Prediction of the pneumonia from the CT lung images by using the multiband google NET CNN

Abstract

Pneumonia is a severe infectious illness which has affected significant bereavement worldwide. This has been prevalent with people who have weak immune systems. The most efficient and sought out method to identify this via imaging is Computed Tomography Scans (CT scans). A disease like Pneumonia can be cured only when treated at the right time. This research involves a simple, innovative and effective methodology to detect pneumonia in individuals with the support of Deep Learning methodologies. With the use of image segmentation, 3D modelling and annotation, we aim at identifying this disease in human lungs. The data used here is obtained from RIDER Lung CT collection. This image data is put under sectioning and pre-processing. The mentioned techniques are done via Laplacian Partial Differential Equation-Based Histogram Equalization and a Weighted Iterative Median Filter. The required features are extracted through recursive isomapping and non-linear component analysis. By using uplift-weighted fuzzy method, the abnormal areas are segmented later. These segmented areas are converted into 3D models for better visualization using the Canny Inductive Frustum model. Finally, abnormalities are classified using the Multiband Google NET CNN classifier. This proposed method shows improved results and offers a generalized approach that can be applied to other similar datasets as well.