Alzheimer’s is categorized as one of the priority types of dementia for public health of Ecuador and the world, it is one of the main causes of disability and family dependence, due to the progressive deterioration of intellectual abilities, caused by a series of genetic factors, environmental, health factors (injuries to the brain, degenerative diseases, etc.) or even to the same lifestyle. However, this pathology could be detected in early stages through the application of artificial intelligence techniques, for the implementation of prototypes of expert systems. This research work addresses a systematic review of literature contrasting related works, which serve as support to show that the extraction of characteristics of an iris image allows to verify the evolution of a pathology, to the point of generating a reference pattern as a result of the study. This project focuses on the development of a prototype prediction of Alzheimer’s levels, in order to provide a clinical contribution to specialists such as the Neurologist, Psychiatrist or Psychologist who address the issue in question. On the other hand, all the information corresponding to the database of images of the iris was obtained in the foundation “CASA AURORA DEL PERPETUO SOCORRO”, these images allowed obtaining the results that are presented in the project. Finally, the information is analyzed through the use of unsupervised training algorithms and image processing.


One of the most significant differences between each classification algorithm is the percentage of accuracy, which may vary according to the characteristics that are implemented in each of them.To achieve an adequate characterization of the subtypes of this pathology, more significant characteristics must be used and more relevant, that is, a better specific treatment based on patterns that may present with these characteristics. The adequate extraction of characteristics is proportional to an adequate recommendation of the pathology in question, can be achieved with an exhaustive work of spatial image processing. In the work in question, the most relevant classifier is grouping of k-Means clusters with an 82.43% of accurate diagnoses, with a 70.11% probability of success for a patient with this pathology to obtain a diagnosis with an assertive result and a 69.77% of a patient in optimal conditions is pathology does not. The values ​​of this study allow us to have a better picture of the situation that is being handled and of the variants that a future could intervene to improve the proposed algorithm. The use of artificial vision tools is limited to the present work , since there is not strictly enough data for its use. Supervised and unsupervised learning can significantly help the crystallization of a project, the parameters that will be used must always be defined by a 75 multidisciplinary group where specialists and technicians in the area are always involved, so that the emission of results always has a significant value.

RECOMMENDATIONS The differentiation between the subtypes of pathologies should be investigated using well-characterized and larger samples, that is, the data The image should be more solid and with a better resolution to avoid influencing the results. Since the potential for image analysis and data quality depend on the criteria setting and decision making when issuing the diagnosis. With the experimentation of digital processing algorithms, an infinite number of parameters that can be useful for any type of investigation can be determined. The error values ​​allow to make a decision about the model and the permissible error range and adapt it according to the approximation criteria. The data generated by the predictive classifier allows feedback to the process of generating the function being modeled

IRIDIOLOGY WITH DIGITAL PROCESSING OF IMAGES ”was carried out by Mr. VEGA GUALLICHICO ROBERTO JAVIER, which has been completely reviewed, analyzed by the content similarity verification tool; therefore it meets the theoretical, scientific, technical, methodological and legal requirements established by the University of the Armed Forces ESPE, which is why I allow myself to accredit and authorize it to support it publicly

Download full abstract: