Skip to content
Imaging Evidence Path

Chapter 7 — Research methodology

This chapter · ~3 h total across 8 lessons · Reading and designing imaging studies · Energy: low to medium · Setup: one paper to dissect (Tran et al., PMID 35726103)

Critical appraisal is a guided learning activity, not an entrance exam. This chapter is a reading-and-design lens: a discipline you hold over any imaging-AI or radiomics study to find what is solid, what is overclaimed, and what a stronger study would look like. The same items tell you how not to make those mistakes in your own work.

The thread this chapter pulls

Chapter 4 taught the validation discipline you build; this chapter is the same discipline applied to reading. Every lesson is one question to ask of a paper — question, unit, reference standard, splits, metrics, claims — and the chapter ends by walking a real paper (Tran et al., PMID 35726103) through all of them. By the end, “this paper reports AUROC 0.94” reads to you as a partial statement pending the rest, and you can design the stronger study the authors should have run.

Lessons

  1. Read once for the story, then reconstruct the study — question, population, inputs, outcome, split, claim.
  2. What exactly is the unit of analysis? — patient/lesion/ scan/slice; repeated measurements.
  3. Where did the labels come from? — reference standard, circularity, reader consensus.
  4. Did the test set really leave the development world? — leakage; internal/external/temporal.
  5. Do the metrics support the clinical claim? — discrimination, calibration, utility, uncertainty.
  6. Recognise an overclaim — “biomarker”, “generalizable”, “clinically useful”.
  7. The reporting and risk-of-bias frameworks — CLAIM, TRIPOD+AI, PROBAST+AI, CLEAR, IBSI.
  8. Worked dissection: Tran et al. — a guided walkthrough of ACL-tear detection on knee MRI, and the stronger study you would design.

The full reading-and-design checklist and the frameworks-by-purpose table are in the methodology reference.

How this chapter connects

  • It is Chapter 4’s validation discipline applied to reading, and the lens for the deployment claims of Chapter 9.
  • Designing your own study and critiquing a paper are the same skill in opposite directions (Ch. 4, lesson 8).

If you only have an hour, lessons 4–5 (test set and metrics) plus the Tran et al. dissection (lesson 8) are the core that transfers to every paper you will read.