Source-linked AI summary
A systematic review of EEG source localization techniques and their applications on diagnosis of brain abnormalities
Shiva Asadzadeh, Tohid Yousefi Rezaii, Soosan Beheshti, Azra Delpak, Saeed Meshgini
TL;DR
EEG source localization addresses an underdetermined inverse problem important for studying brain function and disorders. This systematic review organizes 120 publications across major methodological and clinical application areas, highlighting the influence of head modeling on localization results.
Problem
EEG source localization requires solving a highly underdetermined inverse problem because scalp measurements do not uniquely determine brain sources.
Method
The paper systematically reviews publications using EEG source localization methods for brain functional disorders and categorizes their methodological and clinical applications.
Results
37.5% of reviewed studies addressed statistical solutions to the inverse problem, while 20% focused on epileptic-seizure detection and 18.33% on brain-abnormality diagnosis.
Takeaways & Limitations
More complex head models and patient-specific conductivity models can improve the accuracy of EEG source localization results.
Takeaways & Limitations
Electric-dipole EEG or MEG models cannot localize intracranial mesial temporal spikes.
Abstract
from arXiv · showhide
In recent years, multiple noninvasive imaging modalities have been used to develop a better understanding of the human brain functionality, including positron emission tomography, single-photon emission computed tomography, and functional magnetic resonance imaging, all of which provide brain images with millimeter spatial resolutions. Despite good spatial resolution, time resolution of these methods are poor and values are about seconds. Electroencephalography (EEG) is a popular non-invasive electrophysiological technique of relatively very high time resolution which is used to measure electric potential of brain neural activity. Scalp EEG recordings can be used to perform the inverse problem in order to specify the location of the dominant sources of the brain activity. In this paper, EEG source localization research is clustered as follows: solving the inverse problem by statistical method (37.5%), diagnosis of brain abnormalities using common EEG source localization methods (18.33%), improving EEG source localization methods by non-statistical strategies (3.33%), investigating the effect of the head model on EEG source imaging results (12.5%), detection of epileptic seizures by brain activity localization based on EEG signals (20%), diagnosis and treatment of ADHD abnormalities (8.33%). Among the available methods, minimum norm solution has shown to be very promising for sources with different depths. This review investigates diseases that are diagnosed using EEG source localization techniques. In this review we provide enough evidence that the effects of psychiatric drugs on the activity of brain sources have not been enough investigated, which provides motivation for consideration in the future research using EEG source localization methods.
INTRODUCTION
EEG offers high temporal resolution for brain activity localization, but its inverse problem is nonunique and highly underdetermined. This systematic review examines EEG source-localization techniques, applications to brain disorders, and factors affecting imaging accuracy.
- Motivation: EEG provides high temporal resolution, but mapping scalp measurements to brain sources is nonunique because skull conductivity limits measurable signals.Unlike fMRI, PET, and fNIRS, EEG supports real-time measurement of brain activity.
- Problem formulation: The EEG inverse problem is highly underdetermined because recordings contain fewer sensors than potential source locations and require prior constraints.The forward model represents propagation from source dipoles to electrodes, with additive white Gaussian noise in the measurements.
- Review objectives: The review identifies and investigates publications using EEG source localization to improve understanding of computerized techniques for brain functional disorders.Its research questions address techniques used, diagnosed or treated diseases, accuracy-affecting factors, and future implications.
- Review corpus: 120 papers were included, with 57.5% published between 2011 and 2018, 32.5% by the end of 2010, and 10% by the end of 2000.The review excluded 34 papers using SPECT, CST, or fMRI images.
- Research coverage: 37.5% of the reviewed research addressed statistical solutions to the inverse problem, followed by 20% on epileptic-seizure detection and 18.33% on common source-localization methods for brain abnormalities.Other categories covered head-model effects (12.5%), ADHD diagnosis and treatment (8.33%), and non-statistical improvements (3.33%).
E. Coutin-Churchman et al. [112]
The reviewed studies applied EEG source localization to characterize seizure-onset sources, ADHD-related abnormalities and treatment effects, and diverse psychiatric, neurological, and cognitive phenomena. Findings included clinically relevant source distributions, regional dysfunctions, and method-dependent localization accuracy.
- Epilepsy and seizure localization: 14 patients had 90–100% of seizure spikes within the resection site, while five patients with all sources outside it were not seizure free.Most patients with more than 50% of activity sources within the resection site were seizure free.
- Epilepsy and seizure localization: 76-channel EEG provided higher seizure-onset-zone detection accuracy than arrangements with fewer electrodes, while increasing electrodes enhanced sLORETA localization but yielded diminishing absolute accuracy gains.ExSo-MUSIC favored the strongest activity and performed better for single, deep sources with high signal-to-noise ratio.
- ADHD abnormalities: ADHD children showed delayed P100 and N200 latencies, reduced P100-NoGo amplitude, impaired Go performance, and decreased early NoGo electrical activity, with P100 sources in the occipital area.Visual sensory-processing deficits were also localized to the occipital region.
- ADHD treatment and modulation: Neurofeedback decreased impulsive behavior, while rTMS reduced TMS-evoked N100 amplitude that plateaued after almost 500 pulses and reflected stimulated motor-cortex effects.These findings linked source-localized EEG measures with modulation of ADHD-related behavior and stimulation responses.
- ADHD abnormalities: ADHD-related abnormalities involved reduced P3b amplitude, frontal-polar and temporoparietal differences, altered prefrontal alpha power, and response-selection activity in medial frontal and inferior parietal regions.Medial frontal pacemaker-accumulation processes differentiated ADD from the ADHD-combined subtype.
- Other brain abnormalities: Source localization identified regional patterns across brain abnormalities, including frontal dysfunction in conduct problems, bilateral parietal SREDA sources, a 7 mm mean V1 location error, and reduced frontocentral NoGo-P3 amplitudes in psychopathic traits.Other studies localized meditation-related gamma variation, phobia-related visual, cingulate, insular, and premotor sources, and frequency-specific responses to depression and tonic cold pain.
DISCUSSION AND CONCLUSIONS
EEG source localization is well suited to evaluating brain activity because of its high temporal resolution and supports diagnosis and treatment across several brain abnormalities. The review highlights method-specific limitations while identifying epilepsy and ADHD as major application areas and calling for further methodological development.
- DISCUSSION AND CONCLUSIONS: EEG source localization is suitable for evaluating brain activity during tasks because EEG has relatively high temporal resolution compared with methods such as MRI.The review concerns research on EEG source localization applications in functional brain diseases.
- DISCUSSION AND CONCLUSIONS: LORETA has low spatial resolution and produces blurred images of point sources, while principal component analysis is nearly useless for isolating epileptic spikes and sharp waves.These limitations affect localization quality in EEG recordings from patients with epilepsy.
- DISCUSSION AND CONCLUSIONS: Spatial information about spikes correlates strongly with background signals, reducing the accuracy of hybrid ICA and RAP-MUSIC; ExSo-MUSIC performs better than classical MUSIC.The passage attributes the accuracy problem to correlation between spike and background spatial information.
- DISCUSSION AND CONCLUSIONS: EEG source localization contributes substantially to diagnosing and treating ADHD and epilepsy, with high-resolution EEG determining ictal sources for epileptic surgery.Among compared methods, ECD shows the highest accuracy relative to MUSIC, LORETA, and sLORETA.
- DISCUSSION AND CONCLUSIONS: Future research should address limitations of EEG-based source localization to resolve problems in applied neuroscience and improve diagnosis and treatment of brain disorders.The passage also notes that psychiatric disorders may be diagnosed and treated through brain source localization methods.