- Email: [email protected]
The role of computers in the nuclear medicine laboratories
T h e R o l e o f C o m p u t e r s in t h e N u c l e a r Medicine Laboratories H. I. Glass There are several methods of using c o m p u t e r s in nuclear medicine laboratories. These include the use of special-purpose computers on individual apparatus, linking of apparatus t o a central computer, or the processing of all data off-line on a batch-processing computer. The range of applications of computers in nuclear m e d i cine is examined, and a s u g g e s t i o n con-
cerning the m i n i m u m specification of a centralized nuclear medicine system is proposed. There would appear to be little advantage to be gained at the present t i m e by introducing into nuclear medicine departments computers t h a t w i l l not permit e x p a n s i o n in t h e f u t u r e t o this minim u m proposed specification.
ANY OF THE DATA acquired during investigations carried out in nuclear mediine are amenable to analysis by digital computers. There is little doubt, however, that two comparatively recent developments have been the prime reasons for the increasing demands for the installation of digital computers in departments of nuclear medicine. These are (1) the use of gamma cameras for dynamic studies and (2) the use of automatic sample-counting equipment for numerous hormone and steroid saturation analysis and immunoassay procedures. Two of the special problems in nuclear medicine are the wide variety of activities that fall under this broad heading and the differences in the range of activities performed in individual departments. This results in wide differences in requirements for a minimum computer system to satisfy all physicians working in this area. The continuing reduction in cost of electronic components and the improvements in their performance, specification, and reliability have brought with them a great increase in the use and application of small computers. There are several possible approaches to the problem: (1) use small special-purpose computers on each individual apparatus, (2) provide computer-compatible output on each device for off-line programming on a central system, or (3) attempt to run all the devices on-line to a central computer operating under some form of time sharing or real-time executive system in which each device effectively has full access to a computer system at all times. The relative merits of each of these broad categories will be considered after a general survey of the applications in nuclear medicine which utilize computers (Table 1). This survey will indicate the wide range of data input rates and quantity of data obtained from the equipment normally found in nuclear medicine departments. RADIONUCLIDE SAMPLE ASSAY
Automatic liquid scintillation and automatic gamma counters frequently have either a punched-paper-tape output, a programmable calculator, or an internal computer attached to the machine. The data-processing problem usually involves fairly simple re-
From the Diagnostic Physics Unit, Department of Medical Physics, Royal Postgraduate Medical School Hammersmith Hospital, London, England. H. I. Glass, M.A., Ph.D., F.lnst.P.: Head, Diagnostic Physics Unit, Department of Medical Physics, Royal Postgraduate Medical School, Hammersmith Hospital London, W12, England. 9 1973 by Grune & Stratton, Inc. Seminars in Nuclear Medicine, Vol. 3, No. 4 (October), 1973
303
304
H . I . GLASS Table 1. Fields of Application of Computers in Nuclear Medicine Assay of radioactivity Blood flow Function studies Metabolic studies Whole-body counting Tumor detection Clinical records
Collimator design Activation analysis I mage processing Quantitative scanning Quantitative scintigraphy Radionuclide dosimetry Radioimmunoassay
petitive arithmetic, as in the case of counting multiple gamma-emitting radionuclides, i Occasionally, however, the computer program is also required to carry out curve-fitting and integration procedures. These may arise in liquid scintillation counting utilizing one of several possible quench-correction procedures. The curve normally has a polynomial form, and many commercial counters contain small computers for this purpose. In the immunoassay field, the standard curve that requires to be fitted and interrogated approximates closely, in most assays, to a probit curve; a curve of this type must be constructed for each assay. The content of the unknown samples is estimated by reference to this standard curve. Numerous computer programs exist for carrying out a least-squares fitting procedure to the probit or logit function, possibly the most sophisticated being that described by Rodbard. 2 There would appear to be little or no possibility of automating these more sophisticated calculation procedures with computers of less than 12K 16-bit capacity which may be operated on-line to the automatic counter or off-line in a batch-processing mode. The more sophisticated programs would probably require even more computing power, and if used on-line, the computer might be expected to form part of a system which might receive data from more than one counting apparatus, and which would, in turn, require a further increase in core size of the computer in order to operate in this multiple-input mode. Since the rate of data input is relatively low and the amount of data to be stored and processed is relatively small also, several counters may be multiplexed together using the same input device to the computer, provided some elementary form of buffering is provided. BLOOD FLOW STUDIES
In blood flow studies, the computer may be required to act as an acquisition system from several probes. 3 Immediate presentation of a preliminary analysis is followed by a more detailed analysis, which might consist of a mathematical least-squares fit to the inflow data or clearance data, or possibly both. This analysis often consists of an iterarive procedure such as multiexponential curve fitting. It may be even more complex, involving, for example, a deconvolution procedure. This situation arises, for instance, during the use of an inhalation method for blood flow studies with inert gases. This type of data processing is probably best performed on an off-line system, but the potential of simpler analog methods in this context should not be overlooked. 4 Blood flow studies may also be carried out on the gamma camera, which may have some advantages; however, there are frequently count-rate difficulties with the gamma camera, and in addition the data processing is much more complex (see the article by Brill in the July, 1973, issue). Although small on-line computers may be excellent as dataacquisition buffers, it seems unlikely that an adequate analysis of the data would be possible if more than about six probes were used without the addition of expensive peripheral equipment.
COMPUTERS IN NUCLEAR MEDICINE
305
METABOLIC, TRACER KINETIC, AND FUNCTION STUDIES
In the field of metabolic and tracer kinetic studies, one is concerned with formulating mathematical models to permit the study of metabolism and turnover of various substances or drugs. Simple models have of necessity been postulated, since the quality and scarcity of the data frequently prevent a more detailed analysis to be attempted. The analysis is usually based on the solution of linear or nonlinear differential equations by numerical methods. The analysis is often complicated by a continuously varying input function or by the introduction of variable time delays between different physiological compartments. Several programs have been written for individual systems, and in most cases it would appear more economical to produce a special program to solve a particular kinetic problem. The elaborate program by Berman s permits the study and evaluation of most of the possible relevant models, but it requires a very large installation. The power, versatility, and ease of use of the Berman program, which exists in versions suitable for various computers, render it always worthy of consideration for any problem of this type. For a simpler function study like renography, lung function, liver function, or simple blood flow studies such as cardiac output measurements, a small on-line system would cope adequately with data and might well be the most economical approach, provided the number of patients undergoing this type of investigation is sufficiently high. Once again, the data-acquisition rate presents no major interfacing problem for most computer systems in this type of study. DYNAMIC IMAGING STUDIES
It is a common experience that in many nuclear medicine departments the acquisition of a computer is initiated by the arrival of a gamma camera. In earlier days, several commercial digital gamma-camera systems were produced, all of which suffered from two major defects: (1) the limited power and versatility of the computer and (2) the unsatisfactory nature of the associated display. Recently several gamma-camera systems have become available that have overcome these problems by using more elaborate displays associated with a computer with a core storage of 12K or more and operating with a magnetic disc. Since these are potentially versatile computer systems, the possibility of using such systems for other applications has frequently been raised. This is particularly pertinent since many of the quantitative dynamic studies that were originally envisaged did not yield immediate significant improvements in clinical diagnosis. Budinger 6 has recently reviewed this area with particular reference to one commercial system. Besides these commercial systems, several elaborate and powerful systems have been described with the basic objective of quantitative processing of radionuclide images. 7-9 (See also the article by Brown et al. in this issue.) In most cases, however, the complexity of the clinical problem that has been tackled is an order of magnitude less than the complexity of the system being used to solve it. In the more complex clinical problems, few significant advances appear to have been described up to now. One case where promising results have been obtained is in the assessment of coronary blood flow ~~ by a combination of quantitative uptake estimation and image processing, as was discussed in an earlier issue. STATIC IMAGING The most obvious application of a computer/display system to imaging is the correction of the image for nonuniformity of response of the gamma camera. This technique
306
H . I . GLASS
has not been widely applied for two reasons: (1) the inferior quality of the device on which the processed display is normally presented and (2) the inconvenience of doing this correction routinely on all images. The most effective way of dealing with this problem is to cure it at its source by examining the uniformity at frequent intervals and correcting for any errors by electronic adjustment. Not every group is prepared to carry out this occasionally complex adjustment, while the number of gamma cameras now in operation renders it inevitable that a significant number of cameras are being operated under conditions that are not optimal. The diagnostic errors due to nonuniformity are frequently underestimated, and a routine computer correction of the image is desirable. Image processing by the use of filtering and enhancement procedures has mainly been carried out off-line or on large batch-processing machines, 1~'12 which are often not equipped with suitable display facilities or methods of producing a suitable permanent record. Despite the advent of fast Fourier-transform algorithms, ~3 it seems unlikely that routine image processing will be carried out in any but the larger computer/ display installations. It is important that a scanner be equipped with a digital dataacquisition system (either paper tape or magnetic tape) so that the data may be processed off-line if necessary, with the results then replayed through a suitable display attached to a smaller local machine. One major disadvantage with conventional scanners is the requirement of presetting the operating conditions. In this respect, a commercial system now available seems to offer a significant advance in this field. ~4 This system uses a semiconductor memory to acquire the data, which permits the data to be displayed in an optimum way and to be replayed rapidly onto photographic film. In the case of tomographic scanning, simple data acquisition is inadequate if immediate processed display is required. The earlier mathematical treatment used by Kuhl for processing tomographic images has now been superseded by more complex mathematical analyses ~s'16 requiring fairly sophisticated on-line or off-line computer/display systems. The advent of a recent commercial tomographic scanning system (J and P Engineering, Ltd., Reading, England) has increased the possibility of fully utilizing the commercial image acquisition/processing/display systems that are now available. There seems little or no advantage in using elaborate digital systems for simple qualitative static or dynamic studies on cameras or scanners where cheaper analog magnetic-tape systems, operating in either video or pulse recording modes, are adequate for reproducing the images. A further application of computers to scanning is associated with the increasing use of dual-detector scanners for cluantitative uptake studies for assessing the distribution of new radiopharmaceuticals by whole-body scanning and also in clinical applications, one example of which is the routine estimation of splenic blood volume. 17 These techniques are tedious to perform and analyze by hand, and computer processing of the data is essential. WHOLE-BODY COUNTING
After a period in which the use of whole-body counters in clinical diagnosis had stabilized, it is apparent that recently there has been renewed interest la (see also the article by Andrews et al. in this issue). In some investigations, such as multiple-tracer studies, the availability of a computer is an advantage, and it becomes especially useful if
COMPUTERS IN NUCLEAR MEDICINE
307
spectrum stripping routines must be applied. This type of analysis is easily dealt with as an off-line batch-processing problem provided suitable data-acquisition equipment is available. More recently, the technique of in-vivo neutron activation analysis has been described 19 for the analysis of trace elements in the body. The data analysis and calibration procedures in this technique certainly necessitate the use of spectral analysis by digital computer programs, owing to the complex spectra obtained.
