Intelligent Diagnosis of Lung Cancer and Respiratory Diseases E-Book
79,78 €
Mehr erfahren.
- Herausgeber: Bentham Science Publishers
- Kategorie: Fachliteratur
- Serie: Intelligent Systems in Radiology
- Sprache: Englisch
Intelligent Diagnosis of Lung Cancer and Respiratory Diseases presents information about diseases of the respiratory system and the relevant diagnostic imaging techniques. The book focuses on intelligent diagnostic imaging systems.
The first section of the book deals with the physiological underpinnings of 3 major diseases that affect the respiratory system: tuberculosis, lung cancer and COVID-19. This section also explains the basic principles of artificial Intelligence that support the diagnosis of these diseases.
The next section presents applications of intelligent systems to support the imaging diagnosis of COVID-19 and lung cancer, with emphasis on digital health and telemedicine approaches.
Each chapter is organized into a readable format, and is accompanied with detailed bibliographical information for further reading.
This book is a reference for everyone seeking to understand how artificial intelligence can provide solutions for diagnostic support systems by processing and analyzing radiological images to improve early diagnosis and, consequently, lung disease prognosis.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 453
Veröffentlichungsjahr: 2002
Ähnliche
BENTHAM SCIENCE PUBLISHERS LTD.
End User License Agreement (for non-institutional, personal use)
This is an agreement between you and Bentham Science Publishers Ltd. Please read this License Agreement carefully before using the ebook/echapter/ejournal (“Work”). Your use of the Work constitutes your agreement to the terms and conditions set forth in this License Agreement. If you do not agree to these terms and conditions then you should not use the Work.
Bentham Science Publishers agrees to grant you a non-exclusive, non-transferable limited license to use the Work subject to and in accordance with the following terms and conditions. This License Agreement is for non-library, personal use only. For a library / institutional / multi user license in respect of the Work, please contact: [email protected].
Usage Rules:
All rights reserved: The Work is the subject of copyright and Bentham Science Publishers either owns the Work (and the copyright in it) or is licensed to distribute the Work. You shall not copy, reproduce, modify, remove, delete, augment, add to, publish, transmit, sell, resell, create derivative works from, or in any way exploit the Work or make the Work available for others to do any of the same, in any form or by any means, in whole or in part, in each case without the prior written permission of Bentham Science Publishers, unless stated otherwise in this License Agreement.You may download a copy of the Work on one occasion to one personal computer (including tablet, laptop, desktop, or other such devices). You may make one back-up copy of the Work to avoid losing it.The unauthorised use or distribution of copyrighted or other proprietary content is illegal and could subject you to liability for substantial money damages. You will be liable for any damage resulting from your misuse of the Work or any violation of this License Agreement, including any infringement by you of copyrights or proprietary rights.Disclaimer:
Bentham Science Publishers does not guarantee that the information in the Work is error-free, or warrant that it will meet your requirements or that access to the Work will be uninterrupted or error-free. The Work is provided "as is" without warranty of any kind, either express or implied or statutory, including, without limitation, implied warranties of merchantability and fitness for a particular purpose. The entire risk as to the results and performance of the Work is assumed by you. No responsibility is assumed by Bentham Science Publishers, its staff, editors and/or authors for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products instruction, advertisements or ideas contained in the Work.
Limitation of Liability:
In no event will Bentham Science Publishers, its staff, editors and/or authors, be liable for any damages, including, without limitation, special, incidental and/or consequential damages and/or damages for lost data and/or profits arising out of (whether directly or indirectly) the use or inability to use the Work. The entire liability of Bentham Science Publishers shall be limited to the amount actually paid by you for the Work.
General:
Any dispute or claim arising out of or in connection with this License Agreement or the Work (including non-contractual disputes or claims) will be governed by and construed in accordance with the laws of Singapore. Each party agrees that the courts of the state of Singapore shall have exclusive jurisdiction to settle any dispute or claim arising out of or in connection with this License Agreement or the Work (including non-contractual disputes or claims).Your rights under this License Agreement will automatically terminate without notice and without the need for a court order if at any point you breach any terms of this License Agreement. In no event will any delay or failure by Bentham Science Publishers in enforcing your compliance with this License Agreement constitute a waiver of any of its rights.You acknowledge that you have read this License Agreement, and agree to be bound by its terms and conditions. To the extent that any other terms and conditions presented on any website of Bentham Science Publishers conflict with, or are inconsistent with, the terms and conditions set out in this License Agreement, you acknowledge that the terms and conditions set out in this License Agreement shall prevail.Bentham Science Publishers Pte. Ltd. 80 Robinson Road #02-00 Singapore 068898 Singapore Email: [email protected]
FOREWORD
Breathing is the very first and most fundamental action that we take at birth. This fundamental step usually is taken into lower consideration as time goes by. Whenever any disease rises over the respiratory system, we are once again reminded of this essential value of respiratory function. Lung cancer has been a threat to the respiratory system, deadly for 1.8 million people in 2020, mainly propelled by smoking, with its diagnoses representing approximately 13% of the total cancer diagnosis in the whole world. This peculiar kind of cancer has posed a huge challenge for oncological research due to its high rate of mortality under X-ray screening when compared with other types of cancer. Mainly at its early stage, well-circumscribed, juxta-pleural, juxta-vascular and pleural-tail, observed in a couple of millimeters (e.g. < 3 mm). More recently, studies based on tomography screening have been more successful in reducing mortality.
