This chapter will discuss the digital lung, x-rays, and key components of the x-ray CT scanner to help better understand lung CT AI scanning protocols, and to briefly review the historical progression of advancements in x-ray computed tomography of the lungs from the 1970s through the development of multidetector spiral CT (MDCT) scanners in the early 2000s. Each CT technology advancement improved visual and quantitative assessment of x-ray CT images of the lung. The challenges that needed to be overcome included decreasing the examination time, increasing the spatial resolution of the CT images, and decreasing the x-ray dose to the patient. Each of these improvements, during this timeframe, was very important toward enabling future AI approaches in the diagnosis and assessment of lung diseases.
The Digital Lung
The digital lung is defined as a true three-dimensional (3D) representation of the right and left lungs ( Fig. 2.1A ). The digital lung, using human whole-body x-ray CT scanners available in 2020, is a collection of volume elements (voxels) that collectively represent the lung, and each voxel is assigned a value that corresponds to the average linear attenuation of the x-ray photons by the lung tissue within that voxel ( Fig. 2.1B ). The average volume of both human lungs is about 5 liters. The size of the voxel, using modern MDCT scanners, can be as small as 0.5 mm for each axis. The total number of voxels contained in the lungs, in this case, is 40 million voxels. The first in a series of lung CT AI agents is the computer program running the x-ray CT scanner. Its objective is to generate 3D digital images of the lung that can then be viewed by patients and their physicians. This digital lung is the environment that additional AI agents can observe and, depending on the actions available to the particular AI agent, achieve additional objectives, such as segmenting the lungs from the other thoracic tissue and analyzing the lung voxels for measures of normal and diseased lung tissue (see Chapters 4 through 7 ).
X-ray Computed Tomography
X-ray computed tomography is a transmission computed tomographic method through which a precisely collimated external x-ray beam is transmitted through the human thorax ( Fig. 2.2 ); the location and energy intensity of the x-ray photons that are not scattered or absorbed by thoracic tissue are captured on an x-ray detector designed for this purpose.
The x-ray CT scanner consists of several major components: x-ray tube and electronics, x-ray detector and electronics, CT gantry, patient table, computer hardware, software, and displays ( Fig. 2.3 ). From the very first x-ray CT scanner to the current modern x-ray CT scanners of the 2020 s, they all have needed these components to produce and display a CT image. The advances in CT technology over the past five decades were made possible by advances in each of these key x-ray CT scanner components. It should be emphasized that without the AI software program running the x-ray CT scanner, there would be no practical way to reconstruct the CT images. The advent of relatively small and capable minicomputers in the 1960s made it possible to construct commercially viable x-ray CT scanners. The basic design of CT scans from the 1970s to the present share several common design features. Since the advent of human CT scans, it has been desirable to have the patient lay still in a supine position on the CT “scanning” table. The table moves the patient along their long axis, head-to-toe or z-axis, so the entire body can be scanned, if desired. The need to obtain multiple 1D projections of the supine patient laying in one position on the scanning table led to a CT scanner design where the x-ray tube rotates around the supine stationary subject with its corresponding x-ray detector on the opposite side of the patient ( Fig. 2.4 ). The x-ray tube produces a beam of x-rays and these x-ray beams can have several geometric shapes including pencil beam, broad parallel beam, fan beam, or cone beam, depending on the CT scanner design. The most common current configuration of the x-ray beam is a rotating 50- to 60-degree cone beam ( Fig. 2.5 ). The shape of the x-ray beam determines the shape of the x-ray detector array. The rotation axis of the x-ray tube and x-ray detector assembly is around the z-axis of the subject, and the highly collimated x-ray beam travels through the patient and onto the x-ray detectors. Today, the xy and z-axis width of individual detectors is less than 1 mm depending on the x-ray CT scanner model.
There are other medical imaging techniques that can generate 3D digital images of the lung; including magnetic resonance computed tomography (MRI), single-photon emission computed tomography (SPECT), positron emission computed tomography (PET), and hybrid imaging that combines two of these techniques (e.g., PET/CT, SPECT/CT, PET/MRI). X-ray transmission computed tomography (CT) is the most widely available computed tomographic method and is uniquely capable of rapid 3D imaging of the lungs, with both high image contrast and high spatial resolution, linear density scale over typical lung density values (−1000 HU to 0 HU), low claustrophobia, and low dose radiation (0.15 mGy to 1.5 mGy using modern x-ray CT scanners). The arrival of MDCT scanners in the 2000s, with 64 or more z-axis detector channels, has fueled multiple high-quality research studies that have increased our understanding of lung CT AI, its application in detecting, and assessing lung cancer and diffuse lung disease. Many research studies that have been done using x-ray CT in assessing COPD, ILD, and lung cancer, and these research studies provide the scientific foundations for this book (see Chapters 4 through 8 ).
X-rays
There are two forms of ionizing electromagnetic radiation, x-rays, and gamma rays. They are both high-energy photons capable of breaking covalent chemical bonds and because they can break covalent chemical bonds, are referred to as ionizing electromagnetic radiation. Liquid water is the most abundant molecule in the human body and the ionization energy to remove an electron from water is 11.2 electron volts (eV). A photon with an energy equal to or greater than 11.2 eV is considered ionizing radiation. Covalent organic chemical bonds link together atoms in carbon-based living organisms and breaking these bonds can have serious consequences for the organism. The 3D distribution of energy concentration in units of joules per kilogram of the x-ray photons passing through the body is referred to as ionizing radiation absorbed dose. X-ray and gamma-ray photons are defined by how they are produced. The x-ray is produced outside the nucleus of an atom. In contrast, gamma rays are produced within the nucleus of the atom through radioactive decay of an excited nucleus. The energy of the gamma-ray is equal to the difference in the initial excited energy state of the nucleus and a lower state, or a ground state, of the nucleus.
The differential attenuation of the x-ray beam photons in biologic material provides the mechanism to generate diagnostic medical images of the human body. The differential attenuation of the photons involves elastic (Rayleigh) scattering of x-ray photons, photoelectric absorption of x-ray photons, and Compton scattering of x-ray photons, as the principal mechanisms of differential attenuation of the x-ray beam (using peak x-ray tube voltages (kVp) between 25 kV and 150 kV). The dominant attenuation mechanism of x-rays in lung tissue and other soft tissue using a kVp between 70 kV and 150 kV (typical of non-contrast x-ray CT of the thorax) is Compton scattering. The probability of a Compton scattering event per unit mass of incident x-rays on low atomic number materials (typical of normal lung tissue) is nearly independent of the atomic number (Z) of the scattering lung tissue. The probability of a Compton scattering event per unit volume of incident x-rays on low atomic number materials is proportional to the density of the material. The reason for this is, except for the hydrogen atom, the number of electrons per gram of tissue for low atomic number materials, such as carbon and oxygen, is relatively constant, so the number of electrons per gram of tissue is driven mainly by tissue density; the result is that the attenuation of x-rays in the lung tissue is proportional to the density of the lung tissue. A good friend and colleague of mine, Jim Hogg, MD, has likened the x-ray CT scanner to a lung densitometer, and the above discussion supports this view. Photoelectric absorption of x-rays does come into play when intravenous iodinated contrast material is given to the patient just before the scan is performed. The presence of a high-atomic-number material like iodine (Z = 53, with a K shell orbital electron binding energy of 33 kV) will mean that the absorption of incident x-rays by iodine atoms will depend primarily on the photoelectric effect when incident x-rays interact with an iodine atom.
Important Components of an X-ray Computed Tomographic (CT) Scanner
CT X-ray Tube
The production of x-rays for medical imaging purposes is achieved by using an x-ray tube. The x-ray tube is a vacuum tube that has a very low internal pressure so that the interior of the x-ray tube contains very few atoms that would otherwise attenuate the flow of electrons in the x-ray tube. The x-ray tube contains a cathode, typically a coiled tungsten wire, at one end of the tube and an anode, typically a flat tungsten target, at the other end of the x-ray tube. A current is passed through the cathode tungsten wire, which heats the wire, and the thermionic emission of electrons results in free electrons coming from the cathode. The electrons are accelerated from the negatively charged cathode to the positively charged anode by placing a high voltage between the cathode and the anode. This voltage ranges from 25 kV to 150 kV in medical imaging. When the high-energy free electrons collide with an anode, made of a suitable material, x-ray photons are produced. Tungsten (W, Z = 74) is one of five refractory metals in the periodic table (Nb, Mo, Ta, W, Re) and has the second-highest atomic number (Z = 74) of the refractory metals. The higher the atomic number, the greater the production of x-rays with energies that are high enough for use in x-ray CT of the lung. Tungsten has the highest melting point and lowest vapor pressure of all metals, and at temperatures over 1650°C has the highest tensile strength. Tungsten is a refractory metal with very high resistance to heat and wear, as well as a high atomic number (Z = 74), so it is an ideal anode target material. A Tungsten (90%)-Rhenium (Re, Z = 75) (10%) alloy can be used to increase the resistance to anode surface damage.
