Reduced-channel EEG: classification performance, selection stability and spectral computation cost

A compact EEG headset places fewer electrodes and, if the processor follows that choice, transforms fewer signals. This study asks whether a shorter montage can cut the host time of spectral feature extraction while classification stays within the resolution of one cohort. Eyes-closed recordings from OpenNeuro ds004504 were analysed as 36 participants with Alzheimer's disease and 29 cognitively normal participants. Each person is one vector of 95 relative band powers from 19 channels and five bands. Sets of 4, 6, 8 and 12 channels were chosen inside each outer-training fold and scored with the same L2 logistic regression on five repeated partitions. The 19-channel reference reached a mean balanced accuracy of 0.745 (percentile interval 0.653–0.834) and a mean ROC-AUC of 0.833. Eight-channel mutual-information and mRMR sets reached 0.770. That difference does not meet the predefined rule for a resolved improvement. On the recorded host, batched Welch estimation of 150 synthetic windows took 34.43 ms with 19 channels and 3.64 ms with four, and the coefficients occupy 384 or 84 bytes. The measured saving appears when discarded channels are actually skipped. These results support studying compact montages as an engineering trade-off among electrode count, stability and spectral time. They do not validate a wearable device or an Alzheimer's-specific biomarker: Mini-Mental State Examination scores separate the two groups completely.

Authors

Institutions

Publication Details

Published
2026-11-17
DOI
https://doi.org/10.53388/xiv2026000009
Primary Topic
EEG and Brain-Computer Interfaces
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Reduced-channel EEG: classification performance, selection stability and spectral computation cost

Abdelati Reha, Fatimazahra Jakjoud, Widad Jakjoud
EEG and Brain-Computer Interfaces
preprint

Reduced-channel EEG: classification performance, selection stability and spectral computation cost

Abdelati Reha, Fatimazahra Jakjoud, Widad Jakjoud
preprint en

Abstract

A compact EEG headset places fewer electrodes and, if the processor follows that choice, transforms fewer signals. This study asks whether a shorter montage can cut the host time of spectral feature extraction while classification stays within the resolution of one cohort. Eyes-closed recordings from OpenNeuro ds004504 were analysed as 36 participants with Alzheimer's disease and 29 cognitively normal participants. Each person is one vector of 95 relative band powers from 19 channels and five bands. Sets of 4, 6, 8 and 12 channels were chosen inside each outer-training fold and scored with the same L2 logistic regression on five repeated partitions. The 19-channel reference reached a mean balanced accuracy of 0.745 (percentile interval 0.653–0.834) and a mean ROC-AUC of 0.833. Eight-channel mutual-information and mRMR sets reached 0.770. That difference does not meet the predefined rule for a resolved improvement. On the recorded host, batched Welch estimation of 150 synthetic windows took 34.43 ms with 19 channels and 3.64 ms with four, and the coefficients occupy 384 or 84 bytes. The measured saving appears when discarded channels are actually skipped. These results support studying compact montages as an engineering trade-off among electrode count, stability and spectral time. They do not validate a wearable device or an Alzheimer's-specific biomarker: Mini-Mental State Examination scores separate the two groups completely.

École Marocaine des Sciences de l'Ingénieur (MA)
EEG and Brain-Computer Interfaces
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon