Computer Methods and Programs in Biomedicine, 37 (1992) 299-304 ~ 1992 Elsevier Science Publishers B.V. All rights reserved (1169-26()7/92/$(15.(10
299
C O M M E T CPB01284
Information compression in dynamic radiological studies A a r o K i u r u I Erkki S v e d s t r 6 m 2 and H e i k k i K o r v e n r a n t a 3 I Central Laboratory, Meilahti Clinics, Helsinki Unicersity Central Hospital, Helsinki, Finland and Departments of : Diagnostie Radiology attd 3 Pediatrics, Unicersity of Turku, Turku, Finland
The large volumes of digital image data from radiological examinations demand further research and practical solutions in image compression before PACS solutions in large radiological departments are plausible. But another type of compression of image information is also possible, which has, in fact, been somewhat utilized already in analog form, but which has far better possibilities in the digital world. The number of essential images in dynamic X-ray, nuclear medicine, examinations etc. can be greatly reduced. These examinations producing image series are reviewed in terms of compression of information. 3-D displays are very useful in slice imaging, because they provide a means to see easier inside the human body than a number of slice images side by side. We also provide a new method, dynamic pulmonary imaging with digital fluoroscopy, as an example of the digital possibilities to compress a number of images into parametric images, numbers, histograms and curves. These processes also have other positive consequences information is transformed into a more easily perceptible form.
lnfl~rmation compression; Digital imaging; Image processing; Dynamic examination; Radiology
1. Introduction
The old phrase "an image tells more than thousand words" has an even more profound meaning when applied to the present digital processing techniques; images, series of images and volume data can be transformed into other media. An accustomed interpreter, say a radiologist, is trained to distinguish a signal in an image, rather than to obtain the same information from a corresponding verbal description alone. It is useful to include graphic presentations, and even characteristic parameters and numbers, to this easily perceptible world of images.
Correspondence: Aaro Kiuru, Central Laboratory, Meilahti Clinics, Helsinki University Central Hospital, SF-00290 Helsinkk Finland. Fax; 3580471 4016.
In modern medicine results and findings from patient examinations are expressed in numbers, curves, histograms or images and finally in words. Many clinical examinations exist; from physical palpation and interview through ECG measurements to complicated MR-spectroscopy and physiological PET-examination, where results are expressed in different forms, often as visual images. Medicine along with modern technology is opening a whole new world through image processing and the opportunity of changing image information into alternative media: functional or parametric images, images displayed side by side or over each other (from different modalities), 3-D display modes of volume data and even the utilization of artificial intelligence and expert systems [1] for better utilization of inherited knowledge. A digital system to generate, store, display, manipulate and transfer images (the data base
300
called PACS) together with other patient data (the data based called R1S) can be further connected to the general hospital database (HIS). The term radiology contains customarily all imaging modalities utilizing X-ray, magnetic resonance and ultrasound and sometimes also nuclear medicine. These procedures clearly produce the largest amount of image information in health care. This information will fairly soon be in digital form, when digital storage phosphor type radiography methods replace the analog screen-film systems [2,3]. The term medical imaging, on the other hand includes all images produced in medicine, that is in radiology, nuclear medicine, radiotherapy, pathology, cytology, etc. A medical dynamic imaging examination produces a series of (digital) images, especially when changes caused by a test probe (contrast media, radiopharmaceutical, etc.) or organ movements are measured as a function of time. A great number of characteristic dynamic examinations exist: CT, magnetic resonance, ultrasound and nuclear medicine imaging and in fluoroscopy. Radiological syntheses of long series of images have been made earlier using analog methods; today they are even more appealing using digital techniques. The synthesis can be made physically. actually producing new images, or mentally in the interpreter's brain. The latter method requires much experience and can be difficult to transfer to another interpreter. Therefore, the actual construction of new functional images has been found clinically valuable. In this article we review some of these examinations in terms of a natural compression of information. We consider briefly a new method, dynamic pulmonary imaging, as an example of the possibilities of compressing many images into parametric images and numbers.
