Karakterizacija posnetkov EEG avditornih nalog pri osebah s Parkinsovo boleznijo

Štremfelj, Jana (author), Smrdel, Aleš (mentor)

Parkinsonova bolezen je kronična nevrodegenerativna bolezen, ki močno vpliva na možgane in poslabša motorične sposobnosti bolnika. Spremeni se fizična sposobnost bolnika, zmanjša se mišična gibljivost in nastanejo težave s hojo. Pojavijo se tudi težave s spominom, pozornostjo in vidom. Spremembe motoričnih in kognitivnih sposobnosti je mogoče zaznati v signalih električne aktivnosti možganov, zato ima na tem področju velik potencial analiza elektroencefalografskih (EEG) signalov. V tem delu smo se lotili karakterizacije sprememb signalov EEG pri osebah s Parkinsonovo boleznijo in določili elektrode in značilke, pri katerih so te spremembe najizrazitejše. Uporabili smo bazo posnetkov, ki so bili snemani med izvajanjem različnih avditornih nalog (ang. auditory tasks). Posnetke smo razdelili v dve skupini, skupino oseb s Parkinsonovo boleznijo (SPB) in kontrolno skupino zdravih oseb (KTL). Vsak posnetek smo glede na avditorne naloge razdelili na 4-sekundne segmente, iz katerih smo vzeli 2-sekundne intervale in v slednjih opazovali spremembe. Za vsak posnetek smo iz zaporednih intervalov izračunali značilke povprečje razlike intervalov, Manhattanska razdalja, Evklidska razdalja, koren povprečja kvadratov, vrh frekvence, mediana frekvence in vzorčna entropija. Nato smo izrisali in opazovali spremembe časovnih vrst omenjenih značilk ter pri nekaterih ugotovili ponavljajoče se trende, ki ločujejo skupino SPB od skupine KTL. Sledil je izračun značilk za klasifikacijo na podlagi povprečja razlike začetnih in končnih intervalov. Z uporabo Studentovega t-testa smo identificirali elektrode, pri katerih so razlike med obema skupinama najbolj izrazite. Na podlagi teh rezultatov smo določili skupino značilk, ki najbolje odražajo spremembe med obema skupinama. Za posamezne elektrode smo te značilke uporabili za klasifikacijo signalov EEG in prišli do najvišje klasifikacijske točnosti 90 \%. Kot najboljša mesta elektrod za razločevanje skupin smo na podlagi ugotovitev karakterizacije, z uporabo Studentovega t-testa in klasifikacije določili elektrode TP7, F7, FT7, CP5 in FC5.

• Language: Slovenian
• Keywords: digitalno procesiranje signalov, analiza biomedicinskih signalov, EEG, Parkinsonova bolezen, elektroencefalografija
• Work type: Master's thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2024
• PID: 20.500.12556/RUL-164348
• Publication date in RUL: 23.10.2024

Secondary language

• Language: English
• Title: Characterization of EEG Records of Auditory Tasks in Individuals with Parkinson’s Disease

Abstract

Parkinson's disease is a chronic neurodegenerative disorder that severely affects the brain and decreases the patient's motor abilities. Physical capability declines, muscle mobility worsens and walking difficulties emerge. Issues with memory, attention, and vision are also common. Changes in motor and cognitive abilities can be detected through brain electrical activity signals, making electroencephalographic (EEG) signal analysis particularly promising in this field. In this study, we focused on characterizing EEG signal changes in individuals with Parkinson's disease and determining the electrodes and features where these changes are most prominent. We used a database of recordings captured during various auditory tasks. The recordings were divided into two groups: individuals with Parkinson’s disease (SPB) and a control group of healthy individuals (KTL). Each recording was segmented based on the auditory tasks into 4 second segments, from which 2 second intervals were extracted to observe changes in the EEG signals. For each recording, we calculated features from consecutive intervals, such as the average difference between intervals, Manhattan distance, Euclidean distance, root mean square, peak frequency, median frequency and sample entropy. We then visualized and analyzed changes in the time series of these features, identifying recurring trends that distinguished the SPB group from the KTL group. This was followed by calculating classification features based on the average difference between the initial and final intervals. Using Student’s t-test, we identified electrodes, where differences between the two groups were most pronounced. Based on these results, we determined a set of features that best reflect the changes between the two groups. For each electrode, we used these features to classify the EEG signals, achieving the highest classification accuracy of 90 \%. Based on the findings of the signal characterization, supported by the use of Student’s t-test and classification results, we identified electrodes TP7, F7, FT7, CP5, and FC5 as the most effective ones for distinguishing between the two groups.

• Keywords: digital signal processing, biomedical signal processing, EEG, Parkinson's disease, electroencephalography