The impact of the electrons on a tungsten metal anode produces the x-ray photons using two different methods: characteristic x-ray photon method and Bremsstrahlung x-ray photon method. The characteristic x-rays are generated when an electron with sufficient kinetic energy strikes and ejects an inner electron (K electron orbit) in material with a high atomic number (Z), such as tungsten (W, Z = 74). Subsequently, an outer electron with the same high Z material transitions to the vacancy in the inner electron orbital and, in this process, an x-ray photon is generated with energy equal to the difference in the binding energy of the outer electron and the inner electron that was ejected. X-ray photons greater than 100 eV are termed characteristic photons whose energies are determined by the metal anode target. The most energetic characteristic x-rays from tungsten are due to transitions from the L (57.98 keV, 59.32 keV), or M shell (67.24 keV) to the K shell of tungsten. The Bremsstrahlung, or breaking method of x-ray production, involves the deceleration of electrons as they penetrate a high Z anode target material, such as tungsten, and the deceleration of the electrons generates x-rays with photon energies between the working function of the metal and the keV used in the x-ray tube. The Bremsstrahlung x-rays are usually more abundant than characteristic x-rays in a CT scanner x-ray tube. The efficiency of generating Bremsstrahlung x-rays increases with the atomic number of the anode (Z = 74 for tungsten) and with the kilovoltage that is applied between the cathode and the anode.
Modern x-ray tubes designed to be used with x-ray CT scanners have high power ratings, 5–7 megajoules, to be able to produce a large number of x-ray photons in a very short period of time. The size of the x-ray tube focal spot on the tungsten anode is an important factor in determining the spatial resolution of the CT scanner. A smaller focal size can improve spatial resolution. However, a smaller focal spot size decreases the maximum number of x-ray photons that can be produced in a given period of time. Selection of the optimum focal spot size is an important CT scanning protocol variable for lung CT AI. The focal spot size of current x-ray CT scanners ranges from 1 to 2 mm.
The x-ray tube peak kilovoltage (kVp) and tube current (mA) are two key variables in a lung CT scanning protocol. The kVp determines the x-ray beam peak energy, x-ray beam energy spectrum, and the efficiency of producing x-rays. The increase in efficiency of photon production will increase the ionizing radiation dose to the patient. The higher the kVp, the greater efficiency of producing x-rays for a given tube current. The higher the kVp (between 70 kV and 150 kV), the lower the tissue contrast. It is important in quantitative CT work to agree on the same kVp, since different kVp settings will change the HU values assigned to lung voxels for a given mA. The higher the kVp, the more efficient the x-ray photon production, so for a given mA, a higher kVp will result in higher radiation absorbed dose to the patient. For a given kVp, the tube current will determine the amount of x-ray photons produced. The higher the tube current, the more x-ray photons are produced for a given kVp. A higher tube current results in less image noise, but also higher radiation absorbed dose to the patient for a given kVp.
The x-ray tube current in mA determines the number of electrons that are accelerated between the x-ray tube cathode and anode per unit time. The tube current in mA multiplied by the length of time the x-ray tube is turned on, exposure time in seconds is called the tube current-time product or mAs. The mAs is proportional to the total number of x-ray photons the x-ray tube produces. The higher the mAs, the greater the number of x-ray photons produced and the higher ionizing radiation absorbed dose to the patient. The selection of a mAs value has a direct effect on the image noise and the radiation dose to the patient. Lung CT scanning protocols use mAs values that ensure adequate image quality with the lowest amount of radiation dose to the patient or subject being scanned.
Modern x-ray CT scanners use a bow tie-shaped x-ray beam filter to shape the beam so that all the detector elements in the detector array see a more uniform number of photons per unit area striking them. The use of additional x-ray beam filtration can also modify the spectrum and intensity of the x-ray beam before it is transmitted through the patient. The addition of 5–10 mm thick aluminum filters in x-ray CT imaging will decrease the amount of lower energy x-rays that cannot penetrate the thickness of the thorax and contribute to the formation of the image, and also increase the fraction of higher energy photons that can reach the x-ray detector if they are not scattered or absorbed.