Likewise, lung function is often severely challenged by tuberculosis. A ubiquitous human disease, the drama of tuberculosis inspires many great works of art, from Verdi’s la Traviata, to Victor Hugo’s Les Miserables. Tuberculosis is caused by the bacteria called Mycobacterium tuberculosis. Despite its low-cost techniques to approach it, this disease has yearly killed 1.7 million people out of 10 million infected worldwide.
In addition to the high rate of death by lung cancer, and tuberculosis, in the last 18 months, respiratory system research has faced a worldwide challenge, the COVID-19 pandemic. The acute disease is caused by the virus SARS-CoV-2. Beyond the system function, COVID-19 is a systemic disease over the vascular endothelium. In this relatively short-time, this threat to respiratory function suppressed more than 4.7 million human lives in the world, of which over a million were in the Americas. Respiratory insufficiency is one of the most dangerous symptoms of COVID-19, which demands mechanical ventilation in a significant part of the total hospitalized patients. When a large fraction of the population is infected in a small fraction of time, the demand for mechanical ventilation greatly surpasses the health system’s capacity, dramatically increasing mortality. Therefore, optimal management of the resources related to mechanical ventilation is critical to cope with the COVID-19 pandemic. Due to a wide range of impacts, and current technologies to exchange scientific resources, the response to this pandemic rapidly united efforts from all dimensions of human knowledge, including computer science, with its robust techniques to study the respiratory system.
All of these show us that tools to aid correct diagnosis of respiratory diseases may save uncountable lives every year. In the last decades, several computer-aided diagnosis (CAD) systems have been proposed. These systems can be built using different approaches. When these systems are built based on Artificial Intelligence (AI) approaches, they are called Intelligent Systems, able to learn from the data, which have supported the intelligent diagnosis of respiratory diseases. More recently, the world has witnessed an enormous development of AI techniques to explore large amounts of health data, including image data related to respiratory diseases. The accuracy of these techniques has increased, mostly based on new pre-processing methods and a fast increasing number of images in health data as well.
The COVID-19's final diagnosis is based on lung tomography. Thus, the infection spreading supports a corresponding propagation of the demands for lung tomographies. This coupled-spreading phenomenon has created an unprecedented database of lung tomographies, which, incidentally has supported early-stage lung cancer diagnosis in the same case in which the event trigger was the COVID-19 final diagnosis. Such an unprecedented database of lung tomographies certainly will fuel significant improvement in the accuracy of these AI techniques applied to lung cancer and other respiratory diseases observed in this kind of data.
Despite the fast advances in the past decade, the near future promises a flood of AI applications for research and clinics of the respiratory system, including high-resolution lung 2D/3D-images reconstruction as magnificent support in respiratory diseases diagnosis. Beyond a piece of clear information on the fundamentals of respiratory function and related diseases, this volume provides an opportunity to dive into the understanding of relevant diseases involving respiratory diseases by applying the state of art techniques powered by Artificial Intelligence over data from the respiratory system and other related systems.
PREFACE
This series of books, “Intelligent Systems in Radiology”, aims to present the principles and advances of diagnostic techniques in Radiology based on Artificial Intelligence, from the perspective of the advent of Digital Health. The series consists of three books. Each of them is based on two pillars: one dedicated to theoretical foundations and the other to radiological applications in the real world.
This first book, “Intelligent Diagnosis of Lung Cancer and Respiratory Diseases”, is dedicated to the diagnosis of diseases of the respiratory tract or those that seriously affect the respiratory system. The physiological foundations of the respiratory system and the formation of radiographic images and x-ray computed tomography are presented. Principles of respiratory diseases are also presented, including lung cancer, viral and bacterial pneumonia, tuberculosis, and Covid-19. In addition, the principles of pattern recognition and machine learning and the main theoretical and practical tools are also briefly presented. Software libraries are also mentioned. Additionally, this book presents innovative works and systematic reviews of intelligent applications in the diagnosis of lung cancer, tuberculosis, viral and bacterial pneumonias, and Covid-19.
This book is intended for academics, graduate and postgraduate students in Medicine, Biomedicine, Biomedical Engineering, Computer Science, and those who are interested in Biomedical Computing and breast cancer diagnosis applications.
The first chapter “Principles of Respiratory Diseases - Tuberculosis a brief study” presents a literature review on the field of computer aided mycobacterial detection for pulmonary Tuberculosis. In this chapter, some outstanding studies are described, limitations and points of improvement are also discussed by K. S. Mithra.
The chapter “Physiological basis for the indication of Mechanical Ventilation”, Gonçalves et al., explains the functioning of the respiratory system, as well as several diseases that can affect it, leading to Respiratory Insufficiency (RI). The chapter details RI diagnosis by blood gas analysis, its monitoring by oximeters and capnographs, and treatment by oxygen therapy and mechanical ventilation.
In “Artificial Intelligence in the Diagnosis of Diseases of the Respiratory System”, Seijas & Bezerra present an overview of current artificial intelligence techniques, focusing on deep learning. The chapter reviews needs, software, and databases that can be applied in radiology, especially in diagnosing lung cancer and respiratory diseases.
The chapter “COVID-19: clinical, immunological, and image findings from infection to Post-COVID Syndrome”, from Souza et al., shows multiple aspects of COVID-19. The authors explain the cell infection with SARS-CoV-2, as well as Post-COVID-19 respiratory syndrome. Furthermore, they relate the occurrence of virus variants, laboratory and immunological aspects, the major clinical manifestations and image findings, and all aspects associated with pulmonary damage promoted by the virus.
In “Unveiling distinguished methodologies for the diagnosis of COVID-19”, Rosa et al. bring important discussion regarding molecular and serological tests incorporated into the routine of COVID-19 diagnosis. The authors provide information about different methods used to diagnose COVID-19, including their main technical features, advantages and disadvantages.
Albuquerque et al., in “E-health, M-Health and Telemedicine for the Covid-19 pandemic”, present the world panorama of Digital Health during COVID-19 pandemic, and a geolocalized point of view, evaluating the efforts of countries worldwide. They performed a systematic mapping through Pubmed and analyzed 400 articles. Based on the findings, they concluded that E-health, M-health and Telemedicine were widely used along this period, playing an important role in patients and care provider’s safety.
In “Absolute Images Reconstruction in Heart and Lungs for COVID-19 patients using Multifrequencial Electrical Impedance Tomography System and D-Bar Method”, from Wolff et al., the authors propose the use of multifrequency electrical impedance tomography (MfEIT) in the management of pulmonary disease in ICU beds. Despite having a poor resolution in comparison to other imaging methods, MfEIT displays some advantages, which are especially positive in the context of low income countries. In this study, they applied D-Bar method to perform image reconstruction. The proposed approach proved to be promising for reconstructing conductivity images in patients with COVID-19 in the radio frequency range. Last but not less important, the authors also provide their image reconstruction codes compatible with Matlab and Octave.
The chapter “Lung cancer diagnosis: Where we are and where we will go? Classical and innovative applications in the diagnosis of lung cancer”, Neves et al., provide an overview of the different traditional and emerging diagnostic tools for lung cancer, which is the leading cause of cancer death in both men and women. They believe that an early and accurate diagnosis is critical for successful management. Thereby, the authors bring details of CT imaging, sputum cytology, biopsy, and bronchoscopy.
In the last chapter, entitled “3D Reconstruction of Lung Tumour Using Deep Auto-encoder Network and a Novel Learning-Based Approach”, Vazifehdoostirani & Ahmadi propose a novel approach for 3D tumour reconstruction from a sequence of 2D parallel CT images. With a well-described method, they achieved an approach with reduced complexity and improved accuracy when compared to other 3D lung tumour modelling approaches.
We hope that the work presented in this collection will show some of the new trends in Radiology and intelligent systems regarding respiratory diseases.
Enjoy your reading!
List of Contributors
Principles of Respiratory Diseases - Tuberculosis a Brief Study
K. S. Mithra1,*Abstract
Tuberculosis (TB) is one of the worst lung infections caused by bacteria called Mycobacterium tuberculosis. It is a disease that can be controlled in cases where it is quickly identified and treated. As sputum microscopy is a simple and low-cost approach, most countries use this technique as an initial step in the diagnosis of TB. For this, the patients' morning sputum is collected and submitted to the Ziehl Neelsen staining procedure before the examination. Then a digital microscope is used, where the images of the sputum slides can be recorded for analysis. In this context, numerous research projects have been developed in the field of computer-aided detection of mycobacteria for pulmonary tuberculosis. The survey of these works was discussed here in this work, as well as their limitations.
INTRODUCTION
Tuberculosis (TB) is a dreadful contagious disease caused by an infection with the Mycobacterium tuberculosis (MTB) bacteria. MTB was first discovered in 1882, by Dr. Robert Koch, a German physician and scientist. He received the Nobel Prize in physiology for this discovery and became known as ""The Father of Bacteriology"" [1]. MTB can affect any organ of the human body, especially lungs, via the bloodstream and/or lymphatic system. This scenario leads to pulmonary tuberculosis (PTB). If MTB affects any other part of the body it is called extra pulmonary tuberculosis (ETB). Two major reasons for the persistence of this disease are poverty and HIV infection [2].
When an infected person coughs or sneezes, tiny droplets of TB germs or MTB may come out. In this way, the airborne droplets can infect other people. Each of these people will have a 10% lifetime risk of becoming infected with TB.
Tuberculosis is a leading cause of adult morbidity and mortality worldwide. This fact can be explained by drug resistance and HIV co-infection, both consequences of TB [3]. In India, more than 40,000 people are infected and more than 1,000 die from tuberculosis every day [4].
BASIC FACTS OF TB
Even though TB is noted from the past as one of the human afflictions, still it is one of the leading killers among the infectious diseases, even with the worldwide use of several vaccines and antibiotics. Rapid diagnostic methodologies and appropriate medications are needed especially in developing countries to stem the worldwide epidemic of TB that kills two million people per year [5].
TB Disease
Tuberculosis is one of the most harmful airborne infections after entering the human body through the respiratory system. There are two levels of tuberculosis, which are: Latent tuberculosis (LTB) and Active tuberculosis (ATB), as shown in Fig. (1) below.
Fig. (1)) Levels of Tuberculosis infection.When MTB bacteria enters the human body, the initial level of infection is called LTB. At this stage, the person has no symptoms of the disease. Furthermore, the human immune system fights with the multiplication of bacteria. Then, when the bacteria start to multiply, the ATB stage starts. In this case, emergency care is required. People infected with LTB may remain asymptomatic and never transmit the disease [6]. Only 5-10% of patients with LTB are infected with ATB throughout their lifetime [7]. Thus, as LTB infection poses the risk of ATB disease in the future, the treatment is also recommended for LTB. People most at risk of becoming ATB patients are children under the age of 4, people infected with the Human Immunodeficiency Virus (HIV), people with malnutrition and people with any disease that weakens their immune system.
TB Symptoms
Usually, TB symptoms vary depending on the area in which MTB is growing in the human body. In the majority of cases, the bacterium affects lung regions resulting in PTB. The common symptoms of the PTB disease include:
Weakness, Weight loss, No appetite, Chills, Fever, Night sweatingA bad cough persists for 3 weeks or longerChest painCoughing up blood or sputumInfection of MTB on other parts of the body such as pleura, lymph nodes, abdomen, genitourinary tract, skin, joints, bones and meninges are called ETB [8]. Its symptoms in addition to common TB symptoms are given below.
HeadAche, Confusion, Neck firmness - TB Meningitis;Back pain, Swelling of backbone, Fever - TB in Spinal cord;Blood in urine - TB in Kidney;Swelling in the neck with or without sinus discharge - TB in Lymph node.TB Treatment
It is possible to treat tuberculosis with proper treatment. Whenever a person becomes infected with MTB, initial treatment is with antibiotics such as isoniazid and rifampicin. Then, the patient was monitored for the existence of bacilli, which usually appear after two months of infection [18]. Depending on the number of bacilli present, treatment modalities may vary periodically.
TB Challenges
TB has become one of the most challenging health problems worldwide, mainly because of its two problematic inflectional conditions such as Drug-resistant TB and HIV Co-infection [9].
Drug-Resistant TB
If the initial Anti TB drug such as Isoniazid and Rifampin fails to kill the infected MTB, then the disease is considered to be drug-resistant TB. In Drug-resistant TB (DRTB), bacterium spread in the same way as that of drug-susceptible TB. The common causes of DRTB are:
Incompleteness of full course of TB treatment;Wrong medications provided;Unavailability of proper medications;Poor quality TB drugs.DRTB can be of two types: Multidrug-Resistant (MDR) or Extensively Drug Resistant (XDR) [9]. If the infection is resistant to at least one of primary medications (isoniazid or rifampin), then the state is called MDR-TB. The second line drugs are suggested to cure the disease. Again, if the bacterium is resistant to any one of the second line drugs in addition to primary drugs, it is called XDR-TB. It is estimated that about 4,00,000 people are infected with DRTB every year.