RADIONUCLIDE DOSlMETRY AND LABORATORY RECORDS As reported by Brown, 2~ many laboratories use computers for record keeping and record searching of the procedures carried out. In some laboratories these are also used to prepare reports for inclusion in patients' case notes. The use of computers for more detailed radionuclide dosimetry calculations reflects the continuing concern with the reduction of radiation dose in nuclear medicine procedures and the improved knowledge of the biological turnover and metabolic pathways of the new and increasingly complex radiopharmaceuticals. These dosimetry estimates often require the use of lengthy Monte Carlo calculations, which can be most effectively performed on a large batch-processing computer. The recent revision of the finger dose obtained due to repeated handling of 99mTc strikes a cautionary note. 21 IS THERE AN OPTIMUM SYSTEM? There seems little doubt that the conclusion reached by Cox 22 in 1968 concerning the use of specialized computer systems is still valid today. Briefly summarized, he suggested that significant economics can be achieved by the use of built-in specially programmed small computers to do a specific job. The cost of core memory has been falling steadily for many years and continues to fall. This is reflected in the increasing sophistication of many devices with little or no increase in cost. For example, several commercial multichannel analyzers now incorporate small (up to gK) digital computers as an integral part of the system. These computers are available for use in their own right for other purposes if required. All the usual multichannel analyzer facilities are achieved by software. The disadvantage of the smaller, specially tailored systems is that the economics are only real if the devices are fully used, and that means very frequently used. If this is not likely to be the case, then the slightly less specialized system, exemplified by the computer/multichannel analyzer, is likely to be more useful, since it is more versatile, and this in itself results in increased use. Increased utilization associated with increased versatility brings its own disadvantages, since the equipment is often not available when needed for its original purpose. The way out of this dilemma appears to be to obtain a minimum number of specialpurpose instruments and to link as many pieces of equipment as possible to a minimumsize central computer system equipped specifically with peripherals required for nuclear medicine. This computer system should operate under some real-time executive software system. This would allow simultaneous input from any of the linked apparatus, unless a priority system specifically excludes apparatus in certain special circumstances, e.g., to allow temporarily high data rates to be accepted from a gamma-camera study. Real-time executive programs normally allow running foreground programs or acquisition routines concurrently with background programs that are in effect carrying out batch processing. In this way the system can cope with the wide range of data input rates derived from nuclear medicine equipment, from the slowest rates in radio-
308
H.I. GLASS
nuclide sample counting machines to the higher rates obtained from scanners and the even higher rates encountered in blood flow and function studies using open detectors or the very fast rates obtained on gamma cameras. The other specific requirement for a nuclear medicine department computer system is a digitally controlled display system. It is now possible to buy such systems independently of the computer, although as mentioned above, several such displays are now available from computer manufacturers. These may take the form of television displays driven directly from core or through an intermediate buffering device such as a drum or small disc. SPECIFICATION OF A SUITABLE SYSTEM
To run a computer under a real-time executive system with the number of inputs likely to be encountered in an average-size nuclear medicine department, a minimum system requirement would be a core memory size of 16K plus a magnetic-disc memory. A dual-disc system (one of fixed-head design with a capacity of about 500K words and one exchangeable disc design with a capacity of about 2.5M words) is probably necessary. As the number of input devices increases, an extra 8K words of memory would become highly desirable. To allow for the possibility of off-line processing on larger machines and also as a convenient mass-storage medium, a magnetic-tape handler is useful, although a paper-tape punch may be adequate in some circumstances as a computer-compatible output. Currently, nine-track magnetic-tape handlers are increasingly being used because of their increased storage capacity over seven-track systems; however, many large batch-processing systems will still accept only the latter. The usual magnetic-tape specification offers 45-ips tape speed and 800-cpi tape density; however, 1600-cpi densities are now obtainable on some magnetic-tape handlers. A paper-tape reader and keyboard terminal would also be required. The display should permit the presentation of a picture matrix such that each element corresponds to about one-eighth of the detector resolution (FWHM) of a radionuclide imaging system. In most cases this will result in an image in which the individual picture elements are virtually indistinguishable on a 12-cm oscilloscope. The monochrome display should allow the presentation of a minimum of 16 brightness levels. In the case of those devices where the data input rate is very low, it is possible that a multiplexer which works into a single ADC may result in some financial saving. Similarly; it may be more efficient to batch-process the automatic sample counter output and arrange for the data to be obtained in cheap computer-compatible form such as paper tape. A suggested nuclear medicine system is shown in Fig. 1.
I DISPLAY I 128x 128 . IMEDICINE NUCLEAR I EQUIPMENT
t
16K;16b=t:~ I ~sec
L
ITELETYPE 1
TAPEUNIT
~IIDUAL m m~AG95r~[woris ~ PAPER TAPE READER 400cps I Fig.1.
COMPUTERS IN NUCLEAR MEDICINE
309
DISCUSSION
The computer system which has been proposed might at first appear elaborate and expensive, but it should be considered in the light of the cost of the associated equipment which is normally present in nuclear medicine departments. The presence of at least three imaging devices in a nuclear medicine department is not unusual, and the total cost of the computer and its peripherals shown in Fig. 1 is approximately onethird of that of the nuclear medicine equipment. Furthermore, the recurring cost of radiopharmaceuticals is certainly not insignificant, whereas the running cost of the computer system is essentially its maintenance cost, which may be estimated at approximately 5% of its capital value. With a computer system of this level of complexity, it should be recognized that a scientist or engineer will almost certainly have to be employed if the system is to be effectively utilized, since it is virtually certain that software will not be available from manufacturers to satisfy all the applications likely to be devised in nuclear medicine departments. A complete clinically oriented system having a minimum of knobs and having been designed for maximum ease of use by the clinician would obviously be desirable. However, it seems fair to point out that the term nuclear medicine means all things to all men at the present time, and the activities of individual nuclear medicine departments vary considerably. The possibility of producing a single system with a single set of software which will be acceptable or even useful in all nuclear medicine departments is remote. In the gamma-camera field, at least five commercial computer/display systems are available, all having advantages and disadvantages. It would seem unlikely that a general clinical computer system with packaged software will be available for some time. If it is economically possible, as much data collection and data analysis as possible should be done on-line, as this is without question the most convenient and efficient mode of operation for clinical departments. Whether this is done by use of small local data buffers attached to each device or by a direct on-line link to a larger system is a local decision of logistics, economics, and policy. In general, where batch processing is convenient and acceptable, it should be performed. However, it is extremely important when installing a computer to serve a nuclear medicine department that (1) it should be expandable to at least 32K, (2) it should ultimately be capable of operation under a real-time executive system, (3) it should be equipped with a versatile display, and (4) it should be capable of producing a computer-compatible output (preferably magnetic tape) to allow off-line batch processing of more difficult problems on a larger, faster machine. When building up a system, it is advisable to allow for this expansion possibility. REFERENCES
1. Skrabal F, Arnot RN, Helus F, et al: A method for simultaneous electrolyte investigations in man using bromine-77, potassium-43, and sodium-24. Int J Appl Radiat Isot 21:183, 1970 2. Rodbard D, Lewald JE: Computer analysis of radioligand assay and radioimmunoassay data, in: Proceedings of 2nd Symposium on Research Methods in Reproductive Endocrinology. Stockholm, Karolinska Institute, 1970, p 79
3. Sveindottir E, Lassen NA, Risberg J, Ingvar DH: Regional cerebral blood flow measured by multiple probes: An oscilloscope and digital computer system for rapid data analysis, in Brock M, Feschi C, Ingvar D, et al, (eds): Cerebral Blood Flow. Berlin, Springer-Verlag, 1969, p 27 4. Veall N, Crawley JCW: The 133Xe inhalation technique. Scand J Clin Lab Invest Suppl 102, XI:F, 1968 5. Berman M: The application of multicom-
310
partmental analysis to problems of clinical medicine. Ann Intern Med 68:423, 1968 6. Budinger TF: Clinical and Research Quantitative Nuclear Medicine System, in: Medical Radioisotope Scintigraphy. Vienna, IAEA, 1973 7. Natarajan TK, Wagner HN Jr: A new image display and analysis system (IDA) for radionuclide imaging. Radiology 93:823, 1969 8. Brownell GL, Burnham CA: Numedics, a computer system for processing radioisotope data from multiple sources, USAEC Conf.710425. Oak Ridge, Tenn., USAEC, 1971, p 215 9. Brill AB, Erikson JJ, Lindahl CE: Digital systems for acquisition and storage, in Kenny PJ, Smith EM (eds): Quantitative Organ Visualization in Nuclear Medicine. Coral Gables, Fla., University of Miami Press, 1971, p 339 10. Mclntyre WJ, Bott RE, lshii Y, Pritchard WH: Atraumatic measurement of the distribution of myocardial blood flow using 43K and a scintillation camera, in: Medical Radioisotope Scintigraphy. Vienna, IAEA, 1973 11. Nagai T, Fukuda N, linuma TA: Computer focussing using an appropriate gaussian function. J Nucl Med 10:209, 1969 12. Brown DW: Computer processing of radioisotope scans using Fourier and other transformations. J Nucl Med 11:304, 1970 13. Cooley JW, Tukey JW: An algorithm for the machine calculation of complex Fourier series. Mathematics of Computations 19:297, 1965. 14. Becker J, Graul EH: Computer evaluation and colour display of conventional scinti-
H.I.
