Framework to predict the metabolic syndrome without doing a blood test: based on machine learning for a clinical decision support system

Metabolic Syndrome (MetS) is a cluster of risk factors that increase the likelihood of heart disease and diabetes mellitus, and researchers have recently linked it to worse outcomes for the novel Covid-19 disease. It is crucial to get diagnosed with time to take preventive measures, especially for patients in locations without proper laboratories and medical consultations. This work presents a new model to diagnose metabolic syndrome using machine learning and non-biochemical variables that healthcare professionals can obtain from initial consultations. For evaluating and comparing the model, this work also proposes a new methodology for performing research on data mining called RAMAD. The methodology standardizes the novel model’s comparison with similar classification models, using their reported variables and previously obtained data from a study in Colombia, using the holdout and random subsampling validation methods to get performance evaluation indicators between the models. The resulting ANN model used three hidden layers and only Hip Circumference, dichotomous Waist Circumference, and dichotomous blood pressure variables. It gave an Area under Receiver Operating Characteristic curves (AROC) of 87.75% by the International Diabetes Federation (IDF) and 85.12% by Harmonized Diagnosis or Joint Interim Statement (HMS) diagnosis criteria, higher than previous models. Thanks to the new methodology, diagnosis models can be thoroughly documented for appropriate future comparisons, thus benefiting the studied diseases’ diagnosis. Medical personnel needs to know the factors involved in the syndrome to start a treatment. So, this work also presents the segmentation of metabolic syndrome in types related to each biochemical variable. It uses the RAMAD methodology together with several machine learning techniques to design a framework to predict MetS and their several types, without using a blood test and only anthropometric and clinical information. The results showed an excellent system for predicting six MetS types that combine several factors mentioned above that have an AROC with a range of 71% to 96%, and an AROC 82.86%. This thesis finishes with the proposal of using a SCRUM Thinking framework for creating mobile health applications to implement the new models and serve as decision tools for healthcare professionals. The standard and fundamental characteristics were analyzed, finding the quality attributes verified in the framework’s early stages. Keywords — Metabolic Syndrome, Segmentation, Quine–McCluskey, Random Subsampling validation, RAMAD, Machine learning, Framework, International Diabetes Federation (IDF), Harmonized Diagnosis or Joint Interim Statement (HMS).

• Autores: Barrios Barrios, Mauricio Andrés
• Tipo de recurso: Doctoral thesis
• Fecha de publicación: 2021
• Institución: Universidad del Norte
• Repositorio: Repositorio Uninorte
• Idioma: eng
• OAI Identifier: oai:manglar.uninorte.edu.co:10584/11004
• Acceso en línea: http://hdl.handle.net/10584/11004
• Palabra clave: Metabolic syndrome; Sangre - Análisis
• Rights: openAccess
• License: https://creativecommons.org/licenses/by/4.0/