Classification of noiseless corneal image using capsule networks
- PDF / 1,468,386 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 52 Downloads / 251 Views
(0123456789().,-volV)(0123456789(). ,- volV)
METHODOLOGIES AND APPLICATION
Classification of noiseless corneal image using capsule networks H. James Deva Koresh1 • Shanty Chacko2
Ó Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Classifying a particular image from a data set is a complex work for any image analyst. Generally, the output of medical image scan gives numerous images for analysis. In that, the image analyst has to manually predict a better noiseless image for computer-assisted image process program. Manual verification of all the output images from the scan device consumes a lot of time in predicting the abnormality of a patient. The proposed capsule network for noiseless image algorithm assists the image analyst by classifying the noiseless image from the data set for further computer-assisted image enhancement or segmentation program. The proposed algorithm performance is evaluated and compared with the existing algorithms in terms of accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. Keywords Capsule network Image classification Corneal OCT
1 Introduction Cornea is the outermost layer of an eye whips helps to focus the vision rays to fall over the surface of retina. The abnormal thickness of corneal layer makes the vision rays to fall in front or far away to the retina and it results in blurred vision. A laser surgery corrects up the blurred vision by altering the thickness of corneal layer; for that, a pre-surgical planning is carried out by measuring the thickness of cornea. Corneal optical coherence tomography (OCT) is a non-invasive imaging device widely applied to capture the cross-sectional view of corneal layer. Due to frequency emissions, there are chances for OCT imaging device to capture noisy image, so that the device is developed to take multiple images within a few seconds. An image analyst takes all those images from the OCT
Communicated by V. Loia. & H. James Deva Koresh [email protected] Shanty Chacko [email protected] 1
Department of Electronics and Communication Engineering, Karunya Institute of Technology and Sciences, Coimbatore, India
2
Department of Electrical and Electronics Engineering, Karunya Institute of Technology and Sciences, Coimbatore, India
device for classifying a clear image to undergo with the computer-assisted image process program of thickness measurement. However, in real case, it takes days of time to get the thickness result of a patient due to manual verification of clear noiseless image. A convolution neural network (CNN) algorithm needs large number of images for training to classify an image. To overcome such limitations, a capsule network was structured for classification. Similarly, capsule networks have the ability to encode various parameter details from an image in nature. Capsule network utilizes the vector features of an image for their input and output forms, whereas in CNN, the scalar features are utilized. The vector features are carried out by vomiting the max
Data Loading...
