Assessment and Evaluation of cancer CT images using deep learning Techniques
A. Sheik Abdullah, A. Manoj, Selvakumar Sellamuthu · 2021 2nd International Conference on Secure Cyber Computing and Communications (ICSCCC) · 2021
Image detection and extraction play's a significant role for automation and image analysis process. In the field of medical domain, the extraction of features from real-time images of cancer CT has becoming challenging. The incorporation of machine learning in medical data could explore the salient features that could be used for exploring patterns which then aids in decision making process.Motivation The realm of determining the accuracy of the observed medical image serves as an important factor in accessing image quality and recognition. The quality of image acquisition and analysis lies at quality factors such as range, sharpness, color, fare of lens, artifacts, and tone reproduction. The mechanism of automatic prediction is being useful for a number of practical applications, but if there is a systematic means of accessing the sections of poor quality of subsequent images in medical informatics it can be helpful for assessment and evaluation to determine flags and noise thereby rendering quality-based fusion on all observed images.Objective The objective is to validate the accuracy of cancer CT images using machine learning techniques. The assessment is made in accordance with the deployment of Convolution Neural Network with the enhancement in the image quality features.Methodology The process of assessment and evaluation involves training the dataset with proposed quality metrics. Once trained then it is modeled using CNN with ELM and its training parameters. Once modeled then tested accordingly with automatic prediction with quality factors and detecting sections of poor fusion focusing on interpretation and evaluation.Results and Conclusion The analyses of lung Computerized Tomography (CT) has various interventional variations which can significant degrade the level of accuracy. The NIST FRTV rate has FNMRs of lesser that 3% at the level of 0.01% FMR for higher learning algorithms. When assuming at the level of 0.1% the level of magnitude moves to a higher level of variations. Video frames if received will also have some of the variations in higher level of magnitude. In order to improve the quality metrics, we proposed a new model for predicting the level of cancer using Deep Convolution Extreme learning machine.