Otologic disorders pose a major global health challenge, often leading to hearing impairment that significantly affects communication, cognitive development, and quality of life. Early diagnosis and timely intervention are essential to prevent permanent auditory deficits, particularly in pediatric and elderly populations. However, conventional diagnostic tools such as tympanometry and Wideband Acoustic Immittance (WAI) require the application of external pressure and specialized expertise, limiting their use in large-scale or primary-care screening. To address these limitations, this thesis investigates and characterizes a novel diagnostic technology called Pressure-Less Acoustic Immittance (PLAI)TM, which allows middle ear assessment under ambient pressure conditions, providing a fully non-invasive alternative to traditional methods. Chapter 2 focuses on the physiological characterization of PLAITM parameters and the definition of age-specific reference ranges. A multicentric clinical study involving 218 subjects (360 healthy ears and 76 pathological ears) aged 4 months to 80 years was conducted across six Italian hospitals. Statistical analysis revealed a logarithmic dependence of PLAITM parameters on age, reflecting developmental changes in middle ear anatomy. Based on these trends, three age groups (0–3, 3–12, and over 12 years) were defined, and normative reference intervals were established. Comparisons with Otitis Media with Effusion (OME) cases showed significant diagnostic sensitivity, with key differences in resonance frequency, bandwidth limits, and equivalent volume, confirming the physiological validity of PLAITM. Building on these findings, Chapter 3 explores the diagnostic potential of PLAITM through machine learning–based classification models. Random Forest algorithms trained within each age group achieved macro F1-scores above 0.78, with the best performance (0.84) in children aged 3–12 years. Feature relevance analysis using SHapley Additive exPlanations (SHAP) identified resonance frequency, peak admittance, and canal volume as the most influential predictors, aligning with physiological expectations. Extending beyond diagnostic classification, Chapter 4 presents the development of a lumped element ear model derived from PLAITM data and a physical phantom. These models replicate the mechanical and acoustic behavior of the ear, enabling simulation of both healthy and pathological conditions, including experimentally induced variations. The modeling framework provides a quantitative bridge between measured admittance parameters and underlying biomechanical properties, offering a foundation for advanced diagnostic interpretation and personalized assessment. Overall, the results demonstrate the clinical promise of PLAITM as a reliable, pressure-free, and age-adaptable technique for middle ear evaluation. Its ability to detect functional and pathological changes without external pressure or specialized operators positions it as a valuable tool for early screening and monitoring of otologic disorders. Moreover, the integration of machine learning and physical ear modeling establishes the groundwork for automated, data-driven diagnostic support systems, enhancing accessibility to accurate and non-invasive hearing assessments in both clinical and community healthcare settings.
Le patologie otologiche rappresentano una rilevante sfida sanitaria globale, poichè spesso conducono a deficit uditivi che compromettono in modo significativo la comunicazione, lo sviluppo cognitivo e la qualità della vita. La diagnosi precoce e l’intervento tempestivo sono quindi fondamentali per prevenire danni permanenti, in particolare nelle popolazioni pediatriche e geriatriche. Tuttavia, gli strumenti diagnostici convenzionali, come la timpanometria e l’immittanza acustica a banda larga (WAI), richiedono l’applicazione di pressione esterna e competenze specialistiche, limitandone l’impiego in programmi di screening su larga scala o in contesti di medicina di base. Per superare tali limiti, questa tesi analizza e caratterizza una nuova tecnologia diagnostica denominata Pressure-Less Acoustic Immittance (PLAITM), che consente la valutazione dell’orecchio medio in condizioni di pressione ambientale, offrendo una reale alternativa non invasiva ai metodi tradizionali. Il Capitolo 2 si concentra sulla caratterizzazione fisiologica dei parametri della tecnologia PLAITM e sulla definizione di intervalli di riferimento specifici per età. `E stato condotto uno studio clinico multicentrico che ha coinvolto 218 soggetti (360 orecchie sane e 76 patologiche) di età compresa tra 4 mesi e 80 anni, arruolati in sei ospedali italiani. L’analisi statistica ha evidenziato una dipendenza logaritmica dei parametri PLAITM dall’età, riflettendo le variazioni anatomiche e meccaniche dell’orecchio medio durante la crescita. Sulla base di tali tendenze, la popolazione `e stata suddivisa in tre gruppi d’età (0–3, 3–12 e oltre 12 anni) e per ciascuno sono stati definiti intervalli di riferimento normativi. Il confronto con casi di Otite Media con Effusione (OME) ha mostrato una significativa sensibilità diagnostica, con differenze nei parametri di frequenza di risonanza, limiti di banda e volume equivalente, confermando la validità fisiologica della tecnologia. Sulla base di questi risultati, il Capitolo 3 esplora il potenziale diagnostico dei parametri PLAITM mediante modelli di classificazione basati su tecniche di machine learning. Algoritmi