9.2 Data Review, Reformatting & Montage Re-montaging

9.2 Data Review, Reformatting & Montage Re-montaging

Modern digital long-term video-EEG monitoring (LTM) systems have transformed neurophysiology by completely decoupling data acquisition from data display. In legacy analog EEG, montages, high-frequency filters (HFF), low-frequency filters (LFF), and sensitivity were permanently fixed onto paper at the moment of recording. In digital systems, signals from exploring electrodes are continuously digitized and stored as raw numerical voltage values referenced to a single hardware recording reference (such as $C_z$, $P_z$, or a dedicated reference electrode). This architecture enables the reviewing technologist and epileptologist to re-montage, re-filter, adjust display gains, and manipulate time-bases post-acquisition without modifying the pristine underlying digital master record.

Mastering digital review workflows, automated spike/seizure detection auditing, and digital re-montaging principles is essential for uncovering subtle epileptogenic foci, eliminating obscuring artifacts, and pruning multi-day recordings for permanent clinical archiving.

---

1. Digital Review Workflow & Automated Detection Auditing

Reviewing 24 to 168+ hours of continuous LTM data requires a structured, multi-pass workflow. Relying solely on automated software detection algorithms is dangerous, as mathematical algorithms are subject to both false-positive triggers and critical false-negative misses.

+-----------------------------------------------------------------------------+
|                   DIGITAL LTM DATA REVIEW WORKFLOW PIPELINE                 |
|                                                                             |
|   [STEP 1] SYSTEMATIC SCREENING & EVENT AUDIT                               |
|       - Audit all Patient Push-Button Event Markers                         |
|       - Audit Staff / Nursing Observation Markers                           |
|       - Audit Automated Seizure Detection (ASD) algorithm triggers          |
|       - Review Automated Spike Detection (AED) density and hourly trends    |
|                                                                             |
|   [STEP 2] SYSTEMATIC MANUAL VISUAL PRUNING & SCREENING                     |
|       - Page-by-page visual review of continuous baseline data              |
|       - Rapid screening using compressed time-bases (30s / 60s per page)    |
|       - Identify subtle rhythmic evolving patterns and subclinical seizures |
|                                                                             |
|   [STEP 3] DYNAMIC RE-MONTAGING & WAVEFORM LOCALIZATION                     |
|       - Switch to Longitudinal Bipolar to identify local Phase Reversals    |
|       - Switch to Transverse Bipolar to verify coronal / sagittal axes      |
|       - Switch to Average Reference (CAR) or Laplacian to assess fields     |
|                                                                             |
|   [STEP 4] FILTER & SENSITIVITY OPTIMIZATION                                |
|       - Toggle LFF (0.1 - 1.0 Hz) to assess slow waves / DC shifts          |
|       - Toggle HFF (35 - 70 Hz) to resolve muscle artifact vs. fast spikes  |
|       - Expand time-base (2s - 5s) to resolve polyspikes and fast onsets    |
|                                                                             |
|   [STEP 5] EVENT CLIPPING, PRUNING & PERMANENT ARCHIVING                    |
|       - Clip and export all clinical and electrographic seizures with video |
|       - Clip representative baseline wakefulness, N2 sleep, and REM epochs  |
|       - Prune uninformative continuous baseline data per hospital retention |
+-----------------------------------------------------------------------------+

Automated Algorithms vs. Manual Visual Review

Automated detection algorithms employ mathematical feature extraction, including amplitude thresholding, spectral power density shifts, rhythmicity analysis, and spatial pattern matching:

