This study explores the use of machine learning (ML) to identify brain-specific biomarkers from electroencephalography (EEG) signals, emphasizing the importance of distinguishing between brain and peripheral signals. EEG has long been used to study brain function, but its potential for deriving biomarkers is often limited by physiological artifacts from peripheral sources. The authors propose a framework to conceptualize ML from CNS versus peripheral signals measured with EEG. Using Morlet wavelets, they define traditional brain activity features and alternative inputs for ML. They analyzed over 2600 EEG recordings from public databases to assess the impact of peripheral signals and artifact removal techniques on ML models in age and sex prediction. Their findings show that peripheral signals enable age and sex prediction but explain only a fraction of the performance provided by brain signals. They demonstrate that both brain and body signals, reflected in EEG, allow for prediction of personal characteristics. However, they caution that great care is needed to separate these signals when developing CNS-specific biomarkers using ML. The study highlights the importance of artifact removal in EEG analysis. While basic artifact rejection improved model performance, further removal of peripheral signals using ICA decreased performance. This suggests that peripheral signals may contain predictive information. The authors also show that non-brain signals can be predictive of age and sex, but brain signals are the main driver of prediction performance. They emphasize the need for careful preprocessing to ensure that ML models isolate brain-related signals and avoid using non-brain signals that may contain predictive information. The study also explores the use of wavelet-based methods for spectral analysis and shows that these methods can provide insights into the underlying physiological processes. Overall, the study underscores the importance of distinguishing between brain and peripheral signals in EEG analysis to develop accurate and interpretable biomarkers.This study explores the use of machine learning (ML) to identify brain-specific biomarkers from electroencephalography (EEG) signals, emphasizing the importance of distinguishing between brain and peripheral signals. EEG has long been used to study brain function, but its potential for deriving biomarkers is often limited by physiological artifacts from peripheral sources. The authors propose a framework to conceptualize ML from CNS versus peripheral signals measured with EEG. Using Morlet wavelets, they define traditional brain activity features and alternative inputs for ML. They analyzed over 2600 EEG recordings from public databases to assess the impact of peripheral signals and artifact removal techniques on ML models in age and sex prediction. Their findings show that peripheral signals enable age and sex prediction but explain only a fraction of the performance provided by brain signals. They demonstrate that both brain and body signals, reflected in EEG, allow for prediction of personal characteristics. However, they caution that great care is needed to separate these signals when developing CNS-specific biomarkers using ML. The study highlights the importance of artifact removal in EEG analysis. While basic artifact rejection improved model performance, further removal of peripheral signals using ICA decreased performance. This suggests that peripheral signals may contain predictive information. The authors also show that non-brain signals can be predictive of age and sex, but brain signals are the main driver of prediction performance. They emphasize the need for careful preprocessing to ensure that ML models isolate brain-related signals and avoid using non-brain signals that may contain predictive information. The study also explores the use of wavelet-based methods for spectral analysis and shows that these methods can provide insights into the underlying physiological processes. Overall, the study underscores the importance of distinguishing between brain and peripheral signals in EEG analysis to develop accurate and interpretable biomarkers.