Why this unit

Process-type misclassification is a major source of downstream graph error.

Learning goals

• Distinguish axons and dendrites with morphology + organelle cues.
• Handle ambiguous edge cases with confidence scoring and continuity checks.

Core technical anchors

• Combined cue strategy: morphology, organelles, local connectivity.
• Edge cases: en passant boutons, thin dendrites, truncations.
• Auditability via confidence and correction logs.

Method deep dive: axon-vs-dendrite classification

1. Morphology pass: Branch caliber, tortuosity, spine presence, and process tapering.
2. Organelle pass: Vesicle clustering, microtubule patterning, mitochondrial distribution.
3. Connectivity pass: Input/output pattern and bouton/spine relationships in neighborhood.
4. Continuity pass: Validate interpretation along additional slices and branch points.
5. Decision logging: Capture confidence and alternative hypothesis if ambiguous.

Quantitative QA checkpoints

• Confusion matrix for axon/dendrite labels against adjudicated truth set.
• Error concentration map by tissue region and annotator.
• Rework fraction: percent of labels reversed during secondary review.
• Downstream sensitivity: impact of class errors on motif counts and graph metrics.

Frequent failure modes

• Thin dendrites mislabeled as axons: Add local synaptic-role evidence before finalizing.
• Truncated field-of-view bias: Mark as provisional when continuity evidence is missing.
• Bouton-centric bias: Avoid relying only on vesicle presence without full context.
• Inconsistent team criteria: Calibrate weekly with shared edge-case panels.

Visual training set

Attribution: Pat Rivlin training materials (MICrONS proofreading deck). Some manifest-listed IDs used in planning (`S04`, `S06`, `S10`, `S16`) were not present in extracted thumbnails and were replaced with available neighboring cues.

Course links

• Existing overlap: module04, module09
• Next unit: 07 Glia

Practical workflow

1. Start with morphology cues.
2. Add organelle and synaptic-context checks.
3. Verify continuity in neighboring sections.
4. Assign class with confidence and review note.

Discussion prompts

• Which edge cases most frequently produce classifier disagreement?
• How should teams calibrate confidence thresholds for class assignments?

Related resources

• Journal club list: Technical Track Journal Club
• Shared vocabulary: Connectomics Dictionary

Quick activity

Choose one ambiguous process and document the three strongest cues you used to classify it.

Content library references

• Axon-dendrite classification — Multi-cue protocol, edge cases, confidence scoring
• Axon biology — AIS, boutons, vesicle pools, myelination
• Dendrite biology — Spine types, local translation, organelle gradients
• Organelle annotation cues — Using organelles to resolve ambiguous compartments
• Neuron type identification — Morphological and connectivity-based classification

Teaching slide deck

• Slide draft page: Axons and Dendrites deck draft

Evidence pack: papers and datasets

This unit is anchored to canonical papers and datasets used in connectomics practice. Use these as required preparation before activities.

Key papers

• Harris and Weinberg (2012) - Ultrastructure of synapses in the mammalian brain
• Kasthuri et al. (2015) - Saturated reconstruction of neocortex
• Helmstaedter et al. (2013) - Connectomic reconstruction of the inner plexiform layer in the mouse retina

Key datasets

• MICrONS Explorer
• FlyWire
• neuPrint Hemibrain

Competency checks

• Classify neurites with multi-cue evidence and confidence tags.
• Document ambiguity handoff decisions for adjudication.

Capability development brief

Capability target: Classify neurites using multi-cue evidence and document uncertainty for downstream proofreading.

Required expertise

• Neuroanatomist (neurite morphology and targeting patterns)
• Proofreading expert (error pattern recognition)
• Computational morphologist (feature extraction and validation)

Core concepts to teach

• Multi-cue classification: Combining local texture, branching pattern, vesicles, and synaptic polarity cues.
• Local vs global context: Avoiding overconfidence from small patches by checking longer process trajectories.
• Ambiguity escalation: Structured handoff when cues conflict or confidence remains low.

Studio activity

Neurite Identity Challenge - Reduce false identity calls and improve escalation behavior.

Classify borderline neurites and justify each call with at least two independent cues.

1. Assign preliminary identity and confidence level.
2. Seek confirming or contradictory evidence along the process.
3. Escalate unresolved cases using rubric rules.

Expected outputs:

• Labeled casebook
• Escalation summary

Assessment artifacts

• Axon/dendrite decision rubric with ambiguity pathways.
• Quality log of difficult cases and final adjudications.

Related concepts

Axon vs Dendrite Classification

Apply multi-cue decision rules and confidence tags for process-type calls.

Open in Concept Explorer

reducing identity confusion handling ambiguity