GLASS
grams. Compact News in Nuclear Medicine 3: 55, 1972 15. Kuhl DE, Edwards RQ, RicciAR, Reivich M: Quantitative section scanning using orthogonal tangent correction. J Nucl Med 13:447, 1972 16. Chesler DA: Three dimensional reconstruction technique. (to be published) 17. Hegde VM, Williams ED, Lewis SM, et al: Measurement of splenic red cell volume and visualization of the spleen with Technetium99m. ~to be published) 18. Reizenstein P: Clinical Whole Body Counting. Bristol, U.K., John Wright and Son, 1973 19. Cohn SH, Shukla KK, Dombrowskie CS, Fairchild RG: A total body neutron activation facility employing portable a-n sources for medical research, in: Proceedings of IAEA Expert Panel on ln-Vivo Activation Analysis. Vienra, IAEA (in press) 20. Brown DW, Kirch D, Groome D, et al: General purpose computers for processing of image data, in Kenny PJ, Smith EM (eds): Quantitative Organ Visualization in Nuclear Medicine. Coral Gables, Fla., University of Miami Press, 1971, p 431 21. Henson PW: A note of some aspects of skin contamination by certain radionuclides in common use. Br J Radiol 45:938, 1972 22. Cox JR: Economy of Scale and Specialization in Large Computing Systems. Monograph 93, Biomedical Computing Laboratory, Washington University, St. Louis, Mo., 1968
cerning the m i n i m u m specification of a centralized nuclear medicine system is proposed. There would appear to be little advantage to be gained at the present t i m e by introducing into nuclear medicine departments computers t h a t w i l l not permit e x p a n s i o n in t h e f u t u r e t o this minim u m proposed specification.
ANY OF THE DATA acquired during investigations carried out in nuclear mediine are amenable to analysis by digital computers. There is little doubt, however, that two comparatively recent developments have been the prime reasons for the increasing demands for the installation of digital computers in departments of nuclear medicine. These are (1) the use of gamma cameras for dynamic studies and (2) the use of automatic sample-counting equipment for numerous hormone and steroid saturation analysis and immunoassay procedures. Two of the special problems in nuclear medicine are the wide variety of activities that fall under this broad heading and the differences in the range of activities performed in individual departments. This results in wide differences in requirements for a minimum computer system to satisfy all physicians working in this area. The continuing reduction in cost of electronic components and the improvements in their performance, specification, and reliability have brought with them a great increase in the use and application of small computers. There are several possible approaches to the problem: (1) use small special-purpose computers on each individual apparatus, (2) provide computer-compatible output on each device for off-line programming on a central system, or (3) attempt to run all the devices on-line to a central computer operating under some form of time sharing or real-time executive system in which each device effectively has full access to a computer system at all times. The relative merits of each of these broad categories will be considered after a general survey of the applications in nuclear medicine which utilize computers (Table 1). This survey will indicate the wide range of data input rates and quantity of data obtained from the equipment normally found in nuclear medicine departments. RADIONUCLIDE SAMPLE ASSAY
Automatic liquid scintillation and automatic gamma counters frequently have either a punched-paper-tape output, a programmable calculator, or an internal computer attached to the machine. The data-processing problem usually involves fairly simple re-
From the Diagnostic Physics Unit, Department of Medical Physics, Royal Postgraduate Medical School Hammersmith Hospital, London, England. H. I. Glass, M.A., Ph.D., F.lnst.P.: Head, Diagnostic Physics Unit, Department of Medical Physics, Royal Postgraduate Medical School, Hammersmith Hospital London, W12, England. 9 1973 by Grune & Stratton, Inc. Seminars in Nuclear Medicine, Vol. 3, No. 4 (October), 1973
303
304
H . I . GLASS Table 1. Fields of Application of Computers in Nuclear Medicine Assay of radioactivity Blood flow Function studies Metabolic studies Whole-body counting Tumor detection Clinical records
Collimator design Activation analysis I mage processing Quantitative scanning Quantitative scintigraphy Radionuclide dosimetry Radioimmunoassay
petitive arithmetic, as in the case of counting multiple gamma-emitting radionuclides, i Occasionally, however, the computer program is also required to carry out curve-fitting and integration procedures. These may arise in liquid scintillation counting utilizing one of several possible quench-correction procedures. The curve normally has a polynomial form, and many commercial counters contain small computers for this purpose. In the immunoassay field, the standard curve that requires to be fitted and interrogated approximates closely, in most assays, to a probit curve; a curve of this type must be constructed for each assay. The content of the unknown samples is estimated by reference to this standard curve. Numerous computer programs exist for carrying out a least-squares fitting procedure to the probit or logit function, possibly the most sophisticated being that described by Rodbard. 2 There would appear to be little or no possibility of automating these more sophisticated calculation procedures with computers of less than 12K 16-bit capacity which may be operated on-line to the automatic counter or off-line in a batch-processing mode. The more sophisticated programs would probably require even more computing power, and if used on-line, the computer might be expected to form part of a system which might receive data from more than one counting apparatus, and which would, in turn, require a further increase in core size of the computer in order to operate in this multiple-input mode. Since the rate of data input is relatively low and the amount of data to be stored and processed is relatively small also, several counters may be multiplexed together using the same input device to the computer, provided some elementary form of buffering is provided. BLOOD FLOW STUDIES
In blood flow studies, the computer may be required to act as an acquisition system from several probes. 3 Immediate presentation of a preliminary analysis is followed by a more detailed analysis, which might consist of a mathematical least-squares fit to the inflow data or clearance data, or possibly both. This analysis often consists of an iterarive procedure such as multiexponential curve fitting. It may be even more complex, involving, for example, a deconvolution procedure. This situation arises, for instance, during the use of an inhalation method for blood flow studies with inert gases. This type of data processing is probably best performed on an off-line system, but the potential of simpler analog methods in this context should not be overlooked. 4 Blood flow studies may also be carried out on the gamma camera, which may have some advantages; however, there are frequently count-rate difficulties with the gamma camera, and in addition the data processing is much more complex (see the article by Brill in the July, 1973, issue). Although small on-line computers may be excellent as dataacquisition buffers, it seems unlikely that an adequate analysis of the data would be possible if more than about six probes were used without the addition of expensive peripheral equipment.