2. Dynamic radiological studies Dynamic radiological examinations produce many images. The conventional fluorography of a dynamic heart study with cine filming makes a good example. Other examples are CT and MR image series with many slices and tens of megabytes of information. Due to the fairly small
Fig. 1. Gray scale presentation of colour Doppler capture technique from a femoral artery pseudoaneurysm. See text for details.
image matrix needed in nuclear medicine imaging there are many heart, kidney, brain, etc., examinations for which well-functioning and wellaccepted software analysis have already been in clinical use for a long time. Appropriate modelling of dynamic phenomena in tissues is often used in these cases to compare fitted parameters to normal values, to compress information and to express results in numerical or graphical forms. Ultrasound imaging (except Doppler technique) has not, however, up till now utilized similarly image processing methods, but will probably do so in the future. Figure 1 shows a gray scale presentation of so called colour Doppler capture technique from a femoral artery pseudoaneurysm (diameter about 5 cm) with simultaneous forward and backward flow displayed in colour. The pixels of the image contain the maximum flow values
301
during a measurement period of a few seconds. The dynamic information in over one hundred images has been squeezed to one image.
2.1. Fluoroscopic examinations The analysis of cine filming of stenotic coronary arteries has been the target of extensive research during the last two decades [4]. Today many commercial applications are in routine clinical use in heart catheterization laboratories. A typical dynamic cardiac fluoroscopic study with contrast medium measures the filling of coronary arteries and the function of heart ventricles. Images for the examination are taken, e.g., at the rate of 50 images/s for 10 s in, e.g., four series: this would represent roughly 2000 MB data with 512-matrix and 8 bit intensity (Table 1). The data analysis can produce parameter numbers describing the severity of stenosis and images showing the extent of the stenotic area, e.g., four representative 512-matrix images, all together about 1 MB. The ratio of the number of bytes in the original image series to that of the packed images, curves and parameter values altogether is in the order of 500:1. The process is irreversible. Table 1 shows this estimation together with those of other dynamic studies.
2.2. Slice imaging modalities Consecutive slices are measured in CT, MR, SPECT and P E T imaging; US imaging is also restricted to a thin tissue layer. The slice thick-
ness may vary from 0.5 to about 15 mm. CT and MR imaging is often repeated with contrast media, and a SPECT and PET study can be repeated to see the influence of administered drug, exercise, etc. Even with small image matrixes of 256 x 256 or 512 × 512, the amount of data is often very large. However, slice imaging does not utilize, at present, squeezing methods, in the way meant in this work. Certain studies however, like the measurements of contrast agent dynamics in a single CT slice of an organ have been analysed with, e.g., gamma variate function [5], which means the reduction of image data to one representative curve and corresponding parameters (Table 1). Compression ratio is about 5 : 1. In addition, fast tomographic imaging with spiral CT and ultrafast CT devices [6], as well as MR imaging with fast pulse sequences (utilizing also colour coding of blood flow) [7], are rapidly developing. It is obvious that methods to compress image series in these modalities also will be developed. Typical slice imaging modalities, such as 3-D image handling and display methods (with volumetric rendering, tissue characterization, etc.), are developing quickly with improving image work stations, display screens and software. In addition, modern means of combining information from many imaging modalities are offered by the use of multimedia data bases, still in their early development phase. Artificial intelligence and expert systems for facilitating clinical decision making are already in experimental use, but with restricted use as much further research is needed.
TABLE 1 Examples of compression of typical dynamic radiological examinations: number of images, their size, and corresponding storage capacities are shown The matrix size 5 1 2 x 5 1 2 with 16 bit depth is supposed in CT and MR. Nuclear medicine examinations produce, e.g., 128× 128 with g bit and angiographic examinations correspondingly 512 × 512 with 8 bit images. Examination
N u m b e r of images
Image size (MB)
MB/ EXAM.
Results (images, etc.)
