Open teaching series · free to read, free to teach from

Reading the scan before the read

Five parts on spine MRI for advanced practice providers, residents and students — what to look at, in what order, what the published grading systems actually say, and what routinely fools people. Every claim is checked against the primary source, and every figure is credited in its caption.

Written as the companion handout to a talk given to advanced practice providers at the Michigan Association of Neurosurgeons summer meeting. Read in order if the material is new to you; each part also stands on its own.

  1. Timelines

    How cord compression behaves over time

    Four timelines built from published cohorts: how long a compressed cord takes to become myelopathic, when the damage stops being reversible, what recovery actually looks like after surgery, and how often people should be seen in between. This is the context everything else sits inside.

    Start here →
  2. Part 1

    Reading the compressed cord

    Which sequence is which, an eight-step read you can run every time, and the five findings that change management — then three artefacts that look convincing on one image and evaporate on a second.

    Read part 1 →
  3. Part 2

    Lumbar central stenosis

    Grading the dural sac by what is inside it rather than by measuring it. The Schizas system in full, the evidence that morphology beats cross-sectional area in both directions, how the grade tracks what happened to the patient, and a short drill.

    Read part 2 →
  4. Part 3

    Nerve root and foramen

    Two zones, two grading systems, two sequences. Pfirrmann for the root in the lateral recess, Lee for the root in the foramen — and why the second is graded on T1 when everything before it was graded on T2.

    Read part 3 →
  5. Part 4

    Anatomic checkpoints

    The two things every earlier part quietly assumed: that the cord ends where you expect, and that the level you are naming is the level you are looking at. Conus position, transitional anatomy, and what follows from roughly one spine in nine not being built to the standard plan.

    Read part 4 →

Why this sits alongside the datasets. Part 4 ends where our data work begins. Transitional lumbosacral anatomy is present in a few percent of people and other numerical variants in several more, and both silently renumber every level below them. That is a teaching problem for a clinician and a labelling problem for a segmentation model — and it is the reason CTSpinoPelvic1K was built to be LSTV-aware.

On the education model itself. The open framework behind this work is described in Schehr A, Kim J, Schwing G, OpenSpineConsortium: an open-source framework for medical student engagement in computational spine imaging research, Cureus 2026;18(7):e112661.