Digital imaging

Image manipulation

 

The digital image is thus inferior to a conventional analogue radiographic image when it comes to spatial resolution. This is compensated for by the nature and potential of digital technology. The contrast resolution is superior which is an advantage when the potential for the eye to observe the different shades of darkness is enhanced by the ability to shift the contrast scale, for example on a monitor (Figs. 3,4).

It is possible to carry out a number of manipulations of the digital image in order to enhance the information content of the image. Most of these manipulations can also be carried out with analogue images but this is more cumbersome and time-consuming. As digital images should be assessed on a monitor, simple measures such as changing black and white (Fig. 4 C) or magnification of a detail can be routinely performed.

The goal for image manipulation in radiology is to increase diagnostic accuracy (Fig. 5). In the process of object-image production and final diagnostic image assessment, the image manipulation is included as a quality enhancement. In addition, the potential for different image interpretation techniques is increased compared with conventional film reading. Techniques for both interactive interpretation and automatic image analysis are being evaluated presently.

Figure 4. Examples of different simple manipulations of one digital image of a computed tomography section through the chest. A) A limited density range image with a window level of 0 HU corresponding to water attenuation and a window width of 000 HU Pixels with a value equal to or less than -500 are black, and pixels with a value equal to or above +500 are white. B) The same image with a pulmonary setting. Window level is -600 HU and the window width is between 1000 and -200 HU as in Fig. 3. It can be seen that the soft tissue details disappear in the chest wall while the structure of the lung is visualised compared with a). C) Fig. A) with reversal of the scale so that pixels of +500 HU and above are black, while pixels of -5 00 or below, are white. D) The full density range from -1000 to +1000 HU Observe the diminished contrast resolution between muscle, fat, and glandular structures in the chest wall and breasts. E) The same image as Fig. D) with single application of an edge enhancement algorithm. Some new structures are seen in the lung. F) The same image as Fig. D) with double application of an edge enhancement algorithm. Most of the lung structures seen in Fig. B) are seen but a disturbing back ground noise has been added to other tissues.

Figure 5. Schematic representation of the diagnostic imaging process and image manipulation. Image acquisition from the object, in this case the liver, may give possibly important but vague diagnostic information. Image manipulation will give a processed image that enhances the diagnostic findings - or discards them. Image analysis, normally visual, leads to diagnostic assessment in the form of a report.

 

Darkness level and window width

The simplest form of digital image manipulation is the normal use of darkness (contrast) level and windowing when evaluating CT or MRI images on the monitor. A digital image with 2048 darkness shades will lack contrast if all 2048 levels are visualized with 0 as black and 2047 as white (Fig. 4 d). A darkness level of 1024 and a window of 2048 steps has then been chosen. On a CT the scale normally runs from -1024 to + 1024 Hounsfield units, HU, as seen in Fig. 3.

In order to assess the lungs optimally on a CT chest slice, a darkness level dose to the average CT dens it y of the lungs (between -600 and 900 HU) ought to be chosen (Fig. 4 b). The window width 800 and the level -600 means that -1000 HU is seen as black and -200 HU and above is white (Fig. 3). If the same digital image is used to assess the skeletal details of the chest a window width of 1000 and the level +500 HU will result in a complete grey-scale between 0 and +1000 HU (Fig. 4).

Image subtraction

The subtraction of a pre-contrast film from a radiographic film after contrast medium injection into the arteries - angiography - has been practised for many decades. This technique was especially used when the background to the vascular tree was very irregular or dense as for example in the base of the skull or the upper part of the chest. The pre-contrast film was inverted photographically so that black became white and vice versa and then matched to the post-contrast film so that only the vascular structures were seen.

This procedure is of course both faster and simpler to perform electronically with a computer. Whole sequences of cine background images can be subtracted from moving contrast-filled vascular structures such as the coronary arteries of the beating heart. The technique is called Digital Subtraction Angiography (DSA). The subtraction is often made in real time while the contrast injection is being recorded. A computerized advantage is to be able to find automatically the optimal subtraction orientation of the two images, if a slight movement has occurred between the pre- and post-contrast image.