Compression ratio
Angiogram CT SPECT Dynamic ventilation
4 X 500 20 20 16
0.26 0.07 0.016 0.006
500 1.3 0.32 1
4+ 4+ 2+ 4+
500 : 1 5:1 10 : 1 4:1
plots plots plots plots
302
2.3. Nuclear medicine examinations
Many fascinating methods to process a digital image or image series have been used in nuclear medicine in the last 20 years. These techniques have partly been utilized in other radiological examinations. Heart, kidney, brain, etc., examinations are performed either in dynamic mode in planar geometry or in SPECT slice imaging. There are measurement problems in SPECT connected to gamma ray scatter and attenuation, not found to that degree in, e.g., CT imaging. These have been solved only recently with improving computer power. Image processing methods in nuclear medicine are also improving continuously due to the same reason. PET technology with large computer capacity offers possibilities to assay physiological and metabolic pathways and therefore change biological events to images. SPECT imaging of the brain with radiopharmaceuticals like 99mTc-MHPAO is already in clinical use for the detection of neurological and psychiatric disorders. One way to present all slices simultaneously in one image is the so-called polar
Fig. 2. The polar plot or bull's eye view of a SPECT brain perfusion study with 9~r~Tc HM-PAO. The circles (starting from the centre) represent the horizontal slices (starting from vertex, left image); the front-back and right-left orientation is shown. In the right image the brain activity of this patient is compared to that of normal population.
plot or bull's eyes view. It is actually a map spreading function (Fig. 2), which also shows the patient data compared to normal values. Another example of a nuclear medicine dynamic examination is a cardiac study [3]. The
Fig. 3. Analysis of the series of 16 lung images (256 × 256 × 8 bit). The minimum and maximum values of each pixel were analysed resulting in two functional images. The left image corresponds to FRC and the right one to maximal lung volume during inspiration.
31/3 phase and amplitude analysis [8] is performed to an ECG-gated MUGA (MUltiple Gated Acquisition) series of images (originally 128 matrix, 8 bit images at the rate of 20/s for 600 s, about 0.32 MB, summed to 20 images). The analysis produces the phase and amplitude images representing one heart cycle with numbers describing ejection fraction, altogether about 40 kB. The compression ratio is about 10:1 (Table 1). The summed 20 images are also saved to represent the dynamics of the study.
2.4. Dynamic pulmonary examination A new fluoroscopic imaging device consisting of a microcomputer and a digital image memory unit has been used in dynamic pulmonary studies of experimental and clinical cases [9,10]. Series of digital images (normally 16 or 64) with 256 shades of gray are collected during one to three ventilation cycles at the rate of 6-25 images/s. Small variations in X-ray radiation transmitted through the lungs during tidal volume ventilation are measured. The system not only allows dynamic digital imaging but also regional physiological measurements of pulmonary ventilation. The method gives good dynamic information with adequate spatial resolution. The radiation dose is kept low due to the high kilovoltage and heavy beam filtration technique. The series of 16 or 64 images are analysed in a number of different ways. When a pixel or a small ROI is calculated throughout an image, its immediate surroundings can influence the pixel content. Minimum, maximum and level ((minimum + maximum)/2) images can be calculated. These represent physiological variable, lung volumes and ventilation dynamics. Figure 3 shows a typical normal patient analysis with the FRC (forced residual capacity) and maximal lung volume (during inspiration) images calculated from the series of 16 images. Also, subtracted time interval difference (TID-) images and images relative, for example, to the mean image of the cycle (REL-images) can be produced [9]. The series of images is evaluated dynamically using animation sequences and further analysed using region of interest techniques
resulting in, e.g., curves and histograms [9,10]. The squeezing ratio is, at the moment only reasonable; about 4 : 1 (Table 1).
3. Discussion
Digital methods in radiology allow the reduction of dynamic image data to numbers, curves, histograms, functional a n d / o r parametric images and 3-D displays by utilizing a great variety of already existing computer programs. This is routinely done (and it is a separate issue) prior to the actual compression (reversible or non-reversible) of single images or image series. These methods can be used automatically or interactively, e.g., in an image work station. The achievable compression ratio varies greatly depending on the application (Table 1). The original dynamic images are probably also saved due to their ability to show temporal changes in the study. However, they are probably not sent to the clinician. In addition, the real clinical value of information compression is not yet known, and methods can be expected to be developed much further. There are some drawbacks. It is possible to lose information while compressing a dynamic image series. This can especially be true if nonrecognizable artifacts exist in the series as well as during the development phase of these methods. The misinterpretation of an examination with this condensed data is possible. Compression methods are often irreversible and their interactive utilization in the selection of suitable parameters takes time. There are great potential benefits to be gained in PAC systems with the compression of information from dynamic imaging. Original images, compressed or non-compressed must probably be saved in a basic radiological archive, but a compressed set of images or parameter images with characterizing numbers and curves are faster to transfer to the clinician in a PACS-RIS-HIS network, when needed, and they are easier to comprehend. If an algorithm for information compression is accepted it can be fully automated. The greatest potential in the future is offered by the integration of images and compressed image
3(14 i n f o r m a t i o n with o t h e r p a t i e n t i n f o r m a t i o n from the medical record. T h e compression m e t h o d s also have benefits in education. T h e p r o p e r clinical use of images (in analogy with n u m e r i c a l results from hospital laboratory tests) would actually d e m a n d the c o m p a r i s o n of results with corres p o n d i n g normal values in a database, which in m a n y cases is not yet possible. A general target, therefore, is to be able to c o m b i n e radiological p l a n a r images, series of images a n d slice images with q u a n t i t a t i v e p a r a m e t e r values and curves when appropriate. A universal feature is therefore the a d d i n g of q u a n t i f i c a t i o n data to qualitative image i n f o r m a t i o n . This often causes c o n s i d e r a b l e r e d u c t i o n in data volume, shows the results in a smaller spatial and cond e n s e d t e m p o r a l d o m a i n a n d increases simultaneously the possibilities for easier a n d faster c o m p r e h e n s i o n of the results.