COMPUTERS IN NUCLEAR MEDICINE
305
METABOLIC, TRACER KINETIC, AND FUNCTION STUDIES
In the field of metabolic and tracer kinetic studies, one is concerned with formulating mathematical models to permit the study of metabolism and turnover of various substances or drugs. Simple models have of necessity been postulated, since the quality and scarcity of the data frequently prevent a more detailed analysis to be attempted. The analysis is usually based on the solution of linear or nonlinear differential equations by numerical methods. The analysis is often complicated by a continuously varying input function or by the introduction of variable time delays between different physiological compartments. Several programs have been written for individual systems, and in most cases it would appear more economical to produce a special program to solve a particular kinetic problem. The elaborate program by Berman s permits the study and evaluation of most of the possible relevant models, but it requires a very large installation. The power, versatility, and ease of use of the Berman program, which exists in versions suitable for various computers, render it always worthy of consideration for any problem of this type. For a simpler function study like renography, lung function, liver function, or simple blood flow studies such as cardiac output measurements, a small on-line system would cope adequately with data and might well be the most economical approach, provided the number of patients undergoing this type of investigation is sufficiently high. Once again, the data-acquisition rate presents no major interfacing problem for most computer systems in this type of study. DYNAMIC IMAGING STUDIES
It is a common experience that in many nuclear medicine departments the acquisition of a computer is initiated by the arrival of a gamma camera. In earlier days, several commercial digital gamma-camera systems were produced, all of which suffered from two major defects: (1) the limited power and versatility of the computer and (2) the unsatisfactory nature of the associated display. Recently several gamma-camera systems have become available that have overcome these problems by using more elaborate displays associated with a computer with a core storage of 12K or more and operating with a magnetic disc. Since these are potentially versatile computer systems, the possibility of using such systems for other applications has frequently been raised. This is particularly pertinent since many of the quantitative dynamic studies that were originally envisaged did not yield immediate significant improvements in clinical diagnosis. Budinger 6 has recently reviewed this area with particular reference to one commercial system. Besides these commercial systems, several elaborate and powerful systems have been described with the basic objective of quantitative processing of radionuclide images. 7-9 (See also the article by Brown et al. in this issue.) In most cases, however, the complexity of the clinical problem that has been tackled is an order of magnitude less than the complexity of the system being used to solve it. In the more complex clinical problems, few significant advances appear to have been described up to now. One case where promising results have been obtained is in the assessment of coronary blood flow ~~ by a combination of quantitative uptake estimation and image processing, as was discussed in an earlier issue. STATIC IMAGING The most obvious application of a computer/display system to imaging is the correction of the image for nonuniformity of response of the gamma camera. This technique
306
H . I . GLASS
has not been widely applied for two reasons: (1) the inferior quality of the device on which the processed display is normally presented and (2) the inconvenience of doing this correction routinely on all images. The most effective way of dealing with this problem is to cure it at its source by examining the uniformity at frequent intervals and correcting for any errors by electronic adjustment. Not every group is prepared to carry out this occasionally complex adjustment, while the number of gamma cameras now in operation renders it inevitable that a significant number of cameras are being operated under conditions that are not optimal. The diagnostic errors due to nonuniformity are frequently underestimated, and a routine computer correction of the image is desirable. Image processing by the use of filtering and enhancement procedures has mainly been carried out off-line or on large batch-processing machines, 1~'12 which are often not equipped with suitable display facilities or methods of producing a suitable permanent record. Despite the advent of fast Fourier-transform algorithms, ~3 it seems unlikely that routine image processing will be carried out in any but the larger computer/ display installations. It is important that a scanner be equipped with a digital dataacquisition system (either paper tape or magnetic tape) so that the data may be processed off-line if necessary, with the results then replayed through a suitable display attached to a smaller local machine. One major disadvantage with conventional scanners is the requirement of presetting the operating conditions. In this respect, a commercial system now available seems to offer a significant advance in this field. ~4 This system uses a semiconductor memory to acquire the data, which permits the data to be displayed in an optimum way and to be replayed rapidly onto photographic film. In the case of tomographic scanning, simple data acquisition is inadequate if immediate processed display is required. The earlier mathematical treatment used by Kuhl for processing tomographic images has now been superseded by more complex mathematical analyses ~s'16 requiring fairly sophisticated on-line or off-line computer/display systems. The advent of a recent commercial tomographic scanning system (J and P Engineering, Ltd., Reading, England) has increased the possibility of fully utilizing the commercial image acquisition/processing/display systems that are now available. There seems little or no advantage in using elaborate digital systems for simple qualitative static or dynamic studies on cameras or scanners where cheaper analog magnetic-tape systems, operating in either video or pulse recording modes, are adequate for reproducing the images. A further application of computers to scanning is associated with the increasing use of dual-detector scanners for cluantitative uptake studies for assessing the distribution of new radiopharmaceuticals by whole-body scanning and also in clinical applications, one example of which is the routine estimation of splenic blood volume. 