A Non-Invasive Method for Diagnosis of Dengue Fever using Ultrasound Image of the Liver

Dengue the bite of infected Aedesaegypti mosquitoes. Prediction of Dengue presents great challenge as fever is a serious viral disease that is transmitted through the clinical symptoms overlaps with other conventional fever. The distinctive features of dengue are abnormities in liver and changes in hematocrit. In the proposed work a non-invasive method to identify dengue using ultrasound liver image has been developed. From the ultrasound image of the dengue infected patients, Gall bladder wall thickness and free fluid secretion are the observed distinctive features. The preprocessing methods are necessary to ensure noise free images and compatible for further processing. The ultrasound images endure from speckle noise, hence SRAD filter is used to reduce the speckles. To improve the visibility and distinguish different region of interest in the image the Contrast Limited Adaptive Histogram Equalization (CLAHE) method is adapted. Hybrid of chan-vese and gradient vector flow active contour segmentation method is used to segment the image and performs better segmentation than its individual performance. For Automatic early prediction of dengue Artificial Neural Network (ANN) are trained that can predict the dengue with the given image. The pattern recognition Neural Network is trained with segmented outputs and used for early prediction of dengue fever. The prediction performance is compared with cascade feedforward NN, Radial Basis Function NN (RBFNN).