4. C o n c l u s i o n s W e want to stress the general use of P A C systems, not only for the g e n e r a t i o n , storage, displaying a n d t r a n s f e r r i n g of image i n f o r m a t i o n (together with other p a t i e n t data from and to RIS and HIS), but equally i m p o r t a n t l y to transform dynamic image series a n d image volume information into new forms of i n f o r m a t i o n , which are easier to c o m p r e h e n d . These p r o c e d u r e s can be p e r f o r m e d automatically or interactively, e.g., in an image workstation, this providing i n f o r m a t i o n which has great e d u c a t i o n a l benefits. A review of some existing clinical m e t h o d s to compress image series data has b e e n made. D y n a m i c p u l m o n a r y imaging with digital fluoroscopy is used as a new example to compress image series into p a r a m e t r i c images a n d n u m b e r s .
References [ll P.G. Mutalik, P.R. Fisher, G.G. Welter and H.A. Swett, Expert system advice as a by-product of image acquisition and reporting: obstacles to overcome, CAR'91, (1991), p. 315-3211 (Springer-Verlag, Berlin, 1991). [2] K. Doi, L. Lanzl and P.-J.P. Lin (Eds.), Recent developments in digital imaging, American Association of Physicists in Medicine, Medical Physics Monograph No. 12 (American Institute of Physics, New York, 1985). [3] S. Webb (Ed.), The physics of medical imaging, Medical Science Series (Adam Hilger, New York, 1988). [4] J.lq. Reiber, J.J. Gerbrands, G. Booman, G.J. Troost. A. den Boer, C.J. Slager and J.C. Schuurbiers, Objective characterization of coronary obstructions from monoplane cineangiograms and three-dimensional reeonstructkms of an arterial segment from two orthogonal views, in: Applications of computers in medicine. M.D. Schwartz, ed. 1EEE Cat. No. TH0095-0 (1982) 93-100. [5] L. Axel, P.B. Dean, A.A. Moss and D. Stansberry, Functional imaging of the liver: new information from dynamic CT, Invest. Radiol. 19 (1984) 23-29. [6] K.R. Peschmann, S. Napel, J.L. Couch, R.E. Rand, R. Alei, S.M. Ackelsberg, R. Gould and D.P. Boyd, High speed computer tomography: systems and performance, Appl. Opt. 24 (1985) 4052-4060. [7] S.J. Riederer, R.C. Wright, R.L. Ehman, P.J. Rossrnan, A.E. Holsinger-Bampton, N.J. Hangiandreou and R.C. Grimm, Real-time interactive colour flow MR imaging, Radiology 181 (1991) 33-39. [8] F. Cavailloles, J.P. Bazin and R. Di Paola, Factor analysis in gated cardiac studies, J Nucl. Med. 25 (1984) 1067-1079. [9] A. Kiuru, E. Svedstr6m and I. Kuuluvainen, Dynamic imaging of pulmonary ventilation, description of a novel digital fluoroscopy system. Acta Radiol. 32 (1990) 114119. [10] A. Kiuru, E. Svedstr6m and H. Korvenranta, I. Kuuluvainen, Dynamic pulmonary imaging: performance properties of a digital fluoroscopy system. Med. Phys. 19 (1992) 467-473.