HIV Co-infection
TB is the leading killer of people who had HIV infection, even though both are extremely different types of infections. Globally, 10% of TB incident cases are notified as with HIV co-infection [9].
TB Prevention
Vaccination
The vaccine, namely Bacillus Calmette-Guerin (BCG), was developed in 1920. It mainly focuses on preventing TB in children, as every day around 500 children die from TB. This vaccine has good protection and is applied in all countries with childhood immunization programs. Furthermore, it's a low-cost vaccine.
TB Control Strategies
Giving instructions to infected people about TB, so that they will be aware of its spread and of the importance of taking the full course of medicine to prevent DRTB.Giving proper treatment to cure the TB cases and to stop bacteria spreading from one person to another.Identifying and treating LTB disease in time to prevent the emergence of ATB from LTB.Prevention mechanisms should be adopted in places where the chance of TB infection is greater (e.g. hospitals and prisons).Preventive mechanisms must be adopted in the houses of infected people to minimize cases of person-to-person transmission.HIV-infected adults and children must receive adequate care to avoid TB infection.Malnutrition must be abolished in poor countries for a fruitful control of TB infection.Global TB Report
Every year the World Health Organization (WHO) publishes a global TB report in order to acquire the up-to-date statistics of worldwide TB epidemic and of the progress in prevention, diagnosis, and treatment of the disease. This report is primarily based on data collected from countries and territories. There are 30 high burden countries (HBC) notified in the 2017 global TB report [10] as shown in Table 1.
MYCOBACTERIUM TUBERCULOSIS
The causative agent of this terrible deadly tuberculosis is a pathogenic bacterium called Mycobacterium tuberculosis. This rod-shaped bacterium is a species of the Mycobacteriaceae family shown in Fig. (2). There are different types of these germs. The most common causes TB disease and others cause leprosy and atypical bacterial infections. The MTB cell wall is covered with mycolic acid, which forms a lipid coating on its surface [28]. This coating makes the cells stain resistant and therefore acid-fast staining is applied to identify bacilli. Thus, MTB is known as alcohol-acid resistant bacilli [11].
Fig. (2))Mycobacterium tuberculosis.Mycobacterium Taxonomy
Kingdom: Bacteria Phylum: Actinobacteria Order: Actinomycetales Suborder: Corynebacterineae Family: Mycobacteriaceae Genus: Mycobacterium Species: Mycobacterium tuberculosis
Mycobacterium is the only genus belonging to the Mycobacteriaceae family and currently there are more than 170 species identified as belonging to the Mycobacterium genus [12]. Each of these species has its own genotypic characteristics that distinguish it from species in other genera. MTB and seven closely related mycobacterial species such as Mycobacterium bovis, Mycobacterium africanum, Mycobacterium microti, Mycobacterium caprae, Mycobacterium pinnipedii, Mycobacterium canetti and Mycobacterium mungi together comprise the MTB complex.
Human Pathogens of Mycobacterium
Mycobacterium Tuberculosis- the causative agent of tuberculosis in humans and therefore humans are the great repository of these organisms.Mycobacterium Bovis- the causative agent of tuberculosis in cows and also in humans. The human body becomes infected after ingesting unpasteurized cow's milk. This type of infection can lead to ETB, resulting in bone infections in the human body and resulting in hunched back.Mycobacterium Avium- the causative agent of diseases such as tuberculosis, especially prevalent in HIV-infected people.Mycobacterium Leprae- the causative agent of leprosy.Cell Wall Structure of MTB
The MTB cell wall is distinct among prokaryotes, which plays an important role in determining the bacteria's virulence. About 60% of the MTB cell wall is composed of complex lipids. The cell wall lipid fraction contains three main components such as mycolic acids, string factor and wax-D. The cell wall complex may be composed of peptidoglycan [13].
Mycolic acids- are unique alpha-branched lipids that make up to half the dry weight of the mycobacterial cell envelope. They are strong hydrophobic grains that form a lipid coating for the bacteria that affects the cell surface penetrability properties. It is one of the important determinants of the bacterium's virulence.
Cord factor- determines the serpentine cord level within the morphology of the bacteria. Chains of cells in smears often form serpentine cords based on this factor that is associated with virulent strains of bacteria. It is more poisonous to mammalian cells.
