How EPM Arm Entries, Time and Head Dip Are Calculated

1. What these EPM arm metrics represent

The Elevated Plus Maze has two open arms, two closed arms and a center platform. RodentMetrics reports, for each completed analysis: open-arm entries, closed-arm entries, open-arm time, closed-arm time and center time.

Open- and closed-arm time and entries are commonly interpreted in studies of anxiety-like behavior, on the general premise that rodents tend to avoid open, exposed spaces. RodentMetrics reports these as behavioral measurements only — it does not itself diagnose, score or classify an anxiety-like state from them.

2. How RodentMetrics determines arm occupancy

Every frame's zone is classified from a single tracked point: the tracker's bounding-box center, transformed into the same coordinate space used throughout the analysis. That point is tested against the five region polygons you defined for the maze (open arm 1, open arm 2, closed arm 1, closed arm 2, center); on an exact shared boundary, center takes priority so a boundary point is never silently assigned to an arm.

This bbox-center classification is the only zone system used for official arm time and entries. An experimental body-overlap measurement exists separately in the codebase but is explicitly excluded from official state and does not affect the metrics on this page.

3. How RodentMetrics counts arm entries

An entry is registered once the tracked point's classified zone has been continuously confirmed inside an arm for a minimum duration — 0.20 seconds by default, adjustable per experiment from the arena setup. This is a genuine temporal-confirmation rule, not a same-frame zone check: a single frame or two of noise crossing into an arm does not register as an entry.

Two related rules affect the entry count directly:

• A same-arm return via only a very brief touch of the center platform is treated as tracking jitter and suppressed, rather than counted as a separate exit-then-reentry.
• A direct transition between two arms is never accepted as a valid step unless an actual passage through the center platform was observed in between — it is either bridged through that observed center run, or logged separately as an invalid transition and excluded from the confirmed sequence.

Take the illustrative, synthetic transition sequence Center → Open Arm 1 → Center → Closed Arm 1 (not data from a real recording). Assuming each shown transition is itself already temporally confirmed, this sequence registers exactly one open-arm entry and one closed-arm entry — the starting time in center is not an entry (no transition into it was observed), and each subsequent arm segment is a new entry because a center visit separates it from the last.

4. How time in each zone is calculated

Open-arm time is the combined dwell time in open arm 1 and open arm 2; closed-arm time is the combined dwell time in closed arm 1 and closed arm 2; center time is reported separately from both. Dwell accumulates across half-open frame intervals (each frame contributes from its own timestamp up to, but not including, the next frame's timestamp), bounded by the confirmed transition timeline rather than by raw, unconfirmed per-frame zone guesses.

Frames where the animal was not detected, or where its position did not classify into any of the five regions, are excluded from zone dwell entirely and reported separately as unobserved time — they are never guessed at or assigned to the nearest zone.

5. Head Dip: a separate, pose-based diagnostic measure

Head Dip is not part of the tracker-based measurements above. It uses a distinct, pose-based detector that looks at whether the animal's nose keypoint extends past an open arm's outer edge (or past the center platform's outer edge while the body stays in center), for at least a minimum duration, after the animal has been stably positioned for a short anchoring period. Closed-arm walls are not used as Head Dip edges — that geometry is not modeled reliably enough to score.

RodentMetrics distinguishes two kinds of candidates: unprotected (the nose dips past an open arm's edge while the body itself is in that arm) and protected (the nose dips toward an open arm's edge while the body remains anchored in the center platform, without the body actually entering the arm).

6. Boundary cases and interpretation

A few tracking situations are worth understanding before interpreting a result:

• If a recording begins with the animal already inside an arm, that time counts toward that arm's dwell from the start of the analysis window, but is not itself counted as an entry — no transition into it was observed.
• If a recording ends while the animal is inside an arm, dwell time accrues up to the end of the analysis window without requiring an observed exit.
• A brief, single-frame boundary crossing that does not sustain the 0.20-second confirmation window is not registered as an entry or a zone change at all.
• Missing or unclassified frames never fabricate a zone; they are excluded from dwell and reported as unobserved time rather than assigned to the nearest region.

7. Worked example

Take the same illustrative, synthetic transition sequence from Section 3 — Center → Open Arm 1 → Center → Closed Arm 1 — with example durations assigned to each already-confirmed segment (not data from a real recording):

Total observed time is 16.0 seconds. RodentMetrics would report: 1 open-arm entry, 1 closed-arm entry, 2 total arm entries; open-arm time 5.0s (31.3% of observed time), closed-arm time 8.0s (50.0%), and center time 3.0s (18.8%).

8. What these metrics do not establish

Open-arm avoidance is a widely used behavioral convention, not a diagnostic instrument. RodentMetrics' entry and time measurements alone do not:

• diagnose or clinically validate an anxiety phenotype;
• prove that a treatment or manipulation caused an anxiety-like change;
• constitute a validated psychological or clinical outcome measure.

These measurements describe zone occupancy and transitions derived from confirmed tracking. They should be interpreted alongside tracking/quality information, the experimental protocol, and appropriate group comparisons and statistics — not read in isolation as a conclusion about an animal's emotional state.

9. What to report

When publishing or exporting EPM arm results, RodentMetrics recommends reporting the parameters it actually computes and exports:

• Open-arm, closed-arm and total arm entries
• Open-arm, closed-arm and center time, with the observed-time percentage denominator
• First open-arm entry latency, and whether the session started in an open arm
• The transition confirmation duration used for the session
• Tracking/quality information (e.g. classified behavioral coverage)
• Whether Head Dip diagnostics were enabled and, if so, total/protected/unprotected counts or that they were unavailable
• The RodentMetrics software version the analysis was run with

Further reading

Walf AA, Frye CA. The use of the elevated plus maze as an assay of anxiety-related behavior in rodents. Nat Protoc. 2007;2(2):322-328. doi:10.1038/nprot.2007.44