17 These techniques are tedious to perform and analyze by hand, and computer processing of the data is essential. WHOLE-BODY COUNTING
After a period in which the use of whole-body counters in clinical diagnosis had stabilized, it is apparent that recently there has been renewed interest la (see also the article by Andrews et al. in this issue). In some investigations, such as multiple-tracer studies, the availability of a computer is an advantage, and it becomes especially useful if
COMPUTERS IN NUCLEAR MEDICINE
307
spectrum stripping routines must be applied. This type of analysis is easily dealt with as an off-line batch-processing problem provided suitable data-acquisition equipment is available. More recently, the technique of in-vivo neutron activation analysis has been described 19 for the analysis of trace elements in the body. The data analysis and calibration procedures in this technique certainly necessitate the use of spectral analysis by digital computer programs, owing to the complex spectra obtained.
RADIONUCLIDE DOSlMETRY AND LABORATORY RECORDS As reported by Brown, 2~ many laboratories use computers for record keeping and record searching of the procedures carried out. In some laboratories these are also used to prepare reports for inclusion in patients' case notes. The use of computers for more detailed radionuclide dosimetry calculations reflects the continuing concern with the reduction of radiation dose in nuclear medicine procedures and the improved knowledge of the biological turnover and metabolic pathways of the new and increasingly complex radiopharmaceuticals. These dosimetry estimates often require the use of lengthy Monte Carlo calculations, which can be most effectively performed on a large batch-processing computer. The recent revision of the finger dose obtained due to repeated handling of 99mTc strikes a cautionary note. 21 IS THERE AN OPTIMUM SYSTEM? There seems little doubt that the conclusion reached by Cox 22 in 1968 concerning the use of specialized computer systems is still valid today. Briefly summarized, he suggested that significant economics can be achieved by the use of built-in specially programmed small computers to do a specific job. The cost of core memory has been falling steadily for many years and continues to fall. This is reflected in the increasing sophistication of many devices with little or no increase in cost. For example, several commercial multichannel analyzers now incorporate small (up to gK) digital computers as an integral part of the system. These computers are available for use in their own right for other purposes if required. All the usual multichannel analyzer facilities are achieved by software. The disadvantage of the smaller, specially tailored systems is that the economics are only real if the devices are fully used, and that means very frequently used. If this is not likely to be the case, then the slightly less specialized system, exemplified by the computer/multichannel analyzer, is likely to be more useful, since it is more versatile, and this in itself results in increased use. Increased utilization associated with increased versatility brings its own disadvantages, since the equipment is often not available when needed for its original purpose. The way out of this dilemma appears to be to obtain a minimum number of specialpurpose instruments and to link as many pieces of equipment as possible to a minimumsize central computer system equipped specifically with peripherals required for nuclear medicine. This computer system should operate under some real-time executive software system. This would allow simultaneous input from any of the linked apparatus, unless a priority system specifically excludes apparatus in certain special circumstances, e.g., to allow temporarily high data rates to be accepted from a gamma-camera study. Real-time executive programs normally allow running foreground programs or acquisition routines concurrently with background programs that are in effect carrying out batch processing. In this way the system can cope with the wide range of data input rates derived from nuclear medicine equipment, from the slowest rates in radio-
308
H.I. GLASS
nuclide sample counting machines to the higher rates obtained from scanners and the even higher rates encountered in blood flow and function studies using open detectors or the very fast rates obtained on gamma cameras. The other specific requirement for a nuclear medicine department computer system is a digitally controlled display system. It is now possible to buy such systems independently of the computer, although as mentioned above, several such displays are now available from computer manufacturers. These may take the form of television displays driven directly from core or through an intermediate buffering device such as a drum or small disc. SPECIFICATION OF A SUITABLE SYSTEM