It is now possible to control the kVp and mAs on modern CT scanners in realtime, and these values can be adjusted to provide a consistent signal-to-noise ratio in the CT image and a lower and more uniform radiation dose to the imaged tissues. These are referred to as tube kilovoltage modulation and tube current modulation ( Fig. 2.6 ). The use of mA modulation in lung CT AI work has been successfully implemented recently. The use of kVp modulation is an issue in lung CT AI work when CT image voxel density values need to be consistent across time and patients. The varying kVp will change the x-ray beam energy spectrum, hence the CT image contrast and voxel density values will vary in the lung in a way that is challenging to correct at the present time.
In summary, the important factors regarding the operation of the x-ray tube that need to be factored into a CT scanning protocol include focal spot size, tube current-time product (mAs), peak kilovoltage (kVp), additional x-ray beam filtration, tube current modulation, and kilovoltage modulation.
CT X-ray Beam Shape and Energy Spectrum
The x-ray photon beam has several properties that impact scan time, image contrast, and ionizing radiation dose. Beam intensity and cross-sectional size of the photon beam determine how much tissue can be imaged per unit time. For a given beam intensity, the larger the cross-sectional size of the beam, the shorter time it takes to scan the lungs. The early CT scanners had very narrow pencil or rod-shaped photon beam cross-sections. This limitation was due to the reconstruction algorithms that were used, as well as limitations in the x-ray tube power and available x-ray detectors. The scan times were very long. The size of the x-ray beam cross-section has steadily increased over time as corresponding increases in x-ray tube output, x-ray detector array size, and computer power have been able to deal with large amounts of data per unit time coming from the x-ray detectors.
The x-ray beam photon energy spectrum that irradiates the patient is a major factor in the image contrast of the final CT image. This energy spectrum needs to be carefully selected so that adequate image signal to noise ratio and contrast to noise ratio are present in the lung CT images at the lowest possible radiation dose. The energy spectrum needs to be the same in order to compare one patient to another and to assess the same patient at multiple time points. This is an important concept in quantitative CT of the lung because the values of the linear absorption coefficients, which determine the value assigned to each voxel in the image, are a function of the x-ray beam energy spectrum. To be able to compare voxel values over time from the same lung, or between lungs of different individuals, the x-ray energy spectrum needs to be the same. The photon energy spectrum for x-ray CT scanners is determined by the peak kilovoltage applied across the x-ray tube cathode and anode, and the type and thickness of metal filters placed in the path of the x-rays emitted from the x-ray tube anode. Higher x-ray tube peak kilovoltage (kVp) will shift the peak energy and the entire energy spectrum toward higher energies, and also increase the number of x-ray photons generated for a given mA setting. Lower peak kilovoltage will shift the peak energy and the energy spectrum to lower energies and decrease the number of photons generated for a given mA setting. The addition of metal filters between the x-ray tube anode and the patient can further shape the energy spectrum without increasing the peak energy by filtering out lower energies that are not able to penetrate the patient and reach the x-ray detectors. The peak kilovoltage and metal filters need to be carefully selected to optimize image quality and minimize the radiation dose to the patient or subject.
X-ray CT Detectors
The x-ray CT photon detectors that have been used since the first commercial CT scanners made in the 1970s to the current generation of CT scanners use energy integrating detectors (EID). The total energy of one or more photons that a single EID detector element measures during the measurement time are integrated to provide a total energy signal. This signal is often produced by many different photons of different energies. There are new commercial CT scanners in development that use photon-counting detectors (PCD). The photon-counting CT scanners detect each photon and its energy. The PCD CT scanners compared to the current EID CT scanners have the potential to further reduce radiation dose, increase image contrast and spatial resolution, correct beam hardening artifacts, improve CT intravenous contrast media enhanced imaging, and create new quantitative metrics for lung CT AI.
