Effect of Hyperparameter Optimization technique to Extract the Nuclei from the Cell

Abstract

In this modern era of computer vision technology, image processing plays a vital role in the medical field, and the analysis of medical and biological images through the implementation of image processing is a challenging task along with it also helps the doctors to diagnose the disease as well as start treatment against the disease. In computer vision, especially image processing applications, image segmentation is considered an essential part of this method in which partitioned the digital image into various small segments and from these segments extract the specific object for further investigation with the help of deep learning. The extraction of nuclei from the cell using these modern technologies make a faster method that enables a more rapid cure finding process and reduces the time-to-market of the new drug. At the same time, the accuracy is considered the critical element because reliability and acceptability are based on it. This paper helps in extracting the nucleus from the cell by using image processing techniques and the power of deep learning architectures while it is based on the U-Net nuclei segmentation framework, but the standard model achieved 84.8% testing accuracy, achieved 83.5% training accuracy and also little bit variation in predictive resultant image as compared to binary image but with a hyperparameter optimization technique, the model achieved 97.6 % testing accuracy and 97.5 % training accuracy and the predictive resultant images of the hyperparameter optimization model are more accurate as compared to without a standard model and help researchers community of medical imagery analysis.