Read slides as a page
H01 electron microscopy with object segmentation and original 2 µm scale bar Nanoscale · Module 09

Neuron Morphology and Skeletonization

Teaching Deck

Human cortex · H01
Object segmentation over electron microscopy

H01 release · Lichtman Lab / Harvard & Connectomics at Google · CC BY 4.0
Shapson-Coe et al. (2024) · doi:10.1126/science.adk4858

Module 09 · Nanoscale

Learning Objectives

  • Generate skeleton representations from reconstructed neurites
  • Compute core morphology descriptors
  • Relate morphology metrics to biological interpretation
  • Report morphology uncertainty and classification limits
Module 09 · Nanoscale

Capability Target

Produce a skeleton-based morphology summary with at least three descriptors and one explicit limitation.

Module 09 · Nanoscale

Concept Focus

1) What is skeletonization and why do we need it?

A segmented neuron occupies millions of voxels in the EM volume. To analyze its morphology efficiently, we reduce it to a skeleton: a tree graph where nodes represent points along the neurite centerline and edges represent the path between them. Skeletons compress a neuron's 3D structure from gigabytes to kilobytes while preserving topology (branching pattern, path lengths, connectivity) — at the cost of surface geometry and spine shape, which are discarded.

Module 09 · Nanoscale

Core Workflow

  • Build skeleton from volumetric segmentation using TEASAR or equivalent algorithm.
  • Quality-check the skeleton: prune spurious branches, verify branch points, check for disconnected fragments.
  • Compute descriptors: cable length, branch points, Strahler number, Sholl profile, spine density.
  • Compare against reference patterns: does this neuron match the expected morphology for its putative cell type?
Module 09 · Nanoscale

Core Workflow (continued)

  • Report interpretation confidence: which descriptors are robust, which are affected by reconstruction quality?
Module 09 · Nanoscale

Run of Show (60 min)

  • 00:00-10:00 | Morphology overview
  • 10:00-24:00 | Skeleton extraction demo
  • 24:00-38:00 | Descriptor calculation
  • 38:00-50:00 | Interpretation and caveats
Module 09 · Nanoscale

Run of Show (60 min) (continued)

  • 50:00-60:00 | Competency check
Module 09 · Nanoscale

Misconceptions to Watch

  • Misconception guardrail: a skeleton is a lossless summary of a neuron.
  • Misconception guardrail: morphological measurements are comparable across cells that were proofread to different levels.
  • Misconception guardrail: total dendritic length is a property of the neuron itself, independent of how it was reconstructed.
  • Misconception guardrail: a cell type assigned from morphology alone needs no corroboration from connectivity or molecular identity.
Module 09 · Nanoscale

A skeleton makes measurement choices explicit

Synthetic planar tree. Coordinates are in micrometres.

A (0, 0)
Root

A–B: 3 µm

B (0, 3)
Branch point

C (−4, 3)
B–C: 4 µm

D (4, 3)
B–D: 4 µm

Discuss: Compute total cable length and both root-to-tip path lengths.

Synthetic fixture; exact geometry checked by scripts/check_visual_design.rb.

Module 09 · Nanoscale

A missing branch changes the story

Use the synthetic tree on the previous slide.

  1. Complete example

    Three edges: total length 11 µm.

  2. Truncated example

    Remove B–D: measured length is 7 µm.

  3. Report the limit

    The lower value describes the observed reconstruction.

Discuss: Does a lower measurement prove a smaller biological arbor?

Illustrative teaching scenario; see this module’s lesson and worked example.

Module 09 · Nanoscale

Studio Activity

Scenario: You have skeletons for 10 neurons in L2/3 of mouse visual cortex. Your task is to classify them as pyramidal vs interneuron based on morphology alone, then validate against synapse-based classification (excitatory vs inhibitory output synapses).

Module 09 · Nanoscale

Activity Output Checklist

  • Evidence-linked artifact submitted.
  • At least one limitation or uncertainty stated.
  • Revision point captured from feedback.
Module 09 · Nanoscale

Assessment Rubric

Minimum pass

  • Valid skeleton and descriptor set for all 10 neurons, with at least 3 descriptors each.
  • Every classification carries a stated evidence chain, not just a label.
  • At least one explicit measurement limitation named, tied to a specific reconstruction issue.
Module 09 · Nanoscale

Assessment Rubric

Strong performance

  • Descriptors partitioned into robust versus reconstruction-sensitive for each borderline cell.
  • Mismatches between morphological and synapse-based calls traced to a root cause (truncation, split, spurious branches) rather than logged as disagreement.
  • Per-cell completeness estimated and reported next to every absolute measurement.
Module 09 · Nanoscale

Assessment Rubric (continued)

  • Ratios computed on contained compartments wherever truncation is present.
Module 09 · Nanoscale

Assessment Rubric

Common failure to flag

  • Descriptor list without biological context — numbers with no statement of what they mean for identity.
  • Classification from a single descriptor when the others disagree.
  • Absolute cable length or arbor volume reported for truncated cells without a lower-bound qualifier.
Module 09 · Nanoscale

Exit Ticket

Explain one morphology feature that could be confounded by reconstruction quality.

Module 09 · Nanoscale

References (Instructor)

  • Use module references listed on the module page.
Module 09 · Nanoscale

Teaching Materials

  • Module page: /modules/module09/
  • Session kit: /teaching/sessions/module09/
  • Worksheet: /assets/worksheets/module09/module09-activity.md
Module 09 · Nanoscale

Read, review, or teach

Read these slides as a page

Open the complete lesson

The reading version includes figure descriptions and discussion guidance.

Nanoscale text: CC BY 4.0. Scientific figures retain their individual credits.

Module 09 · Nanoscale

Pre-class preparation (10 min async) Review the data formats content library entry (skeletons section) Install/check NeuroM or equivalent morphology analysis package Minute-by-minute plan 00:00-10:00 | Morphology overview "Why do we care about neuron shape?" — Shape constrains function: a neuron's dendritic arbor determines what inputs it can receive; its axonal arbor determines where it can send output. Show 3 neuron types (pyramidal, basket, Martinotti) and their characteristic morphologies. "Today you'll learn to quantify these shapes from EM data." 10:00-24:00 | Skeleton extraction demo Live demo: take a segmented neuron, run skeletonization, visualize result in Neuroglancer. Walk through SWC format: "Each line is a node. Parent ID tells you the tree structure." Common pitfall: show a skeleton with spurious branches from noisy segmentation. Demonstrate pruning. 24:00-38:00 | Descriptor calculation Hands-on: learners compute 5 descriptors for one neuron using NeuroM or provided scripts. Compare results across the group: did everyone get the same numbers? Discuss sources of variation. Introduce Sholl analysis with live visualization. 38:00-50:00 | Interpretation and caveats "Your neuron has total cable length of 2,100 μm and 47 branch points. Is that a lot?" — Compare to published values for the putative cell type. Discussion: which descriptors are robust to reconstruction errors? (Cable length is sensitive to splits; branch count is sensitive to both splits and spurious branches; spine density is robust if the segmentation boundary is accurate.) "What if 30% of the arbor is outside the volume? How does that change your interpretation?" 50:00-60:00 | Competency check Each learner submits their morphology descriptor table with: At least 3 descriptors with values Putative cell-type classification based on morphology One explicit limitation of the measurement Exit ticket: "Name one morphology feature that could be confounded by reconstruction quality."