Hybrid Image Processing Methods for Medical Image Examination

Venkatesan Rajinikanth, E Priya, Hong Lin, Fuhua Lin

• Publisher: CRC Press
• Year: 2021
• Print ISBN: 9780367534967
• eBook ISBN: 9781000300185
• Topic: Computer Science
• Subtopic: Computer Science General
• Index: Computer Science

In view of better results expected from examination of medical datasets (images) with hybrid (integration of thresholding and segmentation) image processing methods, this work focuses on implementation of possible hybrid image examination techniques for medical images. It describes various image thresholding and segmentation methods which are essential for the development of such a hybrid processing tool. Further, this book presents the essential details, such as test image preparation, implementation of a chosen thresholding operation, evaluation of threshold image, and implementation of segmentation procedure and its evaluation, supported by pertinent case studies. Aimed at researchers/graduate students in the medical image processing domain, image processing, and computer engineering, this book:

• Provides broad background on various image thresholding and segmentation techniques
• Discusses information on various assessment metrics and the confusion matrix
• Proposes integration of the thresholding technique with the bio-inspired algorithms
• Explores case studies including MRI, CT, dermoscopy, and ultrasound images
• Includes separate chapters on machine learning and deep learning for medical image processing

1	Introduction

Recent developments in the science, technology, and medical domains have helped people live better lives. This improved lifestyle is also due to access to a wide variety of facilities, including state-of-the-art medical facilities. Further, a considerable number of vaccinations and preventive medication has helped reduce infectious and communicable diseases. Advanced treatment facilities available in multi-speciality hospitals have also helped in detecting and curing diseases in their early stages. Moreover, scheduled health checkups recommended by doctors also help to identify and cure a number of acute and deadly diseases.

Although considerable measures have been taken to prevent and cure human diseases, the incidence rate of new kinds of infections and communicable/non-communicable diseases is rapidly rising, irrespective of age, race, and gender. To support the early diagnosis and treatment implementation for diseases, a number of diagnostic procedures are proposed and implemented in various disease diagnostic centres and clinics. The medical images recorded using a chosen modality gives required insight for the disease to be identified. Based on this insight, the doctor can plan what treatment is to be implemented [1,2].

Medical imaging can be recorded with a variety of procedures ranging from camera-assisted techniques to radiation-based methods. The choice of methodology depends upon the organ to be examined and the expertise of the doctor. The recorded image can be assessed by the doctor, assisted by a computerised algorithm. Modern image recording facilities support information in digital form, known as digital imaging, thus computer-assisted diagnosis is widely adopted. Aside from this, digital images can be easily stored, retrieved, and processed using a number of techniques [3].

This chapter presents an overview of disease screening procedures for a chosen disease. The invasive and non-invasive image recording procedures existing present in literature and the assessment of the images are discussed briefly along with appropriate examples.

At the present, although a considerable number of medical facilities are available, disease occurrence rates among humans are gradually rising due to various unavoidable causes. These illness in humans can be classified into several categories, such as: (i) infectious disease, (ii) deficiency disease, (iii) hereditary disease, and (iv) physiological disease. The above mentioned diseases can further be grouped as communicable and non-communicable.

Diseases occurring externally in the human body are quite easy to detect and treat when compared with the diseases found in internal organs. A pre-screening procedure is recommended as essential to detect the disease in its premature phase. If the disease and its severity are identified in its premature phase, a treatment procedure could be recommended and implemented to control and cure the disease. This could be could imply less effort compared to a disease diagnosed in its delayed phase.

Diseases found in external body organs, such as the eye and skin, could be examined with a personal checkup by an experienced doctor along with an image supported detection system for further examination. Meanwhile, diseases of internal organs such as the brain, lungs, heart, breast, the digestive system, and blood are normally diagnosed using a chosen imaging method associated with a prescribed imaging modality. The imaging procedure followed for the internal organ needs complete monitoring and should be examined in a controlled environment with a prescribed clinical protocol. After registering the image of the organ, the disease can be diagnosed using a computerized disease examination procedure or a personal check by a clinical expert.

