The application possibilities of digital radiology include the management functions (archiving, transmission, recovery, etc.) and the processing ones, aimed at the “piloted " reconstruction of the image to make it more responsive to production needs. The set of techniques used to achieve this latter goal is known as “image processing.”
The growing development of artificial intelligence is expanding the domain of this discipline to include the procedures of computerized image analysis, the recognition of structures by the operator, up to the automatic reject systems through vision and image analysis.
The image processing then undertakes to operate on one or multi-digital images to increase its diagnostic evaluation, for this reason, it is often also called image enhancement.
The possibility of obtaining a significant increase in the diagnostic quality of the examination after the “post-processing” acquisition is a peculiarity of digital radiology. On the other hand, it does not allow any reworking, forcing the technician to modify the radiological parameters and acquire further images, if the first image obtained is not sufficient for a definitive evaluation. Digital radiology, therefore, guarantees the full exploitation of any acquired image, allowing, within certain limits, the correction of exposure errors and allowing the extraction of all the information present from the image, making visible those that are not perceptible to the naked eye.
The improvement of the image can be obtained in different ways, using very different methods for each operating procedure, to the objective to be achieved, such as the enhancement of the contrast, the detection (edge detection, sharpening, and noise reduction. Most of the techniques act on several images with the use of addition, subtraction, or average algorithms. They use the information contained in the different images to obtain the desired result. The common points between the various techniques essentially reside in the operating procedure; the numbers that encode the gray levels attributed to the pixels of the original image are modified with the application of different functions and deposited in another memory area, where they constitute an alternative image to the original one.
The various techniques differ from each other in the applied mathematical function and in the number, location, and processing order of the pixels to be modified.
One of the most widespread classifications, in fact, distinguishes the techniques in point, regional and global: in the first, conceptually simpler and more widespread, all the pixels are examined individually sequentially, and modified by applying the chosen function to each of them.
In regional processing, the modification of each pixel takes place, taking into account the values of the surrounding ones, arranged in a small matrix of 3x3 or 4x4 (known as the kernel). In global processing, the choice of the new value to be attributed to each pixel occurs after the analysis of the values of the pixels contained in much larger matrices, often coinciding with the entire set of pixels.
The difference between the three types of techniques, therefore, lies in the weight that is attributed, during processing, to the pixels surrounding the one in question.
If it is better to highlight circumscribed, homogeneous density structures, it is necessary to resort to regional processing, because the analysis of the values of the pixels surrounding those in question is the best procedure to identify the belonging of a point to a structure and the boundaries of this. For these reasons, the regional calculations are mainly used to delimit and accentuate margins. To define the average influence of a factor on all points of an image, and modify each pixel according to the value found, one must resort to global processing, which is therefore particularly useful, for example, for the reduction of the noise. By selecting functions and management parameters, it is possible to carry out analyzes on the images, considerably improving the vision of detail. In particular, both the display functions and the processing techniques must be remembered. The latter is used for noise reduction (filtering) and for reconstructions (reformatting and rendering).
- variation of the amplitude and the level of the window of the gray levels
- Inversion of the gray scale (image reverse).
- Enlargement of details (zooming).
- Scrolling the image on the monitor (scrolling).
- Measurements (for the precise definition of distances, angles, areas, gray levels, signal intensity, density, etc.).
Due to the use of this set of techniques, digital images come to allow, even in the presence of a lower spatial resolution, an equal or greater diagnostic efficiency than analog images, precisely in relation to the flexibility and interactivity of the presentation.
It is a disturbing entity, inseparable from the signal, which occurs in all systems and, therefore, also in radiological systems. It expresses the graininess of the system and limits the visibility of low contrast details.
It is due to several factors which with different mechanisms contribute to the degradation of the image:
- Quantum noise is common to both analog and digital systems. It depends on the mechanisms of production of radiation at the level of the x-ray tube and the processes of interaction of radiation with matter. It is due to the stochastic nature of the absorption processes of the incident radiation or of the light. Therefore, the number of absorbed photons varies from point to the point of the detector.
- Electronic noise is due to the imperfection of the various components of electronic systems (resistors, cameras, etc.) and is present in both conventional analog and digital systems. Quantization noise is present only in digital systems for rounding off the analog-to-digital conversion process.
Fully describes the loss of information in the process from acquisition to image display. It is obtained as the relationship between the information available at the input and the information actually supplied by the radiological system. Therefore, the contrast resolution (minimum variation of intensity detectable between contiguous areas of the image) relates to the spatial resolution (the ability of the system to faithfully reproduce small and high contrast radiological details, measured in pl / mm) for a conventional and digital radiological system. In conventional systems, the final contrast resolution is determined by the sensitometric characteristics of the film, by its treatment and obviously by its exposure. In digital systems, the contrast can be manipulated at will with windowing processes (variation of the view window). It thus bypasses the limitation of the human eye, which manages to discriminate less than 20 levels of gray.