How EEG-Interpreting Software Works
Artificial intelligence (AI) can be defined as technology that mimics and even improves upon human intelligence (Wang & Preininger, 2019).
Artificial intelligence has been applied to analysis of various types of human health data. Machine learning and deep learning both fall under the broader category of artificial intelligence.
However, levels of complexity differentiate the two technologies. Machine learning involves more basic functions, while deep learning is a more complex form of machine learning that creates models from the available data (Wang & Preininger, 2019).Subscribe
A review of the extant literature consistently demonstrates that one of the more accurate means of interpreting electroencephalogram (EEG) data appears to be the support vector machine (SVM) model, which is a type of machine learning (e.g., Li et al., 2019; Mahato & Paul, 2020; Mumtaz et al., 2017). Of note, research that examined methods of EEG analysis
analysis between 1988 and 2018 corroborated this finding. This study found that virtually all types of
machine learning models have been utilized to interpret EEGs, and further, supervised learning methods
such as SVM and k-nearest-neighbors (KNN) were most accurate (Hosseini et al., 2021).
Challenges to Model Development
Support vector machine model
Offers greater diagnostic accuracy
Her main problem:
Offers greater diagnostic accuracy
Utilizing deep learning:
Way to minimize the amount of human intervention needed to analyze data
Unlike other types of health-related data, interpreting EEGs presents unique challenges due to the continuous nature of brain waves as well as the large amount of data generated.
Therefore, additional steps are required before accurate interpretation can occur, including denoising and calibration. Another concern in developing accurate predictive models involves measurement error from physical devices.
Contributors to noise on EEGs include equipment interference, motor movements and eye movements. Among the mother wavelet functions tested, coif3 was found to be most effective in addressing these types of noise (Alkareem Alyasseri, 2017).
Use of artificial intelligence to develop a second opinion for the purpose of assisting clinicians in diagnostic decision-making.Subscribe
The depression diagnosis index
Category of automated computer diagnosis that assists clinicians in using EEGs to diagnose depression.
The depression diagnosis index proposed by Archarya and colleagues combines multiple nonlinear features into one numerical score that translates to an objective diagnosis (Archarya et al., 2015a). However, due to the current state of development of EEG technology, the depression diagnosis index may also be considered computer-aided diagnosis, as human clinical judgment may be necessary to fully describe depressive disorder features and specifiers.
The Role of Neuroscience for Software
Without the use of computer software, EEGs are difficult to interpret. Computer–aided diagnosis (CAD) facilitates integration of both linear and nonlinear methods of interpreting EEGs into initial detection of depression as well as confirming human diagnostic assessments.
In addition to improving patient quality of life, the end goal of this product is ultimately to significantly reduce suicide as one of the outcomes of depression.
Neuroscience Software aims to provide a user-friendly
technical solution to facilitate the ability of clinicians to facilitate the ability of clinicians to diagnose accurately and provide timely treatment, as well as monitor response to treatment. This will be accomplished through analysis of both linear and non-linear data.
The software will integrate with a number of existing EEG measurement devices to analyze biomarkers and provide a depression diagnosis index on a scale ranging from 1 to 10, in which 1 indicates low probability of depression and 10 indicates high probability.