Skip to content

M-max estimation and review

Purpose

M-max in MonStim is an estimate derived from M-wave amplitudes across stimulus levels. It is available for normalization only after the M-wave timing, amplitude method, stimulus range, and detected plateau have been reviewed. It is not a direct physiological validation of saturation.

Inputs and scope

For each channel, MonStim takes paired stimulus voltages and M-wave amplitudes calculated with the current amplitude method. Dataset and experiment results aggregate child values; changing method, latency windows, exclusions, or binning can therefore change M-max.

Name and create the M-wave window

M-max uses the first latency window whose name matches a configured M-wave recognition name without regard to case. The shipped names are M-wave, M_wave, M wave, Mwave, M-response, M_response, and M response; use the canonical name M-wave for new work. You can add or remove names globally in File > Settings Center > Global Analysis > Latency windows > M-wave Recognition Names. The list replaces the shipped names, does not extend them, and is not controlled by analysis profiles.

An empty recognition-name list disables automatic M-wave classification. M-max calculation and relative-to-M-max normalization will then have no valid M-response window until you restore or add a name.

Create or adjust the window in Edit > Session/Dataset/Experiment > Manage Latency Windows. The window must cover the intended M-response for the channel and protocol; inspect it on the signal before relying on its amplitudes. Configure only one recognized M-wave window in a session—if several recognized names are present, MonStim uses the first one in the window order.

For the editor workflow, channel-specific starts, and the difference between Session, Dataset, and Experiment changes, see Latency windows.

Step 1: smooth and search for a low-variation run

The M-wave amplitude stimulus-response sequence is smoothed with a third-order Savitzky–Golay filter. If no explicit smoothing length is supplied, MonStim uses int(number_of_points × savgol_window_ratio), enforces a minimum length of 5, and makes the length odd.

MonStim tests contiguous runs beginning at the configured largest window size. It uses the 95th percentile of the filtered curve's absolute amplitudes as a robust curve-amplitude reference. A run is a candidate plateau when its standard deviation is below the configured threshold multiplied by that reference. If no run is found, it retries with one fewer point down to the configured smallest window size. When more than one run qualifies at that size, the later qualifying run is used.

The threshold is a unitless relative-variation fraction. The shipped value, 0.15, allows a plateau-window standard deviation up to 15% of the curve's robust amplitude (95th percentile). This makes plateau detection invariant to a uniform rescaling of the signal, including a change in amplitude units or gain. A qualifying run is an algorithmic low-variation region, not proof of a physiological plateau.

Step 2: estimate from a detected plateau

For the selected run, MonStim computes a maximum, a 95th percentile, a mean of values at/above the 80th percentile (top 20%), and a corrected mean. The corrected mean starts as the run mean and may add the difference between the mean of globally higher values and the mean of non-maximum run values.

Selection order is maximum, 95th percentile, top-20% mean, then corrected mean. A candidate is accepted only when it is no greater than plateau_mean × validation_tolerance. The final value is capped at the global maximum amplitude.

Step 3: no plateau found

MonStim reports M-max unavailable: no plateau detected and does not produce an automatic M-max value. This prevents a high-stimulus heuristic from being used as if it demonstrated saturation. Review the M-wave window, stimulus range, and acquisition quality before deciding whether to repeat the protocol or use a separately documented manual value.

Review before normalization

  1. Confirm the M-wave latency window and amplitude method.
  2. Check that the stimulus range reaches the intended high-response region.
  3. Confirm that a plateau was detected; otherwise, do not use automatic relative-to-M-max normalization.
  4. Record M-max settings and manual overrides with exported results.
  5. Do not compare normalized values from materially different settings without an explicit rationale.

See Configuration reference for the shipped parameters and their defaults.