299
C O M M E T CPB01284
Information compression in dynamic radiological studies A a r o K i u r u I Erkki S v e d s t r 6 m 2 and H e i k k i K o r v e n r a n t a 3 I Central Laboratory, Meilahti Clinics, Helsinki Unicersity Central Hospital, Helsinki, Finland and Departments of : Diagnostie Radiology attd 3 Pediatrics, Unicersity of Turku, Turku, Finland
The large volumes of digital image data from radiological examinations demand further research and practical solutions in image compression before PACS solutions in large radiological departments are plausible. But another type of compression of image information is also possible, which has, in fact, been somewhat utilized already in analog form, but which has far better possibilities in the digital world. The number of essential images in dynamic X-ray, nuclear medicine, examinations etc. can be greatly reduced. These examinations producing image series are reviewed in terms of compression of information. 3-D displays are very useful in slice imaging, because they provide a means to see easier inside the human body than a number of slice images side by side. We also provide a new method, dynamic pulmonary imaging with digital fluoroscopy, as an example of the digital possibilities to compress a number of images into parametric images, numbers, histograms and curves. These processes also have other positive consequences information is transformed into a more easily perceptible form.
lnfl~rmation compression; Digital imaging; Image processing; Dynamic examination; Radiology
1. Introduction
The old phrase "an image tells more than thousand words" has an even more profound meaning when applied to the present digital processing techniques; images, series of images and volume data can be transformed into other media. An accustomed interpreter, say a radiologist, is trained to distinguish a signal in an image, rather than to obtain the same information from a corresponding verbal description alone. It is useful to include graphic presentations, and even characteristic parameters and numbers, to this easily perceptible world of images.
Correspondence: Aaro Kiuru, Central Laboratory, Meilahti Clinics, Helsinki University Central Hospital, SF-00290 Helsinkk Finland. Fax; 3580471 4016.
In modern medicine results and findings from patient examinations are expressed in numbers, curves, histograms or images and finally in words. Many clinical examinations exist; from physical palpation and interview through ECG measurements to complicated MR-spectroscopy and physiological PET-examination, where results are expressed in different forms, often as visual images. Medicine along with modern technology is opening a whole new world through image processing and the opportunity of changing image information into alternative media: functional or parametric images, images displayed side by side or over each other (from different modalities), 3-D display modes of volume data and even the utilization of artificial intelligence and expert systems [1] for better utilization of inherited knowledge. A digital system to generate, store, display, manipulate and transfer images (the data base
300
called PACS) together with other patient data (the data based called R1S) can be further connected to the general hospital database (HIS). The term radiology contains customarily all imaging modalities utilizing X-ray, magnetic resonance and ultrasound and sometimes also nuclear medicine. These procedures clearly produce the largest amount of image information in health care. This information will fairly soon be in digital form, when digital storage phosphor type radiography methods replace the analog screen-film systems [2,3]. The term medical imaging, on the other hand includes all images produced in medicine, that is in radiology, nuclear medicine, radiotherapy, pathology, cytology, etc. A medical dynamic imaging examination produces a series of (digital) images, especially when changes caused by a test probe (contrast media, radiopharmaceutical, etc.) or organ movements are measured as a function of time. A great number of characteristic dynamic examinations exist: CT, magnetic resonance, ultrasound and nuclear medicine imaging and in fluoroscopy. Radiological syntheses of long series of images have been made earlier using analog methods; today they are even more appealing using digital techniques. The synthesis can be made physically. actually producing new images, or mentally in the interpreter's brain. The latter method requires much experience and can be difficult to transfer to another interpreter. Therefore, the actual construction of new functional images has been found clinically valuable. In this article we review some of these examinations in terms of a natural compression of information. We consider briefly a new method, dynamic pulmonary imaging, as an example of the possibilities of compressing many images into parametric images and numbers.
2. Dynamic radiological studies Dynamic radiological examinations produce many images. The conventional fluorography of a dynamic heart study with cine filming makes a good example. Other examples are CT and MR image series with many slices and tens of megabytes of information. Due to the fairly small
Fig. 1. Gray scale presentation of colour Doppler capture technique from a femoral artery pseudoaneurysm. See text for details.
image matrix needed in nuclear medicine imaging there are many heart, kidney, brain, etc., examinations for which well-functioning and wellaccepted software analysis have already been in clinical use for a long time. Appropriate modelling of dynamic phenomena in tissues is often used in these cases to compare fitted parameters to normal values, to compress information and to express results in numerical or graphical forms. Ultrasound imaging (except Doppler technique) has not, however, up till now utilized similarly image processing methods, but will probably do so in the future. Figure 1 shows a gray scale presentation of so called colour Doppler capture technique from a femoral artery pseudoaneurysm (diameter about 5 cm) with simultaneous forward and backward flow displayed in colour. The pixels of the image contain the maximum flow values
301
during a measurement period of a few seconds. The dynamic information in over one hundred images has been squeezed to one image.
2.1. Fluoroscopic examinations The analysis of cine filming of stenotic coronary arteries has been the target of extensive research during the last two decades [4]. Today many commercial applications are in routine clinical use in heart catheterization laboratories. A typical dynamic cardiac fluoroscopic study with contrast medium measures the filling of coronary arteries and the function of heart ventricles. Images for the examination are taken, e.g., at the rate of 50 images/s for 10 s in, e.g., four series: this would represent roughly 2000 MB data with 512-matrix and 8 bit intensity (Table 1). The data analysis can produce parameter numbers describing the severity of stenosis and images showing the extent of the stenotic area, e.g., four representative 512-matrix images, all together about 1 MB. The ratio of the number of bytes in the original image series to that of the packed images, curves and parameter values altogether is in the order of 500:1. The process is irreversible. Table 1 shows this estimation together with those of other dynamic studies.
2.2. Slice imaging modalities Consecutive slices are measured in CT, MR, SPECT and P E T imaging; US imaging is also restricted to a thin tissue layer. The slice thick-
ness may vary from 0.5 to about 15 mm. CT and MR imaging is often repeated with contrast media, and a SPECT and PET study can be repeated to see the influence of administered drug, exercise, etc. Even with small image matrixes of 256 x 256 or 512 × 512, the amount of data is often very large. However, slice imaging does not utilize, at present, squeezing methods, in the way meant in this work. Certain studies however, like the measurements of contrast agent dynamics in a single CT slice of an organ have been analysed with, e.g., gamma variate function [5], which means the reduction of image data to one representative curve and corresponding parameters (Table 1). Compression ratio is about 5 : 1. In addition, fast tomographic imaging with spiral CT and ultrafast CT devices [6], as well as MR imaging with fast pulse sequences (utilizing also colour coding of blood flow) [7], are rapidly developing. It is obvious that methods to compress image series in these modalities also will be developed. Typical slice imaging modalities, such as 3-D image handling and display methods (with volumetric rendering, tissue characterization, etc.), are developing quickly with improving image work stations, display screens and software. In addition, modern means of combining information from many imaging modalities are offered by the use of multimedia data bases, still in their early development phase. Artificial intelligence and expert systems for facilitating clinical decision making are already in experimental use, but with restricted use as much further research is needed.
TABLE 1 Examples of compression of typical dynamic radiological examinations: number of images, their size, and corresponding storage capacities are shown The matrix size 5 1 2 x 5 1 2 with 16 bit depth is supposed in CT and MR. Nuclear medicine examinations produce, e.g., 128× 128 with g bit and angiographic examinations correspondingly 512 × 512 with 8 bit images. Examination
N u m b e r of images
Image size (MB)
MB/ EXAM.
Results (images, etc.)
Compression ratio
Angiogram CT SPECT Dynamic ventilation
4 X 500 20 20 16
0.26 0.07 0.016 0.006
500 1.3 0.32 1
4+ 4+ 2+ 4+
500 : 1 5:1 10 : 1 4:1
plots plots plots plots
302
2.3. Nuclear medicine examinations
Many fascinating methods to process a digital image or image series have been used in nuclear medicine in the last 20 years. These techniques have partly been utilized in other radiological examinations. Heart, kidney, brain, etc., examinations are performed either in dynamic mode in planar geometry or in SPECT slice imaging. There are measurement problems in SPECT connected to gamma ray scatter and attenuation, not found to that degree in, e.g., CT imaging. These have been solved only recently with improving computer power. Image processing methods in nuclear medicine are also improving continuously due to the same reason. PET technology with large computer capacity offers possibilities to assay physiological and metabolic pathways and therefore change biological events to images. SPECT imaging of the brain with radiopharmaceuticals like 99mTc-MHPAO is already in clinical use for the detection of neurological and psychiatric disorders. One way to present all slices simultaneously in one image is the so-called polar
Fig. 2. The polar plot or bull's eye view of a SPECT brain perfusion study with 9~r~Tc HM-PAO. The circles (starting from the centre) represent the horizontal slices (starting from vertex, left image); the front-back and right-left orientation is shown. In the right image the brain activity of this patient is compared to that of normal population.
