Genitsaridi, Eleni
(2021)
Novel approaches for tinnitus subphenotyping: evidence synthesis, standardised assessment, and supervised and unsupervised machine learning applications.
PhD thesis, University of Nottingham.
Abstract
Clinical management of tinnitus is rather challenging and there is yet no cure for most tinnitus cases. It is speculated that tinnitus heterogeneity is hindering progress in scientific understanding and development of treatments. Phenotyping (i.e., assessment of observable characteristics) and subphenotyping (i.e., subgrouping based on differences in observable characteristics) are important for studying heterogeneous conditions like tinnitus. Identifying and defining clinically relevant tinnitus subphenotypes could help achieve transformational advances in the field. This dissertation reports the application of several advanced methodological approaches and has two main aims. The first aim is to contribute to an international standardisation of tinnitus assessment relevant for tinnitus phenotypic profiling and subphenotyping. The second aim is to further our understanding of tinnitus heterogeneity by investigating the presence of robust subphenotypes, consistent across multiple independent datasets.
Two chapters focus on the first aim. Chapter 2 reviews the literature, summarises current knowledge on tinnitus subphenotypes and identifies research gaps. It also summarises methods used so far and presents a novel framework of variable concepts that have been used for tinnitus subphenotyping. Chapter 3 describes the development of a self-report questionnaire intended to be used as a standard for tinnitus phenotyping. This questionnaire was developed through an international collaboration with tinnitus researchers from many centres. The questionnaire is already translated into 9 languages (Albanian, Dutch, French, German, Greek, Italian, Polish, Spanish, and Swedish) and is being used by multiple research teams as a tool for standardised tinnitus assessment.
The second aim is addressed in Chapters 4 and 5. Chapter 4 provides a detailed description of three tinnitus-specific datasets that were subsequently analysed in Chapter 5, and highlights commonalities and differences in the studied populations and the collected variables. Chapter 5 describes a novel data-driven approach for discovering tinnitus subphenotypes. This Chapter reports on a comprehensive unsupervised machine learning methodology applied to the three datasets. Findings indicate that this method was able to identify robust tinnitus subphenotypic patterns.
Finally, Chapter 6 relates the overall findings to the wider context of the published literature and presents suggestions and recommendations for future research. Age, sex, hearing ability, problems with sounds, symptoms of depression, and mandible problems were highlighted as important variables for tinnitus subphenotyping and should be considered for assessment in future tinnitus studies. Overall, this work provides a basis for standardised tinnitus assessment in future studies and gives novel insights into the characteristics of tinnitus subphenotypes.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
