Hybrid U-Net and ADAM Algorithm for 3DCT Liver Segmentation
Suraj Ramesh, J. S. Kanchana, D. David Neels Ponkumar · 2023
Segmentation of images is a computer procedure that involves dividing an image into separate and distinct regions of pixels. This method is often accomplished by using filters or labeling. The U-net architecture, which is based on convolutional neural networks (CNNs), is widely used in the domain of segmentation of images. Nevertheless, a notable limitation of the U-net’s design is in its diminished potential for learning in the more complex aspects of the model. Instead of using the stochastic gradient descent (SGD) method, researchers have proposed an alternative optimization technique called Adam to instruct deep learning models. Adam’s method successfully utilizes the benefits provided by the Momentum and RMS Prop algorithms, allowing it to efficiently manage gradients with low density in situations with high levels of noise. The U-net design may be enhanced with the integration of the Adam algorithm.