• Automated Seizure Detections (ASD): ASD algorithms demonstrate high sensitivity for prolonged, high-amplitude, rhythmic seizures (e.g., $3\text{ Hz}$ generalized spike-wave, robust temporal lobe seizures). However, ASDs frequently trigger false positives during physiologic rhythmic patterns (drowsiness with rhythmic theta, hyperventilation buildup), movement artifacts (chewing, rhythmic rocking, patting, brushing teeth), and electrode pop artifacts. Crucially, ASD algorithms can produce false negatives on low-voltage fast ictal onsets ($>15\text{ Hz}$ low-amplitude beta activity), brief focal seizures ($<10\text{ seconds}$), or subtle frontal lobe seizures obscured by movement.
• Automated Spike Detections (AED): Useful for generating hourly spike-density trend graphs. However, AEDs frequently misclassify sharp physiological sleep transients (vertex sharp waves, POSTS, wicket spikes) and sharp artifacts (lateral rectus spikes, chewing) as epileptiform discharges.

---

2. Digital Re-Montaging Techniques

A montage is a defined mathematical arrangement of channel derivations displaying potential differences across the scalp. Digital reformatting recalculates these differences instantaneously according to Kirchhoff's voltage laws ($V_{A-B} = V_A - V_B$).

+-----------------------------------------------------------------------------+
|                        DIGITAL MONTAGE COMPARISON MATRIX                    |
|                                                                             |
|   MONTAGE TYPE        DERIVATION FORMULA         PRIMARY STRENGTHS / HAZARDS|
|   --------------      -------------------------  -------------------------- |
|   LONGITUDINAL        Ch 1 = Fp1 - F7            - Localizes focal spikes   |
|   BIPOLAR             Ch 2 = F7  - T3              via Phase Reversals      |
|   ("Double Banana")   Ch 3 = T3  - T5            - Cancels widespread noise |
|                       Ch 4 = T5  - O1            - Cancels in-phase activity|
|                                                                             |
|   TRANSVERSE          Ch 1 = F7  - Fp1           - Localizes across coronal |
|   BIPOLAR             Ch 2 = Fp1 - Fp2             axis (left to right)     |
|   (Coronal)           Ch 3 = Fp2 - F8            - Resolves parasagittal vs.|
|                                                    lateral convexity foci   |
|                                                                             |
|   REFERENTIAL         Ch 1 = Fp1 - A1            - Measures absolute voltage|
|   (Ear / Vertex /     Ch 2 = F3  - A1            - Ideal for broad fields   |
|   Linked Ears)        Ch 3 = C3  - A1            - DANGER: Contaminated ref |
|                                                    distorts all channels    |
|                                                                             |
|   COMMON AVERAGE      Ch 1 = Fp1 - AVG           - Unbiased global reference|
|   REFERENCE (CAR)     where AVG = sum(all)/N     - Ideal for focal spikes   |
|                                                  - DANGER: "Ghost Spikes"   |
|                                                                             |
|   LAPLACIAN / SOURCE  Ch 1 = F3 - (Fp1+C3+Fz+F7)/4- High-pass spatial filter|
|   DERIVATION                                     - Emphasizes focal sources |
|                                                  - Suppresses broad fields  |
+-----------------------------------------------------------------------------+

Longitudinal vs. Transverse Bipolar Montages

• Longitudinal Bipolar (Anterior-Posterior / "Double Banana"): Arranged in parallel anteroposterior chains (left temporal, left parasagittal, midline, right parasagittal, right temporal). Each channel subtracts the potential of a posterior electrode from an anterior electrode ($F_{p1}-F_7$, $F_7-T_3$, $T_3-T_5$, $T_5-O_1$). Localization is confirmed when two adjacent channels sharing an exploring electrode display deflections in opposite directions (True Instrumental Phase Reversal).
• Transverse Bipolar (Coronal): Arranged horizontally across the scalp from left to right (e.g., $F_7-F_{p1}$, $F_{p1}-F_{p2}$, $F_{p2}-F_8$; and $T_3-C_3$, $C_3-C_z$, $C_z-C_4$, $C_4-T_4$). Transverse montages are indispensable for determining whether a discharge is maximal over the lateral temporal neocortex ($T_3/T_4$), the parasagittal convexity ($C_3/C_4$), or the midline vertex ($C_z$).

Referential Montages & The Contaminated Reference Hazard

In referential montages, each exploring electrode is paired with a relatively inactive reference electrode (such as the ipsilateral earlobe $A_1/A_2$, the vertex $C_z$, or linked ears $[A_1+A_2]/2$).

• Maximum Amplitude Principle: In an uncontaminated referential montage, the focus of an epileptiform discharge is localized to the exploring electrode that displays the highest peak voltage deflection.
• Contaminated Reference: If the reference electrode itself lies within the electrical field of the discharge (e.g., an active anterior temporal spike spreading to ipsilateral earlobe $A_1$), that spike voltage appears in Input 2 of every left-hemisphere channel. This causes:
  1. Amplitude cancellation at the true focus ($F_7-A_1$ or $T_3-A_1$ shows artificially small deflections).
  2. Appearance of false, inverted "mirror" spikes across healthy, distant channels (e.g., $O_1-A_1$ or $F_{p1}-A_1$).

Common Average Reference (CAR) & "Ghost Spikes"

The Common Average Reference mathematically calculates the average voltage of all recording scalp electrodes at each time sample and uses that value as the reference: $\text{Ref}_{\text{AVG}} = \frac{1}{N}\sum_{i=1}^{N} V_i$ $\text{Channel}_k = V_k - \text{Ref}_{\text{AVG}}$

• Advantage: Eliminates dependence on any single anatomical reference electrode, providing an unbiased display of focal spikes and widespread potential distributions.
• The "Ghost Spike" Phenomenon: If a single electrode generates an exceptionally high-voltage focal discharge (e.g., a $320;\mu\text{V}$ spike at $T_3$), that single spike elevates the mathematical average $\text{Ref}_{\text{AVG}}$ by $\frac{320}{20} = 16;\mu\text{V}$ (in a 20-channel array). When this average is subtracted from all other quiet electrodes ($0;\mu\text{V} - 16;\mu\text{V} = -16;\mu\text{V}$), an artificial, inverted deflection (ghost spike) appears synchronously across all uninvolved channels throughout both hemispheres.

Laplacian (Source Derivation)

The Laplacian montage functions as a high-pass spatial filter. It subtracts the weighted average voltage of the immediately surrounding neighbor electrodes from the central exploring electrode: $V_{\text{Laplacian}}(C_3) = V(C_3) - \frac{V(F_3) + V(P_3) + V(T_3) + V(C_z)}{4}$

• Clinical Application: Effectively eliminates volume-conducted, widespread background rhythms and distant activity, highlighting only the localized, high-gradient cortical sources directly beneath the exploring electrode. It is particularly valuable for resolving focal neocortical spikes and mapping cortical irritative zones during presurgical evaluations.

---

3. Digital Filter Optimization & Artifact Management

Digital filtering enables the technologist to alter the frequency passband dynamically during review. However, inappropriate filter adjustments can introduce significant waveform distortions and diagnostic errors.

+-----------------------------------------------------------------------------+
|                     DIGITAL FILTER CHARACTERISTICS & IMPACT                 |
|                                                                             |
|   FILTER PARAMETER   STANDARD SETTING   ADJUSTED SETTING  EFFECT & CLINICAL |
|   ----------------   ----------------   ----------------  ----------------- |
|   HIGH-FREQUENCY     70 Hz              35 Hz             - Attenuates EMG  |
|   FILTER (HFF)                                              muscle artifact |
|                                                           - Blunts spike    |
|                                                             sharpness &     |
|                                                             reduces amp.    |
|                                                                             |
|   LOW-FREQUENCY      1.0 Hz             0.1 Hz - 0.3 Hz   - Enhances slow   |
|   FILTER (LFF)                                              waves & ictal   |
|                                                             DC shifts       |
|                                                           - Accentuates slow|
|                                                             sweat artifact  |
|                                                                             |
|   LOW-FREQUENCY      1.0 Hz             5.0 Hz            - Attenuates slow |
|   FILTER (LFF)                                              breathing/sweat |
|                                                           - Artificially    |
|                                                             shortens spike  |
|                                                             after-slow wave |
|                                                                             |
|   NOTCH FILTER       OFF                60 Hz (USA) /     - Eliminates line |
|                                         50 Hz (Europe)      frequency hum   |
|                                                           - Can produce     |
|                                                             ringing artifact|
|                                                             on sharp spikes |
+-----------------------------------------------------------------------------+