plot or bull's eyes view. It is actually a map spreading function (Fig. 2), which also shows the patient data compared to normal values. Another example of a nuclear medicine dynamic examination is a cardiac study [3]. The
Fig. 3. Analysis of the series of 16 lung images (256 × 256 × 8 bit). The minimum and maximum values of each pixel were analysed resulting in two functional images. The left image corresponds to FRC and the right one to maximal lung volume during inspiration.
31/3 phase and amplitude analysis [8] is performed to an ECG-gated MUGA (MUltiple Gated Acquisition) series of images (originally 128 matrix, 8 bit images at the rate of 20/s for 600 s, about 0.32 MB, summed to 20 images). The analysis produces the phase and amplitude images representing one heart cycle with numbers describing ejection fraction, altogether about 40 kB. The compression ratio is about 10:1 (Table 1). The summed 20 images are also saved to represent the dynamics of the study.
2.4. Dynamic pulmonary examination A new fluoroscopic imaging device consisting of a microcomputer and a digital image memory unit has been used in dynamic pulmonary studies of experimental and clinical cases [9,10]. Series of digital images (normally 16 or 64) with 256 shades of gray are collected during one to three ventilation cycles at the rate of 6-25 images/s. Small variations in X-ray radiation transmitted through the lungs during tidal volume ventilation are measured. The system not only allows dynamic digital imaging but also regional physiological measurements of pulmonary ventilation. The method gives good dynamic information with adequate spatial resolution. The radiation dose is kept low due to the high kilovoltage and heavy beam filtration technique. The series of 16 or 64 images are analysed in a number of different ways. When a pixel or a small ROI is calculated throughout an image, its immediate surroundings can influence the pixel content. Minimum, maximum and level ((minimum + maximum)/2) images can be calculated. These represent physiological variable, lung volumes and ventilation dynamics. Figure 3 shows a typical normal patient analysis with the FRC (forced residual capacity) and maximal lung volume (during inspiration) images calculated from the series of 16 images. Also, subtracted time interval difference (TID-) images and images relative, for example, to the mean image of the cycle (REL-images) can be produced [9]. The series of images is evaluated dynamically using animation sequences and further analysed using region of interest techniques
resulting in, e.g., curves and histograms [9,10]. The squeezing ratio is, at the moment only reasonable; about 4 : 1 (Table 1).
3. Discussion
Digital methods in radiology allow the reduction of dynamic image data to numbers, curves, histograms, functional a n d / o r parametric images and 3-D displays by utilizing a great variety of already existing computer programs. This is routinely done (and it is a separate issue) prior to the actual compression (reversible or non-reversible) of single images or image series. These methods can be used automatically or interactively, e.g., in an image work station. The achievable compression ratio varies greatly depending on the application (Table 1). The original dynamic images are probably also saved due to their ability to show temporal changes in the study. However, they are probably not sent to the clinician. In addition, the real clinical value of information compression is not yet known, and methods can be expected to be developed much further. There are some drawbacks. It is possible to lose information while compressing a dynamic image series. This can especially be true if nonrecognizable artifacts exist in the series as well as during the development phase of these methods. The misinterpretation of an examination with this condensed data is possible. Compression methods are often irreversible and their interactive utilization in the selection of suitable parameters takes time. There are great potential benefits to be gained in PAC systems with the compression of information from dynamic imaging. Original images, compressed or non-compressed must probably be saved in a basic radiological archive, but a compressed set of images or parameter images with characterizing numbers and curves are faster to transfer to the clinician in a PACS-RIS-HIS network, when needed, and they are easier to comprehend. If an algorithm for information compression is accepted it can be fully automated. The greatest potential in the future is offered by the integration of images and compressed image
3(14 i n f o r m a t i o n with o t h e r p a t i e n t i n f o r m a t i o n from the medical record. T h e compression m e t h o d s also have benefits in education. T h e p r o p e r clinical use of images (in analogy with n u m e r i c a l results from hospital laboratory tests) would actually d e m a n d the c o m p a r i s o n of results with corres p o n d i n g normal values in a database, which in m a n y cases is not yet possible. A general target, therefore, is to be able to c o m b i n e radiological p l a n a r images, series of images a n d slice images with q u a n t i t a t i v e p a r a m e t e r values and curves when appropriate. A universal feature is therefore the a d d i n g of q u a n t i f i c a t i o n data to qualitative image i n f o r m a t i o n . This often causes c o n s i d e r a b l e r e d u c t i o n in data volume, shows the results in a smaller spatial and cond e n s e d t e m p o r a l d o m a i n a n d increases simultaneously the possibilities for easier a n d faster c o m p r e h e n s i o n of the results.
