RT info:eu-repo/semantics/doctoralThesis
T1 Desarrollo y validación de un asistente virtual basado en inteligencia artificial para apoyar la toma de decisión en atención visual primaria.
A1 Stuermer, Leandro
A2 Universidad de Valladolid. Escuela de Doctorado
K1 Optometría
K1 Virtual assistant
K1 Asistente virtual
K1 Artificial intelligence
K1 Inteligencia artificial
K1 Digital health
K1 Salud digital
K1 Optometry
K1 Optometría
K1 32 Ciencias Médicas
AB Primary visual care (PVC) is an essential component for the early detection and timely management of ocular conditions, particularly those that are preventable or treatable with appropriate interventions. In this context, digital technologies have emerged as a strategic opportunity to transform care models, enhancing accessibility, efficiency, and service quality. Digital health, through tools such as mHealth and telehealth, has enabled the integration of solutions across different stages of care, from prevention to follow-up. Artificial intelligence (AI), in particular, has demonstrated significant potential in analyzing large volumes of clinical data, identifying complex patterns, and generating clinical recommendations. While AI has achieved remarkable progress in specialized areas such as ophthalmology, its incorporation into PVC and optometric settings remains limited. This gap motivated the present research, aimed at developing and validating a virtual assistant integrating AI models to support clinical decision-making in optometry.The research was conducted in several phases. Initially, a characterization of digital technologies applied to visual health was carried out through an exploratory review of the scientific literature, revealing a predominance of solutions focused on the retina and diagnostics at secondary and tertiary levels of care, leaving a significant space for innovation in primary care. In parallel, the availability and characteristics of clinical datasets in vision sciences were explored, identifying the absence of specific repositories for optometry and the limited representation of data applicable to primary care contexts. In response to these gaps, a conceptual model of a big data platform for optometry (DAVIH) was designed, conceived to structure, store, and share clinical data in an ethical, interoperable, and secure manner.The central phase focused on the technological development and validation of the virtual assistant. Predictive AI models were generated and trained with real clinical data from patients attended in optometry, using machine learning algorithms and data balancing techniques to optimize the accuracy and robustness of predictions. These models enabled case classification, localization of ocular problems, and detection of binocular vision dysfunctions. Subsequently, the models were integrated into a functional prototype of a virtual assistant in the form of a web application (VICHI), developed with a hybrid client-server architecture and cloud support. The application, designed to be multilingual and accessible from different devices, included modules for model selection, clinical data entry, and presentation of results with reliability indicators.Finally, the influence of the assistant on decision-making was validated through a study involving visual health professionals from different countries. The comparison between an intervention group, exposed to the assistant’s recommendations, and a control group revealed improvements in diagnostic concordance and a reduction in omissions in clinical responses. Additionally, users’ perceptions of the utility and trust in this type of tool were evaluated, showing general positive acceptance.The findings of the thesis highlight the potential of integrating AI-based solutions into PVC, providing a complementary resource to optimize clinical practice and support the training of professionals. Challenges remain regarding data standardization, interoperability, and the integration of digital technologies in resource-limited contexts. Future lines of work include expanding the models to new visual conditions and developing advanced versions of the assistant with natural language processing, consolidating its applicability in both clinical and educational settings.
YR 2026
FD 2026
LK https://uvadoc.uva.es/handle/10324/83262
UL https://uvadoc.uva.es/handle/10324/83262
LA spa
NO Escuela de Doctorado
DS UVaDOC
RD 02-mar-2026