shortening both T1 and T2 | | Gadolinium-enhancing tissue | Bright | Variable | Iso to bright | Dipolar relaxation shortens T1 of adjacent water protons |

Flow, Susceptibility and the Limits of a Single Contrast

Two effects deserve special attention because they let you read physiology, not just tissue. The first is flow. On spin echo, blood moving fast enough to leave the slice between the excitation and refocusing pulses never sees both pulses and produces no echo, appearing as a black flow void. This is a feature: a patent intracranial vessel shows a crisp flow void, and its loss can signal thrombosis or slow flow. Time-of-flight (TOF) MR angiography inverts the logic, suppressing stationary tissue so that unsaturated inflowing spins are the only bright signal, which is the basis of the MRA volume in this lesson.

The second is susceptibility. Local field inhomogeneity from iron, calcium or air accelerates dephasing. The reversible part is captured by the difference between $T_2$ and $T_2^{*}$:

Spin echo refocuses the static part of $\Delta B_{0,\mathrm{local}}$ with its 180 degree pulse, so $T_2$ rather than $T_2^{*}$ governs signal; gradient echo does not, so it blooms on iron and calcium. This is exactly why a chronic microbleed can be invisible on T1, T2 and PD yet starkly black on a susceptibility-weighted gradient-echo image, and why the decision to recommend an added sequence is itself a physics judgment.

Worked Capstone: Reading the Live Multi-Contrast Set

Now apply the routine to real data. The viewer below loads four co-registered volumes from a single subject: T1-weighted, T2-weighted, proton-density, and a flow-sensitive MR angiographic volume. Follow the search pattern exactly. First scroll each volume independently and build its model; only then cross-reference. As you navigate, narrate the signal of CSF, white matter, gray matter, scalp fat and the major vessels, and confirm each against Eq. 18.1 and the grammar table.

Here is the interpretive narrative. On the T1 volume the gray-white differentiation is sharpest because the $T_1$ values of gray and white matter differ substantially; subcutaneous and orbital fat are bright, and the ventricular CSF is dark, exactly the dark-fluid, bright-fat signature the table predicts. Switch to the T2 volume: CSF now inverts to brilliant white as its very long $T_2$ resists decay at long TE, periventricular and cortical fluid spaces are best displayed, and gray matter is slightly brighter than white. The PD volume sits between the two, with the least contrast but often the highest raw SNR because it minimizes both the recovery and decay penalties of Eq. 18.1. Finally, the MRA volume suppresses stationary brain so that inflowing arterial spins dominate, letting you trace the circle of Willis and its major branches; absence of expected flow signal in a vessel would be the finding to chase.

Synthesizing across the four: the cross-sequence signals of every normal tissue are internally consistent. CSF goes dark to bright to intermediate (T1, T2, PD); fat stays bright on T1 and PD; vessels show flow voids on the conventional sequences yet light up on the flow-sensitive MRA. That coherence is itself a finding, it tells you the study is technically sound and free of a tissue with a paradoxical fingerprint. Were a focus to read bright on T1 and dark on T2, you would immediately invoke the blood-age grammar and stage it as early subacute methemoglobin rather than calling it simply a bright spot.

When Artifact Mimics or Masks Disease

The same physics that creates contrast creates artifacts, and a reader who cannot tell them apart will either invent disease or overlook it. Three deserve a standing place in the search pattern. Chemical shift displaces fat relative to water by a fixed number of pixels along the frequency-encode axis, producing bright and dark rims at fat-water interfaces such as the orbit or kidney; recognizing it prevents calling a pseudo-lesion. Motion and pulsation propagate as ghosts along the phase-encode direction, and CSF or vascular pulsation can plant a ghost squarely over normal parenchyma, mimicking a mass or masking a real one. Susceptibility and metal create signal voids with peripheral distortion that can swallow adjacent anatomy entirely on gradient echo while sparing it on spin echo.

Closing the Read and Recommending More

A capstone read ends with a decision, not a description. Three questions close every interpretation. Is the protocol adequate to answer the clinical question? Does any finding have an ambiguous fingerprint that a targeted sequence would resolve? Is contrast needed, given that gadolinium chiefly shortens $T_1$ and must be assessed on T1-weighted images? The recommendation to add fat saturation, a gradient-echo or susceptibility sequence, diffusion, or post-gadolinium T1 is the point where physics, protocol and anatomy converge, and it is what distinguishes a competent description from a useful consultation.