4. C o n c l u s i o n s W e want to stress the general use of P A C systems, not only for the g e n e r a t i o n , storage, displaying a n d t r a n s f e r r i n g of image i n f o r m a t i o n (together with other p a t i e n t data from and to RIS and HIS), but equally i m p o r t a n t l y to transform dynamic image series a n d image volume information into new forms of i n f o r m a t i o n , which are easier to c o m p r e h e n d . These p r o c e d u r e s can be p e r f o r m e d automatically or interactively, e.g., in an image workstation, this providing i n f o r m a t i o n which has great e d u c a t i o n a l benefits. A review of some existing clinical m e t h o d s to compress image series data has b e e n made. D y n a m i c p u l m o n a r y imaging with digital fluoroscopy is used as a new example to compress image series into p a r a m e t r i c images a n d n u m b e r s .
References [ll P.G. Mutalik, P.R. Fisher, G.G. Welter and H.A. Swett, Expert system advice as a by-product of image acquisition and reporting: obstacles to overcome, CAR'91, (1991), p. 315-3211 (Springer-Verlag, Berlin, 1991). [2] K. Doi, L. Lanzl and P.-J.P. Lin (Eds.), Recent developments in digital imaging, American Association of Physicists in Medicine, Medical Physics Monograph No. 12 (American Institute of Physics, New York, 1985). [3] S. Webb (Ed.), The physics of medical imaging, Medical Science Series (Adam Hilger, New York, 1988). [4] J.lq. Reiber, J.J. Gerbrands, G. Booman, G.J. Troost. A. den Boer, C.J. Slager and J.C. Schuurbiers, Objective characterization of coronary obstructions from monoplane cineangiograms and three-dimensional reeonstructkms of an arterial segment from two orthogonal views, in: Applications of computers in medicine. M.D. Schwartz, ed. 1EEE Cat. No. TH0095-0 (1982) 93-100. [5] L. Axel, P.B. Dean, A.A. Moss and D. Stansberry, Functional imaging of the liver: new information from dynamic CT, Invest. Radiol. 19 (1984) 23-29. [6] K.R. Peschmann, S. Napel, J.L. Couch, R.E. Rand, R. Alei, S.M. Ackelsberg, R. Gould and D.P. Boyd, High speed computer tomography: systems and performance, Appl. Opt. 24 (1985) 4052-4060. [7] S.J. Riederer, R.C. Wright, R.L. Ehman, P.J. Rossrnan, A.E. Holsinger-Bampton, N.J. Hangiandreou and R.C. Grimm, Real-time interactive colour flow MR imaging, Radiology 181 (1991) 33-39. [8] F. Cavailloles, J.P. Bazin and R. Di Paola, Factor analysis in gated cardiac studies, J Nucl. Med. 25 (1984) 1067-1079. [9] A. Kiuru, E. Svedstr6m and I. Kuuluvainen, Dynamic imaging of pulmonary ventilation, description of a novel digital fluoroscopy system. Acta Radiol. 32 (1990) 114119. [10] A. Kiuru, E. Svedstr6m and H. Korvenranta, I. Kuuluvainen, Dynamic pulmonary imaging: performance properties of a digital fluoroscopy system. Med. Phys. 19 (1992) 467-473.
