Analysis and Interpretation of Electroencephalogram (EEG) Signals for an Automated Diagnosis of Earlier Onset of Stroke using a Convolutional Neural Network (CNN)

Abstract

Electroencephalogram (EEG) recording is relatively safe for the patients who are in stroke disease, soi tisoftenusedtodetecttheoccurrenceofstrokeinclinicalpractice.TheobjectiveofthispaperistoapplyDeep Learning methodto EEG signal sanalysisinordertoconfirmtheoccurrenceofstrokeduetoinjuriesattack.AnovelapproachusingpolynomialstransformforspectralanalysisofEEGsignalintheTchebychevbasisisimplementedtoobtainspectralcoefficients.Statisticalmetricswereextractedfromthesepolynomialscoefficientstoconstitudethe input vector to a convolutional neural network. The model perform an accuracy of 98 % showing that the methodcan evaluate the condition of brain attack patients and can be a reliable toll of quasi stroke diagnosis.

Introduction

Stroke is a life-threatening medical emergency caused by either a blockage of blood flow (ischemic stroke) or bleeding in the brain (hemorrhagic stroke), leading to the rapid death of brain cells. It is the second leading cause of death and disability worldwide and can result in paralysis, speech impairment, cognitive deficits, memory loss, or death. Early diagnosis is essential because prompt treatment significantly improves patient outcomes.

Electroencephalography (EEG) has emerged as a promising, non-invasive tool for the early detection, diagnosis, prognosis, and rehabilitation monitoring of stroke. EEG records the brain's electrical activity and can detect abnormal neural oscillations before irreversible brain damage becomes visible through conventional imaging techniques. Stroke patients typically exhibit increased slow-wave activity (delta and theta bands) and decreased fast-wave activity (alpha and beta bands), making EEG biomarkers valuable indicators of ischemic brain injury.

The literature demonstrates significant advances in applying artificial intelligence (AI), machine learning (ML), and deep learning techniques to stroke diagnosis and prediction. Ensemble learning models, Random Forest, XGBoost, LightGBM, Support Vector Machines (SVM), Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and transformer-based models have achieved diagnostic accuracies ranging from approximately 85% to nearly 100%. Feature extraction techniques such as wavelet packet transform, functional connectivity analysis, nonlinear EEG measures, and quantitative EEG (QEEG) parameters further improve classification performance.

The proposed study introduces an automated EEG-based stroke diagnosis framework consisting of four stages: EEG data acquisition, feature extraction, feature selection (dimensionality reduction), and AI-based classification. EEG signals were collected from 50 stroke patients and 109 healthy participants using a six-channel EEG device over five-minute resting-state recordings. Signal preprocessing included amplification, filtering, and spectral analysis based on Chebyshev polynomial expansion to obtain informative EEG features.

The experimental platform utilized an Intel Core i5 computer running Anaconda for signal processing and machine learning implementation. Ethical approval, informed consent, anonymization, and secure storage of participant data were maintained throughout the study.

Conclusion

We propose a new health diagnosis sytem that de- tecttheoccurrenceofstrokediseaseswiththeattribute informationofstatisticalmetricextractedfromDicrete Tchebychev Transform (DChT) of raw EEG signal col- lectedfromstrokepatientsandhealthycontrols.Spec- tralanalysis,Extractionofdiscriminativeparameters,Di-mensionality reduction and Classification were performedthrough Tchebychev polynomials transform, Linear dis-criminantanalysisandaoneDimensionalConvolutional Neural Network (1D-CNN). Above all, the health diag- nosissysteminthisstudycandetectandpredicttheoc- currenceofstroke,thusprovidingaccuratepredictictionresultsinasystemthatcanbeimplementedatlowcost. Asaresult,thesystemhasgreatadvantagesasitcanprovideindepthanalysisinformationusefulforthebehavioralcharacteristicsbeforeandduringstrokeattack.In ordertoamelioratethisworkitisnecessaryforfurther studiestointegratesimalaranalysiswithECGandEMG signals.

