THE USE OF ARTIFICIAL NEURAL NETWORKS IN CLINICAL MEDICINE

2016 
Digital Agenda in Serbia involves the introduction of an electronic system for monitoring of the main characteristics of patients, disease progression and treatment outcomes through EHR (Electronic Health Record). Internationally standardized data set contains more than 150 variables, with a tendency to introduce new frequently. In addition to the increased demand for treatment, there are also demands for optimizing the health care system. In order to predict the likelihood of diagnosis, course and outcome of treatment, classically multivariate regression linear logistic model is being used. In recent years, studies indicate that the use of Artificial Neural Networks (ANN) may provide improved results in terms of likelihood of final diagnosis and outcomes that include input variables which, by their nature, have a non-linear interdependence. We reviewed current ANN models, their advantages and disadvantages compared to common regression models and their applicability in clinical practice. Also, we analyzed and suggested models that could possibly optimize the process of diagnosis, predict the cost and duration of treatment and rationalize medical and other resources by reducing the cost/ benefit coefficient per patient.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    34
    References
    1
    Citations
    NaN
    KQI
    []