Independent Vector Analysis of Non-Negative Image Mixture Model for Clinical Image Separation
Chest X-Ray (CXR) plays a significant role in the investigative imaging schemes for diagnosing chest diseases such as lung cancer, tuberculosis, pneumonia, and asthma. The lung nodules are overlaid with ribs in CXRs. Hence, lung nodules and bony structures need to be separated to increase the perceptibility of the infected area and analysis of chest diseases without much trouble. The purpose of this chapter is to separate hard bony structures and soft lung tissue in CXRs. To separate blindly the mixed source, independent component analysis is used for one-dimensional data, whereas independent vector analysis (IVA) is applied to multidimensional data. Therefore, to achieve the separation of bone image and lung image in the CXR, two-dimensional IVA is presented. The performance of IVA is compared with other reported blind image separation (BIS) technique for standard images, dual-energy CXR images as well as conventional CXR images. The IVA algorithm for the separation of dual-energy CXR is better and dynamic.