ABSTRACT
Medical image reconstruction is required for deeper understanding of clinical abnormalities. There are various processes and techniques used to create functional or anatomical images of the human body. Medical imaging is a technique that allows a physician to create a visual representation of interior organs. The results obtained by medical imaging help medical professionals in verification of current treatment, correct diagnosis, and planning of treatment. The medical imaging techniques may be of various types, such as ultrasound, magnetic resonance imaging (MRI), computed tomography (CT) scan, and X-rays. The medical imaging technique is chosen depending upon the type of disease, such as kidney stone diseases, breast cancer, and brain tumor. But, the quality of the medical image is degraded by different types of noise as well as blurriness. In this chapter, an advanced methodology of image processing is proposed for detection of medical images in diseases such as brain tumor, kidney stone, and breast cancer using ultrasound images and MRI images. The RGB medical images are processed and converted to gray images with removal of labels. The image intensity is adjusted to improve the contrast of these biomedical images. Median filtering is utilized to eliminate noise. Discrete wavelet transform (DWT) is used to detect tumors in the brain. To conduct morphological and K-means clustering segmentation, the output filtered medical image is considered. Convolutional neural network (CNN) classifiers are employed to classify the given images as tumor images into two sections, i.e., malignant and benign, normal or abnormal. The final system analysis is carried out to judge the accuracy, specificity, and sensitivity by preparing the confusion matrix. The classification system accuracy obtained is approximately 90%. The presented technique can help doctors in early detection for accurate treatment of patients.