Wax-D- in the cell wall is the main component of lipids and its high concentration can lead to the following characteristics of MTB organisms:
Impenetrability to the stains and dyes;Opposition to many antibiotics, acids, and alkaline compounds;Continued existence of macrophages.Characteristics of MTB
MTB is a straight or slightly curved rod-shaped bacterium with a length of 2-4 micrometers and a width of 0.2 to 0.5 micrometers [14].It is an aerobic organism, thus it needs oxygen for its survival and is therefore found mainly in the upper aerated lobes of the lungs.The existence of these organisms can be in single pairs or small clusters within the infected human body.They are non-mobile organisms and therefore are not able to move or move around in the infected regions of humans.It can remain dormant for years in the human body, resulting in LTB.It started to multiply due to ATB infection and this facultative bacterium can take 15 to 20 hours for its generation.These organisms lack encapsulation as well as spores.Pathogenesis of MTB
MTB is transported in airborne particles called droplet nuclei that measure 1 to 5 microns in diameter. Infectious droplet nuclei are generated when people with pulmonary or laryngeal tuberculosis cough, sneeze, scream, or sing. Depending on the environment, these tiny particles can remain suspended in the air for several hours. MTB is transmitted by air, not by surface contact. Transmission occurs when a person inhales the MTB droplet nuclei and the droplet nuclei pass through the mouth or nasal passages, upper respiratory tract, and bronchi to reach the alveoli of the lungs. The alveolar macrophages then ingest the MTB organism. The power of human immunity inhibits or nullifies these organisms in most cases. Otherwise, it started to multiply and spread through the bloodstream to all parts of the body, resulting in the worst disease.
DIAGNOSIS OF TB
There are different methodologies for detecting MTB infection in the human body which are explained below.
Skin Test
A small amount of fluid called tuberculin is injected into the skin, usually in the forearm region. A small hard red bomb will rise in this part within 44 to 72 hours for positive TB cases. Never worry about latent or active TB infection. This method is highly sensitive but fails in people vaccinated with BCG.
Interferon Gamma Release Assay
It is an adequate blood test to identify latent TB infection, even in people vaccinated with BCG. It measures the immune response of a person with MTB. It takes 24 hours to check immunity and results in positive TB for people with low immunity. This test fails in HIV-positive people and both the skin test and the interferon-gamma release test have low sensitivity for identifying the disease [15].
Chest X-ray
Chest X-ray is examined for any abnormal shadows or hazy appearance during the diagnosis of tuberculosis. MTB infection is confirmed by this appearance. It takes 20 to 30 minutes to perform this test, but it cannot confirm disease accurately in resource-poor settings and therefore this method is of low sensitivity [16].
Serological Test
Here, the blood sample is examined for antibodies. It takes 4 to 5 days to give results and has low sensitivity and specificity in diagnosing the disease. As it produces higher false positive rates, WHO is strongly against to the use of this test [17].
Sputum Examination
It is the first widely used TB test in countries with a high TB epidemic. It is a simple and inexpensive method that takes less than 30 minutes to make a decision. A series of stains are applied to the sputum slide and examined under a microscope to check for MTB organisms. Depending on the number of object bacilli present, the degree of infection for treatment is graduated.
SPUTUM SLIDE PREPARATION
Direct microscopic examination requires quality slides of sputum samples collected from patients. Patients with TB symptoms are advised to produce sputum to the laboratory for examination at three different times.
An on-site specimen during your first hospital visit.Another morning specimen the next day.While submitting the early morning specimen to the laboratory, again one spot specimen was also collected.Equipment needed
The sputum smear should be prepared quickly after collecting or receiving samples. The essential tools needed in the laboratory for slide preparation are provided below [29].
New clean glass slides without scratches and grease.A 50 ml volume container for storing the sputum sample.Bamboo or wooden applicator sticks to spread the sputum on the blades. Wire loops with an inner thickness of 3mm can also be used for this task.A marking pencil to mark the identification number of each specimen on the frosted edge of the slides.Alcohol or sand trap to clean excess sputum on applicator sticks after spreading.Stain support for drying smears.Spirit lamp to light stained slide during staining.The smear must be prepared in manageable collections. In the beginning, the labeling of the slides was done on each slide to differentiate from each other. Then, the collected specimens were applied to the slides with the aid of applicator sticks. The samples were smeared over an area of approximately 1 cm x 2 cm and were made thin enough to be able to read through the slide. Then, the smears were dried in atmospheric air for 15 minutes. The slide was then lightly heated with a flame and then cooled for staining. The image of an actual sputum smear sample in direct view is shown in Fig. (3). The applicator stick was discarded after each use and a new one was used for each sample.
Fig. (3)) Actual sputum direct smear.ZN Staining of Sputum Slide
These staining techniques were first developed by Ziehl and later modified by Neelsen and are therefore called ZN staining. The reagents required for the ZN staining procedure are 1% Carbol Fuchsin, 25% Sulfuric Acid, and 0.1% Methylene Blue, respectively. The Carbol fuchsin reagent was prepared using the following steps.
