Which Test, When
- "Best test" doesn't exist in a vacuum — the right test is the one that answers this clinical question for this patient with the least harm.
- Start with the question, not the scanner. "Is the appendix inflamed?" points you somewhere very different from "is there a stroke?"
- Match the modality to the tissue: bone and acute blood love CT, soft tissue and the nervous system love MRI, and ultrasound shines on babies, bellies, and anything you'd rather not irradiate.
- The big tie-breakers are radiation, speed, cost, availability, and whether the patient can even hold still or fit in the machine.
- When you genuinely don't know, the ACR Appropriateness Criteria already did the homework — use them.
Ordering imaging feels like standing in front of a vending machine with forty buttons and a vague sense of hunger. CT? MRI? Ultrasound? The plain old X-ray that costs about as much as a sandwich? Here's the secret the machine won't tell you: the right answer almost never starts with the machine. It starts with a question.
Ask the question first, pick the tool second
Every good imaging order is really a yes/no question in disguise. "Does this patient have a kidney stone?" "Is there free air in the belly?" "What is this lump?" Once the question is sharp, the test usually picks itself, because each modality is good at a specific kind of looking.
Think of it like choosing a tool from a drawer. You don't reach for the same thing to hang a picture and to fix a faucet. A hammer is fantastic — until the problem is a leaky pipe, at which point it is gloriously useless and slightly dangerous.
The order isn't "patient needs imaging → what's available." It's "what exactly am I trying to find or exclude?" → then the modality.
What each modality is secretly best at
Here's the cheat sheet I wish someone had taped to my forehead on day one. (Each of these has its own deeper page; I'm just sketching the personalities.)
| Modality | Loves | Hates | Radiation? |
|---|---|---|---|
| Radiography (X-ray) | Bones, lungs, free air, line/tube checks — fast and cheap | Subtle soft tissue, overlapping structures | A little |
| CT | Acute bleeds, trauma, stones, "scan everything fast" | Soft-tissue subtleties MRI sees better | Yes — meaningfully |
| MRI | Brain, spine, soft tissue, joints, exquisite contrast | Speed, motion, metal, claustrophobia | None |
| Ultrasound | Gallbladder, pregnancy, neck, kids, anything superficial | Bone, bowel gas, lungs (sound hates air) | None |
| Nuclear medicine / PET | Function and metabolism — is it active, not just present | Crisp anatomy | Yes |
The one-line version: CT is fast and sees blood and bone; MRI is slow and sees soft tissue; ultrasound is free of radiation and free of patience; X-ray is the quick first look.
The tie-breakers that decide real life
Two tests can both answer the question. Now what? You weigh the costs — and "cost" means more than dollars.
- Radiation. A CT carries real dose; an ultrasound or MRI carries none. This matters enormously in kids and in pregnancy. The whole point of ALARA is to not zap people more than the question requires.
- Speed and the sick patient. A crashing trauma patient can't spend 40 minutes holding still in an MRI. CT takes seconds. Sometimes the fast enough test beats the theoretically best test.
- Contrast. Some questions need it, some don't — and contrast has its own risks. Whether to give it is a real decision, covered in contrast vs non-contrast.
- Availability and the human in the machine. The best scanner is the one you can actually get tonight. And a patient with a pacemaker, severe claustrophobia, or who simply won't fit may rule out MRI before you even start.
"More imaging" is not "better care." Every scan you don't need is a dose you didn't give, an incidental finding you didn't have to chase, and a worried patient you didn't create. Sometimes the right answer is no scan at all.
A few worked examples
Patterns stick better than rules, so here's how the logic plays out:
| Clinical question | Reasonable first test | Why |
|---|---|---|
| Suspected kidney stone, flank pain | Non-contrast CT (or US in pregnancy) | Stones are dense; CT finds them fast |
| Right upper quadrant pain, ?gallstones | Ultrasound | Cheap, no radiation, great for the gallbladder |
| Sudden severe headache, ?bleed | Non-contrast head CT | Acute blood is bright and obvious on CT |
| Persistent low back pain with red flags | MRI | Cord and discs are soft tissue — MRI's home turf |
| First-trimester pain or bleeding | Ultrasound | No radiation near a pregnancy; great pelvic detail |
Notice the through-line: the question drives the choice every time. None of this is about which machine is fanciest.
When you genuinely don't know
You will not memorize the right test for every scenario, and you don't have to. The ACR Appropriateness Criteria are exactly this — expert-reviewed ratings of which study fits which clinical scenario, updated over time. When the choice is murky, that's your move: look it up rather than guess.
Don't confuse "the test came back normal" with "I picked the right test." A normal study that couldn't have shown the problem in the first place is falsely reassuring. Matching the test to the question is also about its sensitivity for that specific disease — the wrong tool can miss the thing it was never able to see.
So: start with the question. Pick the modality whose superpower matches the tissue and the urgency. Weigh radiation, speed, and the patient in front of you. And when in doubt, let the people who built the evidence tables make the call. The vending machine is a lot less intimidating once you remember you're shopping for an answer, not a picture.