Electrocochleography is an electrophysiological technique used to record potentials generated in the inner ear and cochlear nerve in response to sound stimuli. This method is useful in the diagnosis and monitoring of various pathologies related to the auditory system, such as auditory neuropathy and Ménière's disease. The electrocleographic signal consists of three components: the microphonic potential (CM), generated by the outer hair cells; the compound action potential (CAP), produced by the cochlear nerve; and the summation potential (SP), generated by the inner hair cells. These components cannot be measured individually in a direct manner: only their sum can be detected. Specific assumptions and techniques are required to separate the individual signals. The aim of this thesis is to present an overview of the methods in the literature and to implement a new method for separating the components of the ECochG signal. The techniques in the literature include the cancellation method, Bayesian filtering, frequency filtering. In the thesis, the development of a proposal for a new method is continued, presented in {cite{tesiLara}, in which the signal is studied in its time-frequency representation using the wavelet transform and automated masks are then designed for the reconstruction of the CM and CAP components (in the first instance also including SP). A dataset of simulated signals is created at different click intensities, both positive and negative. For each intensity, a compression signal and a rarefaction signal are obtained. The main difference between these two signals is that the CM component is in phase opposition, a characteristic on which the cancellation method is based. This analysis is carried out in order to overcome certain limitations present, including the independent analysis of the compression and rarefaction signal in order to decrease the number of sweeps and create an objective and easily reproducible method that does not require the intervention of an expert user to determine the mask parameters. A further difficulty in analysing the ECochG signal is the overlapping of SPs with CAPs, which makes it difficult to distinguish between them. A further aim of this thesis is therefore to propose a method for analysing the two components CAP and SP, extracting values that are important in the clinical context such as, peak value and latency. The proposed methods were tested on the simulated dataset, obtaining good results with the method based on the automated wavelet. Subsequently, the analysis was extended to real data from 8 different patient acquisitions. The results, compared with those of the deletion method, showed some limitations in the correct recognition of the time-frequency areas of the CM and CAP components.
L'Elettrococleografia è una tecnica elettrofisiologica utilizzata per registrare i potenziali generati nell'orecchio interno e dal nervo cocleare in risposta a stimoli sonori. Questa metodica è utile nella diagnosi e nel monitoraggio di varie patologie legate al sistema uditivo, come la neuropatia uditiva e la malattia di Ménière. Il segnale elettrococleografico è formato da tre componenti: il potenziale microfonico (CM), generato dalle cellule ciliate esterne; il potenziale d'azione composto (CAP), prodotto dal nervo cocleare; e il potenziale di sommazione (SP), generato dalle cellule ciliate interne. Tali componenti non sono misurabili individualmente in modo diretto: è possibile rilevare solo la loro somma. Per ottenere la separazione dei singoli segnali sono necessarie specifiche ipotesi e tecniche di separazione delle componenti. L'obiettivo di questa tesi è di presentare una panoramica dei metodi presenti in letteratura e implementare un nuovo metodo per separare le componenti del segnale ECochG. Le tecniche presenti in letteratura includono il metodo di cancellazione, filtraggio bayesiano, filtraggio in frequenza. Nella tesi si prosegue lo sviluppo di una proposta di nuova metodica, presentata in \cite{tesiLara} in cui si studia il segnale nella sua rappresentazione in tempo-frequenza grazie all'utilizzo della trasformata wavelet e si progettano poi maschere automatizzate per la ricostruzione delle componenti CM e CAP (in prima battuta comprendente anche di SP). Viene creato un dataset di segnali simulati a diverse intensità di click, sia positivi che negativi. Per ciascuna intensità, si ottiene un segnale di compressione e uno di rarefazione. La principale differenza tra questi due segnali è che la componente CM risulta in opposizione di fase, caratteristica su cui si basa il metodo di cancellazione. Si svolge quest'analisi per riuscire a superare alcune limitazioni presenti, tra cui l'analisi indipendente del segnale di compressione e rarefazione in modo tale da diminuire il numero di sweep e creare un metodo oggettivo e facilmente riproducibile che non richieda l'intervento di un utente esperto per determinare i parametri delle maschere. Un'ulteriore difficoltà nell'analisi del segnale ECochG è la sovrapposizione di SP con CAP, che rende difficile distinguerli. Un ulteriore obbiettivo di questa tesi è quindi proporre un metodo per analizzare le due componenti CAP e SP, estraendo valori importanti nell'ambito clinico quali, valore di picco e latenza. I metodi proposti sono stati testati su il dataset simulato, ottenendo buoni risultati con il metodo basato sulla wavelet automatizzata. Successivamente, l'analisi è stata estesa a dati reali, provenienti da 8 acquisizioni di pazienti differenti. I risultati, confrontati con quelli del metodo di cancellazione, hanno evidenziato alcune limitazioni nel riconoscimento corretto delle aree tempo-frequenza delle componenti CM e CAP.
Caratterizzazione delle componenti CM, CAP e SP del segnale elettrococleografico e metodiche per la loro stima da misure sperimentali
STEFANI, NOEMI
2024/2025
Abstract
Electrocochleography is an electrophysiological technique used to record potentials generated in the inner ear and cochlear nerve in response to sound stimuli. This method is useful in the diagnosis and monitoring of various pathologies related to the auditory system, such as auditory neuropathy and Ménière's disease. The electrocleographic signal consists of three components: the microphonic potential (CM), generated by the outer hair cells; the compound action potential (CAP), produced by the cochlear nerve; and the summation potential (SP), generated by the inner hair cells. These components cannot be measured individually in a direct manner: only their sum can be detected. Specific assumptions and techniques are required to separate the individual signals. The aim of this thesis is to present an overview of the methods in the literature and to implement a new method for separating the components of the ECochG signal. The techniques in the literature include the cancellation method, Bayesian filtering, frequency filtering. In the thesis, the development of a proposal for a new method is continued, presented in {cite{tesiLara}, in which the signal is studied in its time-frequency representation using the wavelet transform and automated masks are then designed for the reconstruction of the CM and CAP components (in the first instance also including SP). A dataset of simulated signals is created at different click intensities, both positive and negative. For each intensity, a compression signal and a rarefaction signal are obtained. The main difference between these two signals is that the CM component is in phase opposition, a characteristic on which the cancellation method is based. This analysis is carried out in order to overcome certain limitations present, including the independent analysis of the compression and rarefaction signal in order to decrease the number of sweeps and create an objective and easily reproducible method that does not require the intervention of an expert user to determine the mask parameters. A further difficulty in analysing the ECochG signal is the overlapping of SPs with CAPs, which makes it difficult to distinguish between them. A further aim of this thesis is therefore to propose a method for analysing the two components CAP and SP, extracting values that are important in the clinical context such as, peak value and latency. The proposed methods were tested on the simulated dataset, obtaining good results with the method based on the automated wavelet. Subsequently, the analysis was extended to real data from 8 different patient acquisitions. The results, compared with those of the deletion method, showed some limitations in the correct recognition of the time-frequency areas of the CM and CAP components.File | Dimensione | Formato | |
---|---|---|---|
Stefani_Noemi.pdf
accesso aperto
Dimensione
23.24 MB
Formato
Adobe PDF
|
23.24 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/85250