Modeliranje globine anestezije z uporabo indeksa EEG

ZALOŽNIK, MARCEL (author), Karer, Gorazd (mentor)

V magistrski nalogi se osredotočamo na razvoj in vrednotenje modelov za ocenjevanje in vodenje globine anestezije s pomočjo procesiranih elektroencefalografskih (EEG) signalov. Namen raziskave je izboljšati natančnost in učinkovitost vodenja anestezije med kirurškimi posegi, kar je ključnega pomena za varnost in udobje pacienta. V nalogi so najprej predstavljeni farmakokinetični in farmakodinamični modeli za anestezijo, ki opisujejo absorpcijo, distribucijo, metabolizem in izločanje anestetikov ter njihov vpliv na telo. Posebna pozornost je namenjena primerjavi uveljavljenih modelov za propofol in remifentanil. Nato so obravnavani različni indeksi na osnovi procesiranih signalov EEG za merjenje globine anestezije, predvsem indeks BIS (bispektralni indeks) in indeks PSi (indeks stanja pacienta). Obe metodi temeljita na analizi signalov EEG in omogočata oceno stanja pacienta med anestezijo. V okviru raziskave je bil razvit simulacijski model, ki omogoča primerjavo obeh indeksov. V nadaljevanju naloge so predstavljeni novi populacijski modeli na podlagi izmerjenih podatkov, ki bolje prilagajajo globino anestezije izbrani populaciji. Predstavljena je tudi uporaba metod strojnega učenja, kot so nevronske mreže in odločitvena drevesa, za napovedovanje optimalnih parametrov modelov za posamezne paciente. Nato razširimo analizo na vpliv analgetika remifentanila na globino anestezije in predstavimo modificiran model, ki upošteva učinke kombinacije propofola in remifentanila. Rezultati simulacij kažejo izboljšave pri napovedovanju globine anestezije. Nato predstavimo razvoj residualnega modela, ki prilagaja napovedi na podlagi individualnih razlik pacientov. Ta model omogoča sprotno prilagajanje med operacijo, kar izboljša natančnost napovedi in prispeva k varnejši uporabi anestezije. Rezultati raziskave kažejo, da uporaba izboljšanih farmakokinetičnih in farmakodinamičnih modelov lahko bistveno pripomore k natančnejšemu modeliranju globine anestezije. To vodi k zmanjšanju tveganj za zaplete med operacijo in hitrejšemu okrevanju pacientov po posegu.

• Language: Slovenian
• Keywords: anestezija, EEG, farmakokinetični model, farmakodinamični model, strojno učenje, indeks BIS, indeks PSi
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FE - Faculty of Electrical Engineering
• Year: 2024
• PID: 20.500.12556/RUL-161380
• COBISS.SI-ID: 215241219
• Publication date in RUL: 10.09.2024

Secondary language

• Language: English
• Title: Modelling the depth of anaesthesia using an EEG index

Abstract

In this master's thesis, we focus on the development and evaluation of models for assessing and managing the depth of anesthesia using processed electroencephalographic (EEG) signals. The aim of the research is to improve the accuracy and efficiency of anesthesia management during surgical procedures, which is crucial for patient safety and comfort. The thesis first presents pharmacokinetic and pharmacodynamic models for anesthesia, which describe the absorption, distribution, metabolism, and excretion of anesthetics, as well as their effects on the body. Special attention is given to the comparison of established models for propofol and remifentanil. Next, various indices based on processed EEG signals for measuring the depth of anesthesia are discussed, particularly the BIS (Bispectral Index) and PSi (Patient State Index). Both methods are based on EEG signal analysis and allow for the assessment of the patient's state during anesthesia. A simulation model was developed as part of the research, enabling a comparison of both indices. The thesis then introduces new population models based on measured data that better adjust the depth of anesthesia for the selected population. The use of machine learning methods, such as neural networks and decision trees, for predicting optimal model parameters for individual patients is also presented. The analysis is further extended to include the effect of the analgesic remifentanil on the depth of anesthesia, and a modified model that accounts for the combined effects of propofol and remifentanil is introduced. The simulation results show improvements in predicting the depth of anesthesia. A residual model is then developed, which adjusts predictions based on individual patient differences. This model allows for real-time adjustments during surgery, improving prediction accuracy and contributing to the safer use of anesthesia. The research results indicate that the use of enhanced pharmacokinetic and pharmacodynamic models can significantly contribute to more accurate modeling of anesthesia depth. This leads to a reduction in the risk of complications during surgery and a faster recovery for patients after the procedure.

• Keywords: anesthesia, EEG, pharmacokinetic model, pharmacodynamic model, machine learning, BIS index, PSi index