Denoising Segmentation and Classification of Medical Images Using Enhanced Deep Learning Based Methods

Abstract

Biomedical imaging has revolutionized the healthcare system as it helps in analysing the complications in the human body. The advent of computational science and its combination with medical imaging has enabled medical practitioners to provide the best diagnoses. newlineThe recent advancement in this combination has proved to be a boon for human lives as complex inner structures of the body can be analysed using these methods. A lot of non-invasive techniques has been devised in the recent past for serving humanity. MRI, Ultrasound, X-ray, Computed Tomography (CT) are the example of such techniques. MRI is an important modality that focuses on providing structural information and detailed characterization of disease. newlineThe MR images are further analyzed for finding the diseases such as brain tumour, heart vessel structures etc. For inspecting the pathological and anatomical structural changes in the body, the images are further segmented. The image segmentation aims to present the desired region of interest to the clinical practitioners to diagnose the disease. The advantages of this imaging technique are its non-ionization behaviour and better image quality with high tissue contrast resolution, but it is corrupted with the artefacts. These artefacts result from noise, patient body movement etc., which needs to be removed before analysing the images for the disease diagnosis. The MR images are corrupted with Rician noise, which gets induced because of magnetic coils of the receiver circuitry. Noise removal in MRI is of prime importance as it enhances the visual quality of the images. Biomedical image classification is another important task in the field of biomedical imaging. The image classification task allows the medical practitioner to identify different but related symptoms of the disease. It helps in identifying the features of the diseases in various modalities. newline newline