| Popis: |
Measurement in biological and medical scientific fields is a very important issue because a biomedical decision concerning the health status of a patient is based on a valid measurement. Health-related indices provide a reliable way of measuring a person’s behavioral or clinical health in an accurate and reliable way. From a mathematical perspective, health-related indices are quantitative variables composed by component variables, namely features that represent different aspects of a health status aiming to evaluate clinical behaviors or behavioral characteristics. A variety of statistical multidimensional techniques with sufficient performance can be used for the valid classification of individuals as healthy or nonhealthy, apart from using health-related indices. More specifically, several modified statistical classification methods, such as logistic regression, classification and regression tree, neural networks, and data mining elements including machine learning and support vector machines (SVMs), aim to increase the accuracy of classifying individuals in two or more different groups in data sets that have a specific feature or not. The aforementioned methods have been mostly developed recently while informatics' methods became an irreplaceable part of medical research. A short literature review of health-related indices and statistical classification methods used in medical data of specific mental and physical diseases such as dementia, depression, anxiety, cardiovascular, and cancer is presented, as well as, indices for nutritional assessment. In addition, the evaluation criteria used for assessing the classification and prediction ability of indices and classification methods are indicated. Finally, an application in real medical data is given regarding the prediction of a nonalcoholic fatty liver disease risk by providing the corresponding data set description and the results from all of the aforementioned methods for a variety of different scenarios. |
