Research Article
An Efficient Detection of Kidney Stone Based on HDVS Deep Learning Approach
@INPROCEEDINGS{10.4108/eai.7-12-2021.2314490, author={Krishnamoorthy Somasundaram and Suresh M animekalai and Paulraj Sivakumar}, title={An Efficient Detection of Kidney Stone Based on HDVS Deep Learning Approach}, proceedings={Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India}, publisher={EAI}, proceedings_a={ICCAP}, year={2021}, month={12}, keywords={kidney stone deep learning performance metrics binary svm vgn-19}, doi={10.4108/eai.7-12-2021.2314490} }- Krishnamoorthy Somasundaram
Suresh M animekalai
Paulraj Sivakumar
Year: 2021
An Efficient Detection of Kidney Stone Based on HDVS Deep Learning Approach
ICCAP
EAI
DOI: 10.4108/eai.7-12-2021.2314490
Abstract
The Kidney stones (KDS) are the most normal issues in the universe that humans are affected with a high pain and also need an emergency diagnosis. To detect the KDS in an early stage for diagnosis, there are several imaging techniques are provided. For a perfect imaging, the Computer-aided diagnosis (CAD) methods are performed as auxiliary tools to for the support during KDS diagnosis. The computed tomography (CT) images are recently used for an exact diagnosis of KDS. But in the CT images, the prediction accuracy is less with the usage of traditional techniques. Therefore, the deep learning (DL) methods are well versed on handling such images with a greater accuracy. In this paper, the combination of DL methods to perform a both feature extraction and the classification for KDS detection is presented. The feature is extracted with the DL method of Visual Geometry Group network 19 (VGN-19) which is based on convolutional neural network (CNN). Then the classification is done by the Binary Support Vector Machine classifier (SVM) to build a binary model. Thus the proposed model is named as a Hybrid Deep VGN-19 and Binary SVM (HDVS) which is an efficient in the disease prediction with a performance metric of of Recall, precision, accuracy, specificity and F1 score respectively. These performances of proposed HDVS method are compared with the prior advance methods in DL approaches such as Alexnet, GoogleNet, ResNet and SqueezeNet.
