Diagnosis of Type 2 Diabetes Using Electrogastrograms: Extraction and Genetic Algorithm–Based Selection of Informative Features (Preprint)

Abstract

BACKGROUND Electrogastrography (EGG) is a non-invasive electrophysiological measurement procedure followed to measure the frequency and promptness of gastric myoelectrical activity, which is normally considered to investigate the mechanisms of human digestive system. Diabetes can cause alterations in the process of digestion. OBJECTIVE The objective of this work is to extract and select potential informative features from recorded normal and diabetic electrogastrograms for diagnosis of diabetes using EGG signals. METHODS In this work, a total of thirty features of electrogastrograms measured from normal subjects and diabetic cases, were extracted. Further, twenty potential informative features were selected using Genetic Algorithm assisted feature selection process. The extracted features were analyzed for further classification of normal and diabetic EGG signals. RESULTS Results demonstrate that there are distinct variations between the EMG signals recorded from normal subjects and diabetic patients. The investigations reveal that the features namely Maragos fractal dimension and Hausdorff’s box-counting fractal dimension have high degree of correlation with the mobility of normal and diabetic electrogastrograms. CONCLUSIONS Based on the analysis on the extracted features, it is seen that the selected features are suitable for the design of automated classification systems for classification of normal and diabetic cases.

Platform Design of Project Bidding System on the Basic of TCM-KNN and Genetic Algorithm

“…Therefore, when extracting text features, the feature words in the same text can be put into a feature vector representing the text, so as to avoid ignoring the connection between feature items [15][16]. On this basis, this paper proposes a text feature vector on the basic of X 2 statistics, which can not only preserve the correlation between text features, but also distinguish the correlation between features and classes; and uses this vector as the initial population, through multiple rounds of genetic vectors are obtained to improve classification accuracy; through the coordination of crossover operation and mutation operation, global search can be realized and local minima can be avoided [17][18]; according to the characteristics of feature extraction, the fitness function and intersection rules are designed to solve the problem of inappropriate processing of low-frequency words in statistical analysis [19][20]. The flow chart of feature extraction on the basic of GA is shown in Figure 2: The figure 3 shows the average fitness of the population optimized by the GA has reached above 0.935, and these data show that the fitness of individuals in the population is better, and the effect of evolution is better.…”