References

[1] B. Foreman, J. Claassen, K. Abou Khaled, J. Jirsch, D. M.Alschuler, J. Wittman, R. G. Emerson, and L. J. Hirsch,“Generalized periodic discharges in the critically ill: acase-control study of 200 patients,” Neurology, vol. 79,no. 19, pp. 1951–1960, 2012. [2] B. Ovbiagele, L. B. Goldstein, R. T. Higashida, V.J. Howard, S. C. Johnston, O. A. Khavjou, D. T. Lackland, [3] J. H. Lichtman, S. Mohl, R. L. Sacco et al., “Forecastingthefutureofstrokeintheunitedstates:apolicystatementfrom the american heart association and americanstrokeassociation,”Stroke,vol.44,no.8,pp.2361–2375,2013. [4] C.O.Johnson,M.Nguyen,G.A.Roth,E.Nichols,T. Alam, D. Abate, F. Abd-Allah, A. Abdelalim, H. N.Abraha, N.M. Abu-Rmeilehetal., “Global,regional, andnationalburdenofstroke,1990–2016:asystematicanaly-sisfortheglobalburdenofdiseasestudy2016,”TheLancet Neurology, vol. 18, no. 5, pp. 439–458, 2019. [5] H.Jokinen,S.Melkas,R.Ylikoski,T.Pohjasvaara,M. Kaste, T. Erkinjuntti, and M. Hietanen, “Post-strokecognitive impairment is common even after successful clin-ical recovery,” European journal of neurology, vol. 22,no. 9, pp. 1288–1294, 2015. [6] J.Petrovic,V.Milosevic,M.Zivkovic,D.Stojanov,O. Milojkovic, A. Kalauzi, and J. Saponjic, “Slower eegalphageneration,synchronizationand“flow”—possiblebiomarkers of cognitive impairment and neuropathology ofminor stroke,” PeerJ, vol. 5, p. e3839, 2017. [7] T.Aydin,F.N.Kesiktas¸,M.M.Oren,T.Erdogan,Y.C. Ahisha,T.Kizilkurt,M.Corum,?I.Karacan,S.Öztürk, andG.Bahat,“Sarcopeniainpatientsfollowingstroke:Anoverlookedproblem,”InternationalJournalofRehabilita- tionResearch,vol.44,no.3,pp.269–275,2021. [8] N.Wright,K.Newell,K.H.Chan,S.Gilbert,A.Hacker,Y. Lu, Y. Guo, P. Pei, C. Yu, J. Lv etal., “Long-term ambi-ent air pollution exposure and cardio-respiratory disease inchina:findingsfromaprospectivecohortstudy,”Environ-mental Health, vol. 22, no. 1, p. 30, 2023. [9] S.Benghanem,N.Kubis,E.Gayat,A.Loiodice,E. Pruvost-Robieux, T. Sharshar, A. Foucrier, S. Figueiredo, V. Bouilleret, E. De Montmollin et al.,“Prognostic value of early eeg abnormalities in severestrokepatientsrequiringmechanicalventilation:apre-planned analysis of the spice prospective multicenterstudy,” Critical Care, vol. 28, no. 1, p. 173, 2024. [10] Y.Jin,J.Li,Z.Fan,X.Hua,T.Wang,S.Du,X.Xi,andL. Li, “Recognition of regions of stroke injury using multi-modalfrequencyfeaturesofelectroencephalogram,”Fron- tiers in Neuroscience, vol. 18, p. 1404816, 2024. [11] Y.Hu,Y.Wang,R.Zhang,Y.Hu,M.Fang,Z.Li,L.Shi,Y. Zhang, Z. Zhang, J. Gao et al., “Assessing stroke re-habilitationdegreebasedonquantitativeeegindexandnonlinearparameters,”CognitiveNeurodynamics,vol.17,no. 3, pp. 661–669, 2023. [12] L. Sutcliffe, H. Lumley, L. Shaw, R. Francis, and C. I.Price,“Surfaceelectroencephalography(eeg)duringtheacute phase of stroke to assist with diagnosis and predictionofprognosis:ascopingreview,”BMCemergencymedicine,vol. 22, no. 1, p. 29, 2022. [13] M.S.N.AgLamat,M.S.H.AbdRahman,W.A. WanZaidi,W.N.N.W.Yahya,C.S.Khoo,R.Hod,andH.J.Tan,“Qualitativeelectroencephalogramanditspre-dictorsinthediagnosisofstroke,”FrontiersinNeurology,vol. 14, p. 1118903, 2023. [14] Z. Keser, S. C. Buchl, N. A. Seven, M. Markota, H. M.Clark, D. T. Jones, G. Lanzino, R. D. Brown Jr, G. A. Wor-rell,andB.N.Lundstrom,“Electroencephalogram(eeg)with or without transcranial magnetic stimulation (tms) asbiomarkersforpost-strokerecovery:anarrativereview,”Frontiers in neurology, vol. 13, p. 827866, 2022. [15] A. L. Schneider and K. G. Jordan, “Regional attenuationwithout delta (rawod): a distinctive eeg pattern that can aidinthediagnosisandmanagementofsevereacuteischemicstroke,”Americanjournalofelectroneurodiagnostictech- nology, vol. 45, no. 2, pp. 102–117, 2005. [16] B.ForemanandJ.Claassen,“Quantitativeeegforthede-tectionofbrainischemia,”Criticalcare,vol.16,pp.1–9,2012. [17] S.FinniganandM.J.VanPutten,“Eeginischaemicstroke:quantitativeeegcanuniquelyinform(sub-)acuteprog-nosesandclinicalmanagement,” Clinicalneurophysiology,vol. 124, no. 1, pp. 10–19, 2013. [18] S. Finnigan, A. Wong, and S. Read, “Defining abnormalslow eeg activity in acute ischaemic stroke: Delta/alpha ra-tioasanoptimalqeegindex,”ClinicalNeurophysiology,vol. 127, no. 2, pp. 1452–1459, 2016. [19] R.A.Chowdhury,G.Pellegrino,Ü.Aydin,J.-M.Lina,F.Dubeau,E.Kobayashi,andC.Grova,“Reproducibilityofeeg-megfusionsourceanalysisofinterictalspikes:Rel-evanceinpresurgicalevaluationofepilepsy,”Humanbrain mapping, vol. 39, no. 2, pp. 880–901, 2018. [20] F.DeVicoFallani,L.Astolfi,F.Cincotti,D.Mattia,D. laRocca,E.Maksuti,S.Salinari,F.Babiloni,B.Vegso,G.Kozmannetal.,“Evaluationofthebrainnetworkorga-nizationfromeegsignals:apreliminaryevidenceinstrokepatient,”TheAnatomicalRecord:AdvancesinIntegrative Anatomy and Evolutionary Biology: Advances in Integra- tiveAnatomyandEvolutionaryBiology,vol.292,no.12,pp.2023–2031,2009. [21] Z. Ip, G. Rabiller, J. He, Z. Yao, Y. Akamatsu, Y. Nishi-jima, J. Liu, and A. Yazdan-Shahmorad, “Cortical strokeaffectsactivityandstabilityoftheta/deltastatesinremotehippocampalregions,”in201941stAnnualInternational Conference of the IEEE Engineering in Medicine and Bi- ology Society (EMBC).IEEE, 2019, pp. 5225–5228. [22] M. W. M. G. A. J. SP Finnigan, SE Rose, “Correlation ofquantitative eeg in acute ischemic stroke with 30-day nihssscore:comparisonwithdiffusionandperfusionmri,”p.5,2019. [23] J. J. C. H. C. P. H. L. YA Choi, S Park, “Machine-learning-basedelderlystrokemonitoringsystemusingelectroen-cephalography vital signals,” p. 5, 2023. [24] X.L.H.Liu,“Eegdatasetsofstrokepatients,”2023. [25] G.M.D.H.T.B.N.W.J.Schalk,“Eegmotormovement,”2009. [26] D.d’HelsinkiLaennec,“Principeséthiquesapplicablesauxrecherchesmédicalessurdessujetshumains,”2002. [27] D.Tchiotsop,“Modélisationspolynomialesdessignauxecg.applicationàlacompression.”2007. [28] B.H.etA.LOUAIL,“Applicationsdespolynômesorthog-onauxpourrésolutiondeséquationsintégrales,”2021. [29] V.Rossetto,“Outilsnumériques,”2021. [30] A. Subasi and M. I. Gursoy, “Eeg signal classification usingpca, ica, lda and support vector machines,” Expertsystems withapplications, vol. 37, no. 12, pp. 8659–8666, 2010. [31] K.Fukunaga,Introductiontostatisticalpatternrecogni- tion.Elsevier, 2013. [32] L. C. D. Nkengfack, D. Tchiotsop, R. Atangana, V. Louis-Door, and D. Wolf, “Eeg signals analysis for epilepticseizures detection using polynomial transforms, linear dis-criminantanalysisandsupportvectormachines,”Biomed- ical Signal Processing and Control, vol. 62, p. 102141,2020. [33] R. C. González, “Towards an automated portableelectroencephalography-basedsystemforalzheimer’sdis-easediagnosis,”Ph.D.dissertation,InstitutNationaldelaRecherche Scientifique (Canada), 2018. [34] R.Chetana,A.S.Rao,andK.Mahantesh,“Applicationofconv-1dandbi-lstmtoclassifyanddetectepilepsyineegdata,” International Journal of Advanced Computer Sci- ence and Applications, vol. 14, no. 6, 2023.

Copyright

Copyright © 2026 Romain Atangana, Danwe Tsobsala, Amstrong Emini Me Zenanga, Daniel Gams Massi, Daniel Tchiotsop, Godpromesse Kenne. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.