Initially, 10 mg of fuchsin basic dye was transferred to the 250 ml conical flask.Then, 100 ml of alcohol was added and the dye was dissolved by the addition of water heated to about 60°C. This mixture is called solution 1.Then, 50 mg of phenol crystals were transferred to a 1000 ml conical flask.Then 500 ml of distilled water is added to make solution 2. The phenol crystals were dissolved by giving moderate heat.Solutions 1 and 2 were mixed with distilled water to make 1000 ml of solution.Finally, the solution was filtered and labeled with 1% Carbol fuchsin reagent.The next 25% sulfuric acid reagent was prepared following the steps below.
Initially, 750 ml of distilled water was poured into the flask.Then 250 ml of concentrated sulfuric acid was added slowly and the solution was heated and labeled as 25% sulfuric acid.The generation of the methylene blue reagent was given in the steps below.
Methylene blue of 1 mg in weight was measured and transferred to a one liter flask.Then 1000 ml of distilled water was added to dissolve the methylene blue reagent by stirring the solution well.Finally, the solution was filtered and named 0.1% methylene blue reagent.ZN staining on the sputum slides was performed after the reagent generation process. The various staining steps involved are given as follows.
1. The sputum slides were placed on the staining support so that a maximum of 12 slides could be kept. The slides with the identification number were placed in an order such that there was at least one finger space between the slides.
2. Starting at the edges, the entire surfaces of the slides were smeared with Carbol Fuchsin reagent.
3. The blades were heated slowly until steam appeared. The heated slides were left free for ten minutes to dry.
4. After ten minutes the slides were washed with water. Excess water can be removed by tilting the blades.
5. Then, the bleaching agent (25% sulfuric acid reagent) was poured onto the slides and left for 3 minutes.
6. After 3 minutes, the slides were washed and tilted to remove all water content from the slides.
7. 0.1% methylene blue reagent was poured onto each slide for 1 minute.
8. Finally, the slides were carefully rinsed under running water to remove excess stains and kept on the support block to air dry.
BRIGHT FILED MICROSCOPY
Brightfield sputum microscopy requires very simple laboratory facilities and therefore was an inexpensive substitute for the complex and expensive sputum culture methodology. Zeinl Neelsen (ZN) staining was used for conventional brightfield microscopy. It has been the main diagnostic technique for TB disease for over 100 years. The brightfield microscope is shown in Fig. (5). Image acquisition for computer-aided MTB algorithms was performed using a charge-coupled camera mounted in the optical path of the microscope, if the microscope is not digital.
Fig. (4)) ZN stained microscopic sputum image. Fig. (5)) Bright Field Microscope.These images are sent to a digital computer for bacilli detection through various image processing techniques. For sputum slides, imaging is usually performed under both conventional (brightfield) and fluorescent microscopes. Improving the number of fields on the sputum slides that are analyzed under a microscope can improve the results for further treatment of the disease. A minimum of one hundred fields must be checked before reporting a negative sputum result. The microscopic image of the ZN-stained sputum smear is shown in Fig. (4). After staining with ZN, bacilli objects resemble red colored rods and can exist in groups or individually. Sometimes they can be slightly curved structures clearly visible from the blue background. The carbol-fuchsin reagent changes the bacillus objects to red color.
Infection Level Recording
Bacillary objects observed under a microscope should be counted to determine the severity of the infection. The accuracy of manual counting depends on the ability of the microscopic analyst or laboratory technician working in the laboratory. The grading of the level of infection was standardized by WHO as 1+, 2+ and 3+ depending on the bacilli present in varied microscopic fields and which were shown in Table 2.
FLUORESCENT MICROSCOPY
Fluorescent microscopy requires Auramine staining of sputum slides to detect the bacilli objects.
During the year 2011, WHO introduced the Light Emitting Diode, based on fluorescent microscopy, aiming to replace the conventional fluorescent microscopy. The auramine staining requires 0.1% auramine, 0.5% acid alcohol, 0.3% methylene blue or 0.5% potassium permanganate, and 1 or 2 volumes of decolorizing objects per each volume of stain. The image of a fluorescent microscope is shown in Fig. (6).
Fig. (6)) Fluorescent Microscope.Auramine staining on sputum slides is performed in the same way as that of ZN staining. However, the reagents used for preparation are different. The various steps are given as follows.
1. All the smear slides are arranged with a finger distance apart and filtered auramine is applied on the smear side of each slide.
2. After 20 minutes, the slides are cleaned with water and acid alcohol is applied on each slide.
3. After 1 to 2 minutes, the slides are again cleaned and methylene blue is applied on each slide.
4. Finally after 1 minute, they are washed with water and kept away from sunlight for air dry. The slides are examined only after completely air dried.
