Sensorineural hearing loss is primarily associated with damage to the sensory hair cells of the Organ of Corti, highly specialized mechanosensitive cells characterized by limited regenerative capacity. Understanding how auditory cells respond to controlled acoustic stimulation is crucial for gaining insight into the mechanisms underlying cellular vulnerability and adaptation to mechanical stress. This thesis investigates the effects of in vitro mechanical stimulation on murine auditory cells subjected to four different stimulation conditions (0, 30, 40, and 50 Hz). Following stimulation, samples were fixed, labeled with fluorescent probes, and imaged using fluorescence microscopy. Quantitative parameters regarding cell, nucleus, and cytoskeleton morphology were extracted from the images using a standardized image-processing pipeline that included segmentation and structural analysis. The resulting dataset was analyzed using a pipeline comprising statistical analysis, feature engineering, automated feature selection, data filtering, and a comparison of various supervised machine learning algorithms. The objective was to identify the most informative morphological features and evaluate the models' ability to distinguish between the different mechanical stimulation conditions. The results demonstrated that introducing new derived features and applying selection and filtering techniques improved model performance compared to using only the original morphological characteristics, thereby revealing patterns associated with the different stimulation conditions. This study confirms the potential of integrating quantitative image analysis with machine learning to support the study of cellular responses to acoustic stress and provides a reproducible methodology applicable to image-based biological analyses.
La perdita dell'udito di tipo neurosensoriale è principalmente associata al danneggiamento delle cellule ciliate sensoriali dell'Organo del Corti, cellule meccanosensibili altamente specializzate e caratterizzate da una limitata capacità rigenerativa. Comprendere la risposta delle cellule uditive a stimolazioni acustiche controllate è fondamentale per approfondire i meccanismi coinvolti nella vulnerabilità cellulare e nell'adattamento allo stress meccanico. In questo lavoro di tesi sono stati studiati gli effetti della stimolazione meccanica in vitro su cellule uditive murine, sottoposte a quattro differenti condizioni di stimolazione (0, 30, 40 e 50 Hz). Dopo la stimolazione, i campioni sono stati fissati, marcati con sonde fluorescenti e acquisiti mediante microscopia a fluorescenza. Dalle immagini sono stati estratti parametri quantitativi relativi alla morfologia della cellula, del nucleo e del citoscheletro attraverso una pipeline standardizzata di elaborazione delle immagini comprendente segmentazione e analisi strutturale. Il dataset ottenuto è stato analizzato mediante una pipeline di data analysis composta da analisi statistiche, ingegnerizzazione delle caratteristiche (Feature Engineering), selezione automatica delle feature, filtraggio dei dati e confronto di diversi algoritmi di Machine Learning supervisionato. L'obiettivo è stato quello di individuare le caratteristiche morfologiche maggiormente informative e valutare la capacità dei modelli di distinguere le differenti condizioni di stimolazione meccanica. I risultati hanno mostrato che l'introduzione di nuove feature derivate e l'applicazione di tecniche di selezione e filtraggio hanno migliorato le prestazioni dei modelli rispetto all'utilizzo delle sole caratteristiche morfologiche originali, consentendo di evidenziare pattern associati alle diverse condizioni di stimolazione. Lo studio conferma il potenziale dell'integrazione tra analisi quantitativa delle immagini e Machine Learning come strumento di supporto allo studio delle risposte cellulari allo stress acustico e fornisce una metodologia riproducibile applicabile ad analisi biologiche basate su immagini.
Sviluppo di modelli predittivi basati su Machine Learning per l’analisi di cellule uditive dell'Organo del Corti (OC-k3) sottoposte a stimolazione acustica
BIASIOLO, GIORGIA
2025/2026
Abstract
Sensorineural hearing loss is primarily associated with damage to the sensory hair cells of the Organ of Corti, highly specialized mechanosensitive cells characterized by limited regenerative capacity. Understanding how auditory cells respond to controlled acoustic stimulation is crucial for gaining insight into the mechanisms underlying cellular vulnerability and adaptation to mechanical stress. This thesis investigates the effects of in vitro mechanical stimulation on murine auditory cells subjected to four different stimulation conditions (0, 30, 40, and 50 Hz). Following stimulation, samples were fixed, labeled with fluorescent probes, and imaged using fluorescence microscopy. Quantitative parameters regarding cell, nucleus, and cytoskeleton morphology were extracted from the images using a standardized image-processing pipeline that included segmentation and structural analysis. The resulting dataset was analyzed using a pipeline comprising statistical analysis, feature engineering, automated feature selection, data filtering, and a comparison of various supervised machine learning algorithms. The objective was to identify the most informative morphological features and evaluate the models' ability to distinguish between the different mechanical stimulation conditions. The results demonstrated that introducing new derived features and applying selection and filtering techniques improved model performance compared to using only the original morphological characteristics, thereby revealing patterns associated with the different stimulation conditions. This study confirms the potential of integrating quantitative image analysis with machine learning to support the study of cellular responses to acoustic stress and provides a reproducible methodology applicable to image-based biological analyses.| File | Dimensione | Formato | |
|---|---|---|---|
|
Biasiolo_Giorgia.pdf
Accesso riservato
Dimensione
8.88 MB
Formato
Adobe PDF
|
8.88 MB | Adobe PDF |
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/20.500.12608/116349