Síntesis empírica de meta-análisis, técnicas de optimización multi-objetivo y minería de datos aplicadas al diagnóstico de la Enfermedad de Alzheimer

Resumo

In recent years, there have been important advances in scientific research on AD. Section 1 of this thesis, specific estimates of the effect sizes of the association between different modifiable risk factors and AD are provided using meta-analytic techniques. Section 2 deals with secondary prevention aimed at the early diagnosis of AD. Therefore, a specific methodology is described to develop a model generator system from medical diagnoses established in a database. Results of the first part: Pooling all the effects extracted from the meta-analyses performed with the different risk factors, the largest effect sizes are for: Depression (OR = 2.46) and Stroke (OR = 2.27). Another of the highest effect sizes is found between the association between LDL cholesterol type and AD (OR = 2.55). Results of the second part: stroke variable was the most relevant variable (predict at 74.87%) in all classifiers. Arterial hypertension also constituted an important variable in the SVM classifier. Furthermore, antiarrhythmic drugs were considered a protective factor in all classifiers. Finally, the smoking variable generates controversy in the different models, since in some it is classified as a protective factor and in others as a risk factor for AD. To conclude, this thesis applies diverse methodologies with different samples to demonstrate the predictive robustness of risk factors associated with the development of AD