To manipulate an image

The possibilities of performing mathematical manipulations on digital images are more or less unlimited. In practice, only relatively few manipulations are used, primarily edge enhancement and contrast equalization. They are used to even out the contrast span over the whole image, to enhance contours that can be difficult to see, and to even out irregularities in homogenous structures. The reason for evening out the contrast span is to be able to assess equally structures that are located in very dark or very light areas on the original image.

 

Figure 6. Examples of simple radiological measurements performed on digital images by the modality computer. A) Distances 1, 2, and 3 within the chest are given in centimetres (DI) together with the angle in degrees between the line (indicating the distance) and the vertical direction (AN). B) Angles in degrees between the indicated lines are given on the image. C) The circular Region of Interest (ROI) in the lung indicates an average attenuation of-815 HU (ME) with an average pixel deviation from the mean of 70. 76 (SD). The surface of the circle is also calculated (AR) and is 30.73 cm2.

 

The methods for image manipulation are mathematically based on a recalculation of each pixel based on the values in surrounding pixels. Squares of pixels, e.g. 3 times 3 or 5 times 5, are used to influence the value of the middle pixel. Edge enhancement gives an image that mimicks the manipulation made by the eye sending visual information to the cerebral cortex. This means that the second derivative of the densitometric curve is used around an edge resulting in an extra bright zone on the light side of the edge and an extra dark zone on the dark side (Figs. 4 e, t).

Radiological measurements

The ruler and the protractor have always been radiological tools. Measurement in radiology was then advanced when the development of ultrasound introduced the ability to make simple measurements such as distances and angles between identified points on the image. This capability was especially important when no relative size estimate was available on the screen. Subsequently, area and examination- specific measurements were developed.

The same options are nowadays accessible for most screen oriented modalities independent of whether the images are analogue or digital. Simultaneous measurements of multiple distances (Fig. 6 a) and angles (Fig. 6 b) can be obtained. Both regularly and irregularly shaped surfaces can also be analysed with respect to area, mean dens it y (e.g. HU attenuation), and the standard deviation of the density (Fig. 6 c).
In the future the ability to make measurements on the image will be combined with normal values for the measurement related to measures such as patient age, height, or weight. It ought, for example to be possible to measure the projected area or volume of a kidney and relate the result to an appropriate parameter of body size. This should also be true for cardio-thoracic ratio or cardiac volume per square meter body surface area, etc.

PACS

As mentioned above this abbreviation means "Picture Archiving and Communications System". A first step in the development of PACS in a hospital or health care organization, is a HIS or "Hospital Information System". The corresponding system in a radiological department is called RIS or "Radiological Information System". Such computer systems contain data about the patient, e.g. name, address, previous examinations, modalities, and diagnoses, scheduled visits, referring physician, ward, etc. When linked to a PACS and the units that produce digital radiological images they form the basis of a digital radiological unit.

A PACS technically contains five parts:
1. The communication network with the image sources (Modalities)
2. A registry of examinations and patients and the archive for storage of such demographic data
3. Programs for implementing the demonstration and manipulation of images
4. An archive of all images
5. A unit for communication via a telephone connection or computer net to other digital systems

It is important that the patient administration system also keeps track of the digital images in order to maintain professional secrecy and order in the system. The access time for an examination in a PACS depends on many factors. This time is longer if multiple image sources are connected, if complex manipulations are made, or if frequent requests for previous examinations are made.

In a completely developed PACS, radiological conferences are performed on image screens rather than photographic films. During a transition period analogue films and image screens often coexist and both may be used during conferences.

Archival of digital images

Developments in the computer field have made it possible to store large numbers of digital images even if a very large memory capacity is necessary. A binary figure is called a bit. In most instances eight bits form one character (a decimal figure, a letter or comparable entity) and is called a byte. One kilobyte is 1024 bytes and a megabyte is about one million bytes in a computer. The hard disc of an ordinary personal computer contains of the order of 100 megabyte.

One image with a spatial resolution of 1000 x 1000 pixels uses one megabyte of memory with the possibility of 8 bite or 256 levels of contrast resolution. As larger radiological departments produce millions of images per year the required computer archive is enormous if all images are digital.

In order to reduce storage requirements the digital image information is normally compressed in one of two ways. In the first type of compression an image identical to the original can be obtained from reversibly compressed data. The gain in computer storage is about 50 to 70 % and the compression is in the order of two or three times. The reduction can be described in a simplified way as putting together all neighbouring pixels with the same value and storing them as one piece of information, indicating the start and the end of these pixels.