In most cases, semi-automated/automated disease detection procedures are implemented to speed up the diagnostic process. The report prepared with these techniques are used as supporting evidence regarding the patient, which will help the doctor during decision making and treatment planning process. Further, the availability of the computing facility helps to develop a large number of computer-aided detection procedures, which considerably reduce the burden on doctors during conventional disease detection and also during mass screening processes.

This book aims to discuss the details of the Artificial-Intelligence (AI) based disease detection procedures mainly developed by Machine-Learning (ML) and Deep-Learning (DL) techniques. Further, this book also presents the details of Hybrid Image Processing (HIP) methods implemented to enhance the detection accuracy for a class of clinical images.

1.1 Introduction to Disease Screening

Diseases are medical emergencies where the unrecognised and untreated will cause various problems, including death. These diseases are classified either as communicable or non-communicable depending on the occurrence rate and its nature. Diseases of the human body can be diagnosed in a variety of procedures, and a visual check is preferred due to its accessibility. The disease detection procedure in external organs is easy while also helping identify the severity. In some moderate/acute cases, along with a personal check, a suitable signal/image-based disease evaluation is also recommended by the doctor to verify and confirm the disease.

Diseases in vital internal organs are more when compared with external organs, hence more care needs to be taken during diagnosis. Most of the diseases in internal organs such as the heart, lungs, brain, kidney, respiratory tract, stomach, and blood are normally diagnosed using carefully chosen bio-signalling/bio-imaging procedures. The bio-imaging-based assessment helps attain more insight regarding the organs to be examined compared to the signal-based techniques. Hence, medical imaging-assisted disease diagnosis has emerged as a common and recommended technique. In this method, an image modality is considered to record and examine the internal organ.

Developments in medical imaging techniques helped to achieve a two-dimensional (2D) or the three-dimensional (3D) picture of the organ in greyscale or RGB form. These images would help the physician to get an insight regarding the disease in the body and also helps to track the progression of the disease with respect to time. Identification of the disease in its premature phase is very essential to plan for the appropriate treatment. Treatment of the premature phase disease is easy compare to other stages and hence a number of scheduled screening procedures could be planned and conducted at an early stage. The scheduled body screening will help to identify number of diseases in its premature phase, even though the symptoms are absent.

As discussed earlier, the disease in humans can be commonly classified as (i) communicable and (ii) non-communicable diseases. Each disease will have its own symptom; and the patient will immediately approach the physician when he/she experiences a disease symptom. The doctor will examine the patient with the prescribed protocol existing to identify the disease based on the symptom as well as the difficulty experienced by the patient.

The doctor will suggest a range of preliminary diagnostic procedures to confirm the disease and assess the severity level. The procedures executed to test the patient for confirmation of the disease is technically known as Disease-Screening (DS) process and it varies according to the disease to be detected. The procedures commonly employed in DS involves (i) personal check by an expert, (ii) clinical test ranging from sample collection and testing of the bio-signal/bio-image-based methods, (iii) intermediate level detection based on the bio-signals and images collected from the patients, and (iv) verification of report by the doctor for authentication of disease.

The overview of the clinical level diagnosis of the disease in human is presented in this section with appropriate block diagrams.

Figure 1.1 illustrates the initial-level verification and recommenda...

Table of contents

• Cover
• Half Title
• Series Page
• Title Page
• Copyright Page
• Contents
• Preface
• Author
• Chapter 1 Introduction
• Chapter 2 Image Examination
• Chapter 3 Image Thresholding
• Chapter 4 Image Segmentation
• Chapter 5 Medical Image Processing with Hybrid Image Processing Method
• Chapter 6 Deep Learning for Medical Image Processing
• Chapter 7 Conclusion
• Index