Random Forest, addestrati per ciascun gruppo d’età, hanno raggiunto macro F1-score superiori a 0.78, con prestazioni massime (0.84) nei bambini tra 3 e 12 anni. L’analisi di interpretabilità condotta tramite SHapley Additive exPlanations (SHAP) ha identificato la frequenza di risonanza, il picco di ammettenza e il volume del condotto uditivo come predittori più influenti, in linea con le aspettative fisiologiche. Estendendo l’analisi oltre la classificazione diagnostica, il Capitolo 4 presenta lo sviluppo di modelli fisici e a elementi concentrati dell’orecchio derivati dai dati PLAITM. Tali modelli riproducono il comportamento meccanico e acustico dell’orecchio, consentendo la simulazione di condizioni sia fisiologiche sia patologiche, incluse variazioni sperimentali indotte. Questo approccio modellistico fornisce un collegamento quantitativo tra i parametri di ammettenza misurati e le proprietà biomeccaniche sottostanti, ponendo solide basi verso un’interpretazione diagnostica avanzata e personalizzata. Complessivamente, i risultati dimostrano il potenziale clinico di PLAITM come metodo affidabile, atraumatico e adattabile all’età per la valutazione dell’orecchio medio. La capacità di individuare variazioni funzionali e patologiche senza l’uso di pressioni esterne nè di personale specializzato ne fa uno strumento promettente per lo screening precoce e il monitoraggio delle patologie otologiche. Inoltre, l’integrazione di tecniche di machine learning e di modelli fisici dell’orecchio getta le basi per lo sviluppo di sistemi di supporto diagnostico automatici, migliorando l’accessibilità a valutazioni uditive accurate e non invasive sia in ambito clinico sia in contesti di assistenza territoriale.
Valutazione e modellizzazione della funzionalità uditiva tramite un metodo innovativo di misura dell’immittanza acustica senza pressione
BASSI, FRANCESCO
2026
Abstract
Otologic disorders pose a major global health challenge, often leading to hearing impairment that significantly affects communication, cognitive development, and quality of life. Early diagnosis and timely intervention are essential to prevent permanent auditory deficits, particularly in pediatric and elderly populations. However, conventional diagnostic tools such as tympanometry and Wideband Acoustic Immittance (WAI) require the application of external pressure and specialized expertise, limiting their use in large-scale or primary-care screening. To address these limitations, this thesis investigates and characterizes a novel diagnostic technology called Pressure-Less Acoustic Immittance (PLAI)TM, which allows middle ear assessment under ambient pressure conditions, providing a fully non-invasive alternative to traditional methods. Chapter 2 focuses on the physiological characterization of PLAITM parameters and the definition of age-specific reference ranges. A multicentric clinical study involving 218 subjects (360 healthy ears and 76 pathological ears) aged 4 months to 80 years was conducted across six Italian hospitals. Statistical analysis revealed a logarithmic dependence of PLAITM parameters on age, reflecting developmental changes in middle ear anatomy. Based on these trends, three age groups (0–3, 3–12, and over 12 years) were defined, and normative reference intervals were established. Comparisons with Otitis Media with Effusion (OME) cases showed significant diagnostic sensitivity, with key differences in resonance frequency, bandwidth limits, and equivalent volume, confirming the physiological validity of PLAITM. Building on these findings, Chapter 3 explores the diagnostic potential of PLAITM through machine learning–based classification models. Random Forest algorithms trained within each age group achieved macro F1-scores above 0.78, with the best performance (0.84) in children aged 3–12 years. Feature relevance analysis using SHapley Additive exPlanations (SHAP) identified resonance frequency, peak admittance, and canal volume as the most influential predictors, aligning with physiological expectations. Extending beyond diagnostic classification, Chapter 4 presents the development of a lumped element ear model derived from PLAITM data and a physical phantom. These models replicate the mechanical and acoustic behavior of the ear, enabling simulation of both healthy and pathological conditions, including experimentally induced variations. The modeling framework provides a quantitative bridge between measured admittance parameters and underlying biomechanical properties, offering a foundation for advanced diagnostic interpretation and personalized assessment. Overall, the results demonstrate the clinical promise of PLAITM as a reliable, pressure-free, and age-adaptable technique for middle ear evaluation. Its ability to detect functional and pathological changes without external pressure or specialized operators positions it as a valuable tool for early screening and monitoring of otologic disorders. Moreover, the integration of machine learning and physical ear modeling establishes the groundwork for automated, data-driven diagnostic support systems, enhancing accessibility to accurate and non-invasive hearing assessments in both clinical and community healthcare settings.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_dottorato_Bassi_v2_fin.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
4.93 MB
Formato
Adobe PDF
|
4.93 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/364646
URN:NBN:IT:UNITS-364646