The x-ray CT detectors are located opposite the x-ray tube. The x-ray CT detectors detect photons that are not scattered or absorbed by the tissues. These photons are transmitted through the tissue, without being scattered or absorbed, and then are detected by the x-ray detectors. The different tissues scatter/absorb the x-ray photons differently, so the number of and energy of x-ray photons that impact the x-ray detectors is related to the thickness and specific composition of the tissues imaged. The EID type detector elements are typically made from high-density ceramics containing rare earth materials, for instance, gadolinium oxysulfide (Gd 2 O 2 S). The ceramic material is at the front of the individual detector element and absorbs the transmitted x-ray photons coming from the patient. Visible light is emitted by the ceramic material. This light is detected by an array of electronic solid-state photodiodes that register the location and total energy of the absorbed photons during a very short measurement time interval. The electronic signal from the photodiodes is then digitized by an analog to digital converter and these digital signals are sent to the CT scanner’s computer.
The physical size and shape of the detector array are determined by the shape of the x-ray beam. The current x-ray CT scanners use a narrow-angle cone beam, so the x-ray detector is a two-dimensional curved surface. The detector curvature is optimized to minimize differences in path length from the x-ray source to a given detector element. The size of the individual x-ray detector elements determines the maximum spatial resolution possible for the CT scanner. The rotation speed of the CT x-ray tube detector pair determines how fast the thorax can be scanned; all other factors held constant. The current generation of commercial CT scanners have a 360-degree rotation time as short as 0.25 seconds.
CT Gantry
The CT gantry has a large aperture at its center for the patient and the patient table to transit ( Fig. 2.6 ). The modern CT gantry consists of two concentric cylinders, slip ring design, where the x-ray tube, x-ray detectors, and electronics, including the high voltage generator, rotate around the patient continuously. The control signals and scan data are transmitted from the rotating cylinder to the outer stationary cylinder ( Fig. 2.7 ). The x-ray tube and corresponding detector array maintain an accurate fixed alignment with each other as they rotate around the patient. The x-ray photons that pass through the object are detected by the corresponding detector array located opposite the x-ray tube that is aligned with the detector array. Most CT scanners have a single x-ray tube and corresponding x-ray detector, but Siemens has made dual source CT scanners for over 15 years, which have two x-ray tubes and two corresponding x-ray detectors 90 degrees apart ( Fig. 2.8 ). The current third-generation dual-source CT scanners have effective scan times that are half the scan times of similar single-source CT scanners. The dual-source CT scanner enables dual-energy CT imaging as well.
CT Table, Isocenter, Scan Pitch, and Scanning Modes
The CT table in modern x-ray CT scanners is computer controlled and can position the patient very precisely on the x, y, and z-axis. Isocenter is an important CT scanning protocol parameter and refers to positioning the patient at the geometric center of the CT gantry aperture. The table can be lowered for ease of getting the patient on and off, and the table can be raised to precisely position the patient so that the center mass of the patient’s thorax is at the isocenter of the CT gantry aperture. It is very important to scan the patient at the isocenter for visual or quantitative CT applications. During spiral CT scanning, the CT table needs to accurately and precisely position the patient in response to commands from the CT scanner. This includes accurate and precise positioning of the patient while the table moves at relatively high speeds in the z-axis direction. The combination of the continuously rotating x-ray tube, x-ray detector within the CT gantry, and the moving CT table enables spiral CT scanning modes where the patient is moved continuously through the CT gantry, and a spiral path is traced out by the x-ray tube and x-ray detector relative to the patient. The tightness of this spiral path is the scan pitch. Pitch is an important CT protocol metric that needs to be set in the CT scanning protocol. Pitch is defined as the table feed distance per 360-degree rotation of the x-ray tube and x-ray detector array divided by the z-axis width of the x-ray detector array.
Pitch=(Table Feed Distance per 360-degree rotation)/(z-axis width of the x-ray detector array)
A table speed of 80 mm/second, 360-degree gantry rotation time of 0.5 seconds, and a z-axis detector array width of 40 mm would result in a spiral CT scanning pitch of 1.0. A pitch of 1.0 in spiral CT scanning mode generates CT images that are similar to contiguous axial scanning mode (see Scanning Modes). The radiation dose is inversely proportional to the pitch. Pitch values higher than 1.5 enable faster scanning, lower radiation dose, and lower image quality. Pitch values less than 1.5 results in slower scanning, higher radiation dose, and higher image quality ( Fig. 2.9 ). The pitch value usually chosen for lung CT AI are 1.0 or very close to it.