Bright Field with Fluorescent Microscopy
Fluorescent microscopy had 10% more sensitivity than bright field microscopy. In addition, it has reduced staining time and the work load of lab staff. In contrast, it is costly compared with bright field technique. Thereby, most medical laboratories around the world continue to use bright-field microscopy. Despite this, WHO recommends that steps be taken for switching to fluorescent microscopy gradually.
COMPUTER AIDED DIAGNOSIS OF TB
Computer Aided Diagnosis (CAD) is a digital methodology which helps physicians and lab technicians for diagnosing diseases and to make decisions within a short time period. It provides information or reports based on images and other relevant data about the disease [19]. With CAD systems, physicians are able to detect diseases earlier and they can monitor the progress of the treatment efficiently [20].
Medical Imaging
Medical imaging is the process of generating pictorial representation or images of parts of the human body. They play an important role for CAD systems. For TB disease, the medical imaging is performed mostly with the following diagnostic procedures.
Sputum examination - Digital microscopes attached with a computer are used for generating microscopic views of sputum slides [21].Chest Imaging - The techniques used for chest imaging are conventional radiography, computed tomography, magnetic resonance imaging, positron emission tomography, fluorodeoxyglucose positron emission tomography [22, 23].Sputum Image Processing
Sputum image processing is used to automatically analyze the microscopic image of stained sputum slides for the detection of MTB organisms. The most common workflow of sputum image processing includes the following techniques:
Preprocessing - Image preprocessing is performed as an initial step since it makes the image suitable for the next steps. It enhances suspect regions of the image, providing an early and efficient identification of diseases [24].Segmentation - Image segmentation is used to identify certain image regions possessing the same features or properties such as pixel intensities, color, texture etc. [25]. Thus it is used to select the bacilli objects present in the microscopic sputum image.Feature Extraction - It is used to extract object oriented feature values, to get more pixel information about the object present in the image. In sputum images some of the common bacilli features that can be extracted prior to classification are area, density, perimeter, compactness, axis lengths etc. [26].Classification - It is an efficient technique used to group the data into various classes depending on their extracted feature values. For sputum images, the extracted feature values were fed into classifiers like convolutional neural networks, deep belief networks etc. to recognize whether TB bacilli is present or not [27].EVALUATION OF VARIOUS CAD SYSTEMS
The existing automatic TB identification methods were evaluated based on their execution performance, algorithms used, noise reduction methodologies applied, separation of overlapping objects, identification of level of TB infection by bacilli counting and the data repository used for system development. The report of this evaluation is presented on Table 3. Also the availability of those parameters for different existing systems which support both light field and fluorescent microscopy are shown in Table 3. The parametric evaluation of the existing methodologies was also discussed within this section. The availability of existing works with the various current research requirements was discussed as follows.
Image preprocessing - Usually image preprocessing is done before performing segmentation and classification [54]. It gives enhanced images suitable for further processing steps and its availability in existing literature can be viewed in Table 3. Only 10% of the existing literature supports preprocessing of sputum images before performing segmentation and classification.Overlapping bacilli objects processing - The separation of overlapping bacterial objects is a complex task. It is needed for determining the level of infection through the number of bacilli objects. From the literature survey of related articles it was found that only 25% of existing works consider those objects while processing and only very few consider separation of those objects.Bacilli count based diagnosis - The counting of total bacterial objects present in the specimen is a part of estimating the level of TB infection. Most of the existing works in the literature only identify the existence of bacterial objects, and cannot determine the level of infection within the human body. That is, few works in the literature support the counting of bacilli.Performance Evaluation of Bright Field Microscopic Methods
The various automatic diagnosis systems developed for bright field microscopic models were analyzed with respect to their detection accuracy, sensitivity and specificity. The accuracy values of the different works were plotted using charts given in Fig. (7). The system's ability to correctly identify bacilli objects was measured using the sensitivity metric. In addition, the identification of non-bacilliferous objects was measured using specificity. Only a few of these highlighted systems evaluated their performance with these metrics, which were plotted on the graph provided in Fig. (8).
Fig. (7)) Accuracy chart for existing automatic TB diagnosis methodologies. Fig. (8)) Performance Chart for various methodologies in automatic bacilli identification. (a) Sensitivity Chart (b) Specificity Chart.Within the existing techniques, the highest accuracy value for bright field microscopic images was 98% and sensitivity was 100%, using hybrid multi-layer perceptron networks and moving k-means clustering technique [41]. Furthermore, it was observed that these systems can accurately detect single bacilli objects, but fail to identify and count the overlapping bacilli objects present in the image. Again, in papers [34] and [38], the accuracy value seems to be higher. In paper [34
