| Popis: |
In the modern era, the field of medicine is witnessing an enormous growth in methods of diagnosis to the methods of treatment and the abundance of information which is not reaching the front lines in an appropriate time deters the ability to correctly choose an optimal line of treatment. The explosion of the availability of information about the methods of diagnosis and various treatment options requires some sound computational framework which enables to optimally use the patient-specific diagnosis as well as a suitable treatment, and one such computational framework is artificial intelligence (deep learning), which performs the tasks with the help of computers, normally requiring human intelligence such as image processing, speech recognition, audio and/or visual perception, translation, and decision making. The recent development of Artificial Intelligence has provided such a platform to handle, compile, analyze, and interpret vast data in order to enable doctors to shift the focus from diagnosis and identification of patient-specific treatment to actual patient care. In this paper, we present an effective and efficient diagnosis system using Artificial Neural Network (ANN) for Thyroid Disease (TD) diagnosis by first training the model based on training data set and then use the model to predict the presence of disease. We have compared the prediction results of three different methods i.e. Multiple Regression, Random Forest (Ensemble Learning Method for Classification), and Artificial Neural Network (Deep Learning). When the accuracy results are compared for the prediction results, ANN delivers the highest accuracy as compared to the other two methods. The accuracy predictions for the Test Dataset of ANN, random forest and Multiple Regression are 98.22%, 22.46%, and 2.42% respectively. Given the thorough design and problem formulation, powerful computational frameworks such as an artificial neural network can help problem specific decision making as it yields very accurate prediction results as compared to other traditional models. Future scope as described in this paper, outlines the potential research area and the possibility of integration of artificial neural network for providing patient-specific personalized diagnosis and feasible and optimal treatment options which will minimize the risk and the time of recovery. |
