Brain Activity with EEG Frequency Bands Detecting Depression

Detecting Depression in Brain Activity with EEG Frequency Bands

Introduction

According to Steiger and Kimura, sleep and waking electroencephalogram (EEG) can act as anti-depressive therapy and depression biomarkers. The complex EEG signals are challenging to interpret and derive the associated figures hence a need for computer-aided diagnosis (CAD). For this reason, non-linear methods and chaos theory are often used in diagnosis. Specification of the value is essential to allow depression detection. Several responses can be used to evaluate the significance of using the EEG signals in depression detection. What if the neurons oscillated in unison and at a specific frequency on a scale detectable by EEG? How would the modeling help in depression detection?

Clinical Relevance

The EEG value that specifies the existence of depression in the brain can be defined to help act as the reference point. The divergence from the threshold value can determine the depth and severity of depression in the brain. Clinical detections guide these values. The detected value can determine the type of therapy effective for the depressed patient. According to Acharya et al., the EEG modeling for signals in the left side of the brain are sufficient to detect depression. The retrieved signals are compared with the normal clinical levels. These models are shown below.

The collected data is captured and processed, a model developed, and the output classified to determine if the signals depict any forms of depression. The extracted data\\\\\\\'s ranking uses differential figures such as fractal dimension, high-order spectra, and detrended fluctuation analysis with the CAD system. It is up to the clinicians to determine the most effective form of interpretation based on existing medical data. Non-linear methods are easier to use because the EEG outcome is used as a confirmation tool of the diagnosis. Besides, EEG signal processing can be used in early depression detection.

Methods

Newson and Thiagarajan explain that deep learning algorithms are used. The fusing of interhemispheric asymmetry is done and correlated with EEG values. Suppose we have 32 subjects, 16 with major depressive disorders and 16 who act as controls. The theta, alpha, and beta frequency bands are extracted. The connecting features are examined and segmented to fit the EEG signals. Structural matrices from the three frequency bands define a mixed matrix. Such an approach helps establish high accuracy levels of at least 94.13 %, 93.52 % specificity, and 95.74 % sensitivity. Depending on the modeling used during matrix development, these values can vary. However, the EEG model gives a high resolution, and functionalities inform the variation in values in the brain. When recording the EEG values, the subjects sit on an armchair and adopts a relaxed mode where they are awake for 3 minutes. The room temperatures are regulated to remain within 23 degrees margin. The EEG headset is then used to capture the data. The data is captured from 64 brain channels.

Framework

According to Newson and Thiagarajan, there are four important components in the EEG analysis. First is preprocessing the EEG signal and segmenting it, extraction of features id one, the feature matrix is created, and lastly, the output is classified. Theta frequency band is classified between 4 and 8 Hz, alpha 8 to 13 Hz, and beta 13 to 40 Hz. Changes in these bands are used to determine the asymmetric interhemispheric changes.

Analysis

EEG signals are non-stationary and vary with time. For better results, those signal segments are spread into unit second intervals. This allows monitoring signals at different states because the brain\\\\\\\'s response will vary based on the prevailing conditions. In some instances, the EEG signal will give a significantly high value and low in some other cases. Thus, a short interval is essential.

Limitations