1. Few shot learning for cross domain ckd and prediction based on homomorphing filter with tuna swarm optimization.
- Author
-
Stella, A. and Kumari, P. Vasanthi
- Subjects
SUPERVISED learning ,CHRONIC kidney failure ,COMPUTED tomography ,TUNA ,REFLECTANCE - Abstract
Chronic Kidney Disease (CKD) involves illnesses that harm your kidneys and lessen their capacity to maintain your health. Most of the supervised machine learning methods are employed to construct in-domain CKD prediction with the use of labelled datasets. However, employing a classifier that has been trained with labelled images for a certain domain to categorize a picture of CKD on a different domain sometimes leads to poor results, while images that are present in the training domain might not appear in the test domain. So, Deep learning approaches are developed for cross-domain CKD prediction. Initially, CT scans for CKD are gathered and pre-processed using homomorphing and frost filters to enhance the picture quality. A frost filter is utilized to eliminate noise from the original image, and a homomorphing filter is utilized to improve the image contrast level of the noise removal image. In a homomorphing filter, reflectance component values change much over the range between (0, 1). So, the optimal value of the reflectance component is selected based on tuna swarm optimization. Gradient vector flow is then used to segment the pre-processed picture. The segmented picture is then used as an input in the classification process. This model was created to identify cross-domain based CKD using a few shot learning model based on the Attentive Squeeze network (ASNet). According to the experimental study, the proposed approach achieves 96% of accuracy. Consequently, the developed model is the best option for detecting cross domain based CKD prediction with better accuracy. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF