Use of Artificial Neural Network for Abnormality Detection in Medical Images

Abstract

Cancer is one of the diseases the human race has been battling against. Several researches and technologies are being updated and implemented for the detection and cure of this obliterating disease. The cure of this disease is significantly dependent on its early detection when the tumor is in the early stage and is detectable. Thus, relative classification criteria would benefit the radiologist in determining the malignancy at an early stage. With this view, this current chapter will help the radiologist as well as oncological physician to quickly identify and determine the stage of cancer as well as whether the tumor is being either malignant or benign. The texture feature estimation and morphological analysis have been done with lot of chest X-ray images such as pulmonary nodule (PN), non-pulmonary nodule (NPN) and unclassified (tuberculosis, pneumonia, etc.) images retrieved from the Japanese Society of Radiological Technology (JSRT), public biomedical database. This feature is then applied to expert classification analysis system such as artificial neural network (ANN). This chapter implements classification analysis procedures to improvise the image classification of X-ray images, which will enhance the detection probability at a very early stage. Thirty percent of radiologist fails to detect the malignancy in the beginning stage. There are also possibilities of reducing false-positive results. These false-positive (FP) results can be due to inter-observatory analysis errors resulting from different faults in rib vessel and its structuring. Thus, the reduction in false-positive (FP) images and the increased true-positive (TP) images is important for an accurate analysis of the X-ray.