In the second type of compression the compression factor can be up to 40, or even more. This means, however, that the recalculated image differs somewhat from the original. In the case of higher compression ratios the differences between the original and final images may be marked and could influence their diagnostic usefulness. There is thus a balance between image quality on the one hand and computer memory on the other.

A second such balance exists between retrieval time and cost. If the image has to be retrieved immediately from the memory the cost is high, especially if the archive is large. Using tape, laser- or CD disks that might be searched for and mounted prior to image retrieval will increase the time cost but reduce the economic cost.

Image communication

Digital images in a PACS are transmitted between image producing modalities, image workstations, image screens for conferences, and computer archives. The large volumes of data makes great demands on the communication network. The network of special computer lines can be separate for PACS but sometimes when there are large distances between image producing units and computers and work-stations, the general computer network of the hospital is used. In such a case the demands for professional secrecy is high and the large amount of image data tend to block even medium capacity networks. In such cases optical cables can be used for transmission of images since such cables have large capacity and better security.

The first system for data communication in a hospital often deals with administrative data and comprises a computer in contact with a computer terminal. The components of PACS are so complex and take up so much computer memory that image communication is between computers. The main reason for this is the time needed to produce an image on screen. If each image were transmitted to the viewing station when it was requested, the time required would be unacceptably long. For this reason whole image packages are transmitted to the work station at the same time as the first image is made ready for viewing and manipulation.

Teleradiology

One extreme of image communication within PACS is teleradiology or transmission of digital radiological images between radiology departments or to a referring unit over the telephone network. It is not yet very common but can be used for consultations between radiologists or when radiological examinations are performed without a radiologist on site.

The radiological evaluation is made after image transmission over a telephone line. It is, however, important that the clinical data and other information is given verbally or in written form. One line of development is to use teleradiology to enable for the radiologist on call to perform most of his consultations at home. Another is to have sub specialized radiological service available for large are as via teleradiology. Radiological conferences with smaller referring units or practices with no radiologist can be performed without travel if the consultation is made over the telecommunication network.

Most current teleradiological systems are either connected to digital archives or to a video camera or laser digitiser that digitises an analogue film and records the data in a separate teleradiological memory. In the video camera case it is important for the quality of the diagnostic image that possible magnifications of parts of the original image are made through zooming with the video camera and not on the transmitted image. If the magnification is made on the teleradiologically transmitted image the spatial resolution is much lower.

The equipment on the receiving side depends upon the application. Normal and high resolution screens as well as laser printers for films can be used.

The digital radiological department

A digital radiology department only using digital images and screens would have a branched or circular network connecting all involved functions. These are l) image producing units (modalities), 2) image workstations, 3) the archive, and 4) a central or divided computer system.

The image producing units include MRI and CT machines, gamma cameras, digital ultrasound, and image plate systems. In addition, there are digitizing units where analogue images on film will be digitised into the PACS of the department. There is probably also a need for laser printers to produce analogue films to be sent outside the institution.

The image workstations are also of different types. The simplest requiring a minimum of computer power is directly connected to the examination room. It is used as a check to ensure that the image contains the appropriate part of the anatomy, correct projection, etc. The next type is used for demonstration during conferences and may consist of multiple screens placed to resemble a conventional film alternator. Conventional viewing boxes should also be available in the conference area to allow hard copy analogue film.

The third type of workstation is intended for the diagnostic work. The monitor screen has to be of high quality with good resolution and sufficient brightness and frame rate. The advantages of high spatial resolution is lost if the brightness is inadequate and will necessitate a higher monitor frame rate. The computer capacity must be sufficient to perform all types of image manipulations fast. The conferences can be prepared at this diagnostic workstation, with appropriate image manipulation, such as selecting relevant recording of magnified areas of importance, reduced number of, informative images of magnified areas, etc.

The PACS archive requirements differ for the patient/examination demographic database and for the digital image data. The most recent examinations ought to be immediately available and thereafter there is a progressively diminishing retrieval frequency of older images with time. It might be acceptable to wait for a couple of minutes for older films and even longer for educational cases, research material, etc.

 

Tatsuo Kumazaki and Hans Ringertz