Image processing in medical diagnosis

In medicine, having the right perspective means recognizing diseases earlier and treating them better, or even saving lives. Modern medical technology is unthinkable today without image processing for diagnostic purposes or real-time monitoring of operations. Image processing can also be used to advantage in areas such as medicine or life sciences. Image processing technologies are mainly used in the field of diagnostics, early tumor detection, and therapy progress monitoring. Known applications are, for example, examinations with radioscopy, computed tomography (CT), magnetic resonance imaging (MRI), or ultrasound on the patient. 

Image processing methods can also support laboratory medicine and diagnostics, for example, the use of optical 3D measurement technology in the dental field or for skin cancer screening in the form of whole-body scanners. Image processing allows the improvement of diagnostic information and the obtaining of quantitative information that can be used to monitor the evolution of diseases and the response to therapies. The digital image processing procedures can be grouped into four categories (by points, by areas, geometric and global).

Point processing

It is used to modify the contrast, varying the pixel value according to a rule applied to the corresponding original value. The most used procedure in practice is the reworking of gray and color levels to improve contrast.

Processing by zones (or areas)

It uses mathematical operations (filtering) as a result of which the value of each pixel in the processed image depends on the value of the surrounding pixels. Zone processing often has the purpose of softening the contrast (smoothing) or making the images sharper. The most common procedure to make the image softer is 9-point smoothing, in which the value of each pixel is recalculated on the basis of the average of the values ​​of the eight elements that surround it in a 3 × 3 matrix (9 points). , it is possible to use filters that make the profiles of the structures more evident (edge enhancement) to facilitate the detection of fine details in mammograms, chest radiographs, and the visualization of bone structures.

Geometric processing

It includes simple movements and complex procedures such as rotation, mirror arrangement, magnification, and deformation. Among these, the translation and series realignment of cross-sections and volumetric data play a fundamental role in the co-recording of images obtained with the same technique at different times or of images obtained with different techniques such as PET and CT, to integrate morphological information and functional (fusion imaging).

Comprehensive processing

It includes analytical processing procedures, among which the classification of tissues (segmentation), which has the greatest interest, identifying homogeneous classes of tissues, then allows us to measure their volumes. The segmentation of brain structures, for example, is used to measure the volumes of gray and white matter and any pathological tissues.

The segmentation operations could also be performed manually by expert operators by tracing the profiles of the areas of interest. But for complex structures and multiple injury diseases, the manual operation would take an unacceptable amount of time for routine applications. Therefore computer-aided methods are indispensable.

Among the various imaging techniques, MRI is the most suitable for segmentation procedures, due to its multi-parametric nature that allows algorithms to analyze multiple parameters to classify tissues. Fully automatic multi-parametric (multispectral) MRI segmentation procedures are available that can be used to objectively identify pathological conditions, such as brain atrophy and changes in brain volumes over time and during therapy in tumors, in degenerative diseases, such as Alzheimer's disease, and in that inflammatory of the central nervous system such as multiple sclerosis.

Three-dimensional reconstruction

The reconstruction of three-dimensional (3D) images is the most spectacular aspect of the application of tomographic techniques. Three-dimensional images play an important role both in teaching (allowing interactive learning of normal and pathological anatomy) and in the clinic (for the guidance of biopsies and interventional procedures and for the planning of surgical and radiotherapy interventions.) Alongside the specific software developed by the manufacturers of diagnostic devices, there are many freeware solutions that allow excellent 3D processing on a PC.

Some procedures require preliminary tissue segmentation. The most used procedures that do not require preliminary segmentation are multiplanar reconstruction (MPR, Multiplanar reformation, or multiplanar reconstruction). It is used mainly on CT and MRI data, to display plans oriented differently than those of the scan, through interpolation and resampling of data; the maximum intensity projection (MIP) which creates projection images (two-dimensional), of structures that have a distinctly different signal from the surrounding ones, which are eliminated from the representation. In MIP images, the resulting pixels are those that have the highest value when the studied volume is read by a virtual ray coming from an external observation point. MIP reconstructions are used for the evaluation of MRI angiography.

 The main procedures that require a preliminary classification of the tissues are the surface or volume renderings that include elaborations such as the extraction of surfaces with equal values, the binary classification of voxels, and the direct rendering of the volume. The latter is based on a semitransparent gel-like model of the 3D data set, and with this procedure, each voxel is associated with transparency and a level of colors. In volume rendering, the observer can interactively modify the display parameters (color, brightness, and opacity of the tissues). 

 In the surface rendering, the separation profiles between the two structures are displayed. Shaded surface display (SSD), in which the intensity of each pixel is calculated using the orientation of the surface with respect to a virtual lighting source, it can be used for angiography (as an alternative to MIP), to study a field of surgery, for functional MRI and virtual endoscopy. 

 The latter is based on virtual reality applications and allows the visualization of the internal surfaces of organs, such as the intestine, bladder, respiratory tract or internal walls of blood vessels, starting from CT or MRI images after 3D processing, with the ability to navigate within the organs, just like when using an endoscope for functional MRI and virtual endoscopy. 

Author: Vicki Lezama