Filter Mechanics and Clinical Rules:

• High-Frequency Filter (HFF) Adjustments: Lowering the HFF from $70\text{ Hz}$ to $35\text{ Hz}$ suppresses excessive myogenic (EMG) muscle artifact during motor seizures or agitation. However, because spikes and sharp waves contain high-frequency Fourier components ($>40\text{ Hz}$), a $35\text{ Hz}$ filter will blunt the pointed peak, reduce measured spike amplitude, and artificially increase measured duration ($>70\text{ ms}$), potentially causing a true spike to resemble a rounded sharp wave or background theta transient.
• Low-Frequency Filter (LFF) Adjustments: Standard LFF is set at $1.0\text{ Hz}$ (time constant $\approx 0.16\text{ seconds}$). When evaluating structural lesions, post-ictal focal slowing, or slow DC potential shifts at seizure onset, decreasing the LFF to $0.1\text{ to }0.3\text{ Hz}$ reveals ultra-slow delta activity. Conversely, increasing the LFF to $5.0\text{ Hz}$ to remove sweat or respiration artifact will distort spike-wave complexes by abolishing the after-coming slow wave and shifting waveform phase.
• 60 Hz Notch Filter: The notch filter eliminates continuous $60\text{ Hz}$ electrical line interference. While digital notch filters have narrow rejection bandwidths ($59.5\text{ to }60.5\text{ Hz}$), relying on the notch filter can mask poor electrode contact (high impedances) and can introduce "ringing" artifacts on high-amplitude sharp transients. Technologists should always correct the physical impedance source before enabling the notch filter.

---

4. Display Time-Base & Sensitivity Scaling

Dynamic adjustment of display scaling is critical for resolving different neurophysiological phenomena across multi-day records.

+-----------------------------------------------------------------------------+
|                  TIME-BASE & SENSITIVITY OPERATIONAL GUIDE                  |
|                                                                             |
|   PARAMETER       SETTING          CLINICAL APPLICATION & UTILITY           |
|   ---------       -------          ------------------------------           |
|   TIME-BASE       10 sec / page    - Standard ACNS clinical review speed    |
|   (Epoch Length)  (30 mm/s equiv)  - Balances background & spike morphology |
|                                                                             |
|                   30 - 60 sec/page - Compressed review for long trends      |
|                   (Trend View)     - Cyclic sleep cycling, burst suppression|
|                                    - Identifying slow rhythmic ictal onsets |
|                                                                             |
|                   2 - 5 sec/page   - Expanded review for high-frequency     |
|                   (Zoomed View)    - Resolving polyspike components & HFOs  |
|                                    - Separating spike from lateral rectus   |
|                                                                             |
|   SENSITIVITY     7 uV / mm        - Standard default calibration scaling   |
|                   (Standard)       - Normal adult background (20 - 70 uV)   |
|                                                                             |
|                   2 - 3 uV / mm    - Electrocerebral Inactivity (ECI) / Coma|
|                   (High Gain)      - Low-voltage fast ictal onsets / ECI    |
|                                                                             |
|                   15 - 30 uV / mm  - High-voltage pediatric EEG / Spasms    |
|                   (Low Gain)       - Hypsarrhythmia, severe spike bursts    |
+-----------------------------------------------------------------------------+

Clinical Applications of Scaling Adjustments:

