Computed tomography has become the standard thoracic imaging technique after chest radiography performed on the first line. Technological developments have made it possible to shorten the duration of image acquisition to a few seconds and to image the entire thorax in patients who cannot hold a long apnea. The purpose of this article is to present the fundamental principles of the formation of the CT image, the different embodiments, and the constraints of this examination.
CT is used to measure the absorption of X-rays within an explored volume, using multiple incidences of an X-ray beam. The CT scanner, commonly called "scanner," consists of a rotating assembly composed of an X-ray tube facing rows of detectors, this assembly rotating around the patient. The number of rows (“bars”) of these detectors is increasing with each new generation of detectors. The digital information from the different angular incidences during the rotation is processed by a computer to define the value of the X-ray attenuation coefficient of each pixel of the axial image obtained (the axial plane being perpendicular to the head – feet axis) of the patient). The image matrix is a table composed of n rows and n columns defining a number of elementary squares or pixels. The current matrices are most often in 512×512. Whatever the sectional imaging modality (CT, MRI, ultrasound, etc.), an image always has a certain thickness, and each pixel on the screen, therefore, represents a unit of an elementary volume called a voxel. In CT, each voxel in the reconstruction matrix corresponds to an attenuation coefficient.
The attenuation coefficient (often called density) of a voxel is measured in Hounsfield units (UH) on a scale from −1000 for air to+3000, the 0 correspondings to the density of the water, +1000 being the density of the compact bone. The density of a tissue structure varies between 20 and 80 HU and a fatty structure between - 40 and - 120 HU.
The human eye distinguishing only 16 gray levels, the range of 2000 densities of the human body (from –1000 to +1000 HU) cannot be seen simultaneously on the screen. Among this wide range of densities, we define a window of densities that will be effectively translated into gray levels on the screen. A window is defined by its width (W for width), and it's level or center (C).
For the study of the mediastinum, where the structures have close densities (tissue: 20 to 80 HU; pure liquid structures: 0 to 20 HU; fatty structures: −120 to −40 HU), a "tight" window of width w=400 HU, and centered by the average density of the mediastinal structures C=0 HU). The densities shown range from −200 to +200 UH.
•X-ray computed tomography is a radiological imaging technique for obtaining axial sections.
•"Strips" are the number of rows of detectors that pick up the residual X-ray beam.
•The pixel is the elementary square at the origin of the formation of the image.
•The voxel is the unit of volume corresponding to a pixel.
•The attenuation coefficient (or density) of a voxel is measured in Hounsfield units (UH).
•The size of the window used depends on the density of the tissue studied.
Until 1989, only the sequential acquisition existed, a single axial cut (perpendicular to the patient's axis) is acquired with each rotation. Then the examination table advances and stops again for the realization of the cut next. With this mode, in an apnea, only a few cuts are made, and several apneas are necessary to cover the chest. There is the spacing between each of these cuts (example: 1 cut of 1mm thick every 10mm).
The volume acquisition was developed in the early 1990s, the principle is the displacement of the examination table, during the continuous rotation of the tube – detector pair. The tube is describing a helix shape around the patient; this type of acquisition is also called "helical." It is the most commonly used acquisition mode today, and it allows you to image the entire chest in less than 15 seconds on most current machines and requires only one apnea.
The quality of an image depends on the spatial resolution, the contrast resolution, and the possible presence of artifacts.
The resolution, in contrast, is the ability to differentiate two structures whose density is close (low natural contrast, which is the case in the mediastinum). The contrast resolution is altered by image noise, which visually translates into a granularity on the image. This noise disturbs much more the visualization of structures with low natural contrast as in the mediastinum than the visualization of the lung. Noise can be measured on an image by the standard deviation of the densities measured within a homogeneous structure. This noise increases with:
•The decrease in the dose of X-rays, hence sometimes a discreet granularity on a thoracic scanner "low dose of irradiation," carried out within the framework of systematic screening.
•Reducing the reconstructed cutting thickness. Therefore, in addition to the series of fine images of the pulmonary parenchyma, we reconstruct a series of thicker images for the mediastinum, less noisy.
•The image reconstruction filter: The mediastinal series is reconstructed with a high-resolution contrast filter, which reduces noise (smoothed image). As a corollary, the spatial resolution decreases, and even by applying a parenchymatous windowing to the mediastinal series. The pulmonary structures will appear less precisely than on the pulmonary series with a high spatial resolution filter.
Spatial resolution is the distance below which CT can no longer separate two points. The spatial resolution is defined by the size of the voxel. In high resolution "standard" mode, the voxel size is 0.625×0.625mm in the axial plane, the third dimension of the voxel being the reconstructed cutting thickness (0.7 to 1mm for fine analysis of the pulmonary parenchyma).
It depends on factors intrinsic to the machine, but also on selected factors:
•The thickness of the reconstructed image: the finer the cut, the smaller the thickness of the voxel. Reconstructed cutting thickness of 3mm or more is insufficient to analyze invasive pneumonia;
•The size of the matrix: most often 512×512. Very high spatial resolution with a voxel size of 0.3–0.4mm can be obtained using a 768 2 matrix, but it is exceptionally useful in thoracic imaging and requires a large capacity for storing digital data;
•The reconstruction filter: the choice of a high spatial resolution filter increases the visibility of details while increasing noise. The (moderate) increase in noise is not a problem for viewing the pulmonary parenchyma, which has high natural contrast. On the other hand, even by applying mediastinal windowing to the series of parenchymal images, the noise will make the mediastinal structures much less visible;
•The reconstruction field the width of the thorax.
•The dose of X-rays.
•Volume (or helical) acquisition replaced sequential acquisition.
•The quality of the image depends on the spatial resolution, the contrast resolution, and any artifacts.
•Spatial resolution is the distance below which CT can no longer separate two points.
•The resolution, in contrast, is the ability to differentiate two structures whose density is close.
•Noise is an alteration of the image, which visually translates into a granularity.
The CT allows us to know the density of each voxel of the patient. The first step is always the reconstruction of transverse axial images, the number of which now reaches several hundred for the study of the thorax alone. Only a selection of these images is usually reprographic on film, and increasingly on paper. These digital images that can be viewed on-screen can have several supports, CD-ROM and DVD, and Picture Archiving and Communicating System (PACS), allowing the precise comparison of a patient's scanners, and to judge the evolution of his pathology. This digital data of axial images called "native" can be processed by a computer to improve the visualization of certain signs, but also to quantify the extension of certain pathologies.