• Compressed Time-Bases ($30\text{ to }60\text{ seconds/page}$): Enables rapid visual scanning of hours of data. Periodic discharges (LPDs, GPDs), sleep architecture cycling (NREM vs. REM transitions), burst suppression ratios in ICU sedation titration, and the gradual evolution of slow rhythmic ictal patterns become immediately visible as distinctive spatial "striping" or rhythmic envelopes.
• Expanded Time-Bases ($2\text{ to }5\text{ seconds/page}$): Indispensable for dissecting complex waveforms. When evaluating spike-wave complexes, an expanded time-base separates closely spaced polyspikes, allowing measurement of individual spike durations ($20\text{ to }70\text{ ms}$) and identification of High-Frequency Oscillations (HFOs: ripples $80\text{--}250\text{ Hz}$ and fast ripples $250\text{--}500\text{ Hz}$) at the seizure onset zone.
• Sensitivity Scaling: Standard sensitivity is $7;\mu\text{V/mm}$ ($50;\mu\text{V}$ produces a $7.14\text{ mm}$ deflection). In pediatric patients with hypsarrhythmia or high-voltage generalized spike-wave discharges exceeding $300;\mu\text{V}$, channels overlap and block one another; reducing sensitivity to $15\text{ to }30;\mu\text{V/mm}$ eliminates trace clipping. In suspected electrocerebral inactivity (ECI / brain death) or profound coma, sensitivity must be increased to $2;\mu\text{V/mm}$ for at least 30 minutes to verify the absence of cerebral activity exceeding $2;\mu\text{V}$.

---

5. Event Clipping, Pruning & Long-Term Archival Workflows

Multi-day LTM recordings generate massive digital data files (tens to hundreds of gigabytes per patient). Storing unpruned raw video-EEG indefinitely exceeds institutional storage area network (SAN) capacity. Technologists must execute structured data clipping and pruning workflows according to ACNS guidelines and hospital retention policies.

+-----------------------------------------------------------------------------+
|                   DATA CLIPPING & PRUNING PROTOCOL MATRIX                   |
|                                                                             |
|   [1] SEIZURE & CLINICAL EVENT CLIPS                                        |
|       - Enough pre-event baseline to establish earliest onset                |
|       - Complete clinical and electrographic event                            |
|       - Post-event context through meaningful recovery per protocol           |
|       - Synchronized video, audio, and polygraphy when recorded                |
|                                                                             |
|   [2] REPRESENTATIVE PHYSIOLOGICAL BASELINE SAMPLES                         |
|       - 10 to 15 minutes of resting awake background with eye open/closed   |
|       - 10 to 15 minutes of Stage N2 sleep showing sleep spindles & K-cmplx |
|       - 10 minutes of Stage N3 slow-wave sleep                              |
|       - 10 minutes of Stage REM sleep with polygraphic EOG and EMG traces   |
|       - Full duration of provocative maneuvers (HV, Photic Stimulation)     |
|                                                                             |
|   [3] DATA RETENTION & MEDICO-LEGAL MANDATES                                |
|       - Follow the approved HIM/legal retention schedule for the governing  |
|         jurisdiction, record type, and any litigation or regulatory hold.   |
|       - Verify the retained master and audit trail before source deletion.   |
|       - Never calculate a pediatric destruction date from a universal       |
|         age-plus-years formula.                                             |
+-----------------------------------------------------------------------------+

Clinical Onset vs. Electrographic Onset Marking

When clipping an event, the technologist must place two separate, unambiguous electronic markers:

1. Electrographic Onset (EO): The exact millisecond timestamp corresponding to the earliest rhythmic, spectral, or morphological deviation from the ongoing baseline background (e.g., regional electrode decremental event, paroxysmal low-voltage fast beta activity, or rhythmic sinusoidal theta train).
2. Clinical Onset (CO): The exact timestamp corresponding to the earliest observable behavioral, motor, speech, or autonomic change on video, or the moment the patient signals an aura via push-button activation.