To run a computer under a real-time executive system with the number of inputs likely to be encountered in an average-size nuclear medicine department, a minimum system requirement would be a core memory size of 16K plus a magnetic-disc memory. A dual-disc system (one of fixed-head design with a capacity of about 500K words and one exchangeable disc design with a capacity of about 2.5M words) is probably necessary. As the number of input devices increases, an extra 8K words of memory would become highly desirable. To allow for the possibility of off-line processing on larger machines and also as a convenient mass-storage medium, a magnetic-tape handler is useful, although a paper-tape punch may be adequate in some circumstances as a computer-compatible output. Currently, nine-track magnetic-tape handlers are increasingly being used because of their increased storage capacity over seven-track systems; however, many large batch-processing systems will still accept only the latter. The usual magnetic-tape specification offers 45-ips tape speed and 800-cpi tape density; however, 1600-cpi densities are now obtainable on some magnetic-tape handlers. A paper-tape reader and keyboard terminal would also be required. The display should permit the presentation of a picture matrix such that each element corresponds to about one-eighth of the detector resolution (FWHM) of a radionuclide imaging system. In most cases this will result in an image in which the individual picture elements are virtually indistinguishable on a 12-cm oscilloscope. The monochrome display should allow the presentation of a minimum of 16 brightness levels. In the case of those devices where the data input rate is very low, it is possible that a multiplexer which works into a single ADC may result in some financial saving. Similarly; it may be more efficient to batch-process the automatic sample counter output and arrange for the data to be obtained in cheap computer-compatible form such as paper tape. A suggested nuclear medicine system is shown in Fig. 1.
I DISPLAY I 128x 128 . IMEDICINE NUCLEAR I EQUIPMENT
t
16K;16b=t:~ I ~sec
L
ITELETYPE 1
TAPEUNIT
~IIDUAL m m~AG95r~[woris ~ PAPER TAPE READER 400cps I Fig.1.
COMPUTERS IN NUCLEAR MEDICINE
309
DISCUSSION
The computer system which has been proposed might at first appear elaborate and expensive, but it should be considered in the light of the cost of the associated equipment which is normally present in nuclear medicine departments. The presence of at least three imaging devices in a nuclear medicine department is not unusual, and the total cost of the computer and its peripherals shown in Fig. 1 is approximately onethird of that of the nuclear medicine equipment. Furthermore, the recurring cost of radiopharmaceuticals is certainly not insignificant, whereas the running cost of the computer system is essentially its maintenance cost, which may be estimated at approximately 5% of its capital value. With a computer system of this level of complexity, it should be recognized that a scientist or engineer will almost certainly have to be employed if the system is to be effectively utilized, since it is virtually certain that software will not be available from manufacturers to satisfy all the applications likely to be devised in nuclear medicine departments. A complete clinically oriented system having a minimum of knobs and having been designed for maximum ease of use by the clinician would obviously be desirable. However, it seems fair to point out that the term nuclear medicine means all things to all men at the present time, and the activities of individual nuclear medicine departments vary considerably. The possibility of producing a single system with a single set of software which will be acceptable or even useful in all nuclear medicine departments is remote. In the gamma-camera field, at least five commercial computer/display systems are available, all having advantages and disadvantages. It would seem unlikely that a general clinical computer system with packaged software will be available for some time. If it is economically possible, as much data collection and data analysis as possible should be done on-line, as this is without question the most convenient and efficient mode of operation for clinical departments. Whether this is done by use of small local data buffers attached to each device or by a direct on-line link to a larger system is a local decision of logistics, economics, and policy. In general, where batch processing is convenient and acceptable, it should be performed. However, it is extremely important when installing a computer to serve a nuclear medicine department that (1) it should be expandable to at least 32K, (2) it should ultimately be capable of operation under a real-time executive system, (3) it should be equipped with a versatile display, and (4) it should be capable of producing a computer-compatible output (preferably magnetic tape) to allow off-line batch processing of more difficult problems on a larger, faster machine. When building up a system, it is advisable to allow for this expansion possibility. REFERENCES
1. Skrabal F, Arnot RN, Helus F, et al: A method for simultaneous electrolyte investigations in man using bromine-77, potassium-43, and sodium-24. Int J Appl Radiat Isot 21:183, 1970 2. Rodbard D, Lewald JE: Computer analysis of radioligand assay and radioimmunoassay data, in: Proceedings of 2nd Symposium on Research Methods in Reproductive Endocrinology. Stockholm, Karolinska Institute, 1970, p 79
3. Sveindottir E, Lassen NA, Risberg J, Ingvar DH: Regional cerebral blood flow measured by multiple probes: An oscilloscope and digital computer system for rapid data analysis, in Brock M, Feschi C, Ingvar D, et al, (eds): Cerebral Blood Flow. Berlin, Springer-Verlag, 1969, p 27 4. Veall N, Crawley JCW: The 133Xe inhalation technique. Scand J Clin Lab Invest Suppl 102, XI:F, 1968 5. Berman M: The application of multicom-
310
partmental analysis to problems of clinical medicine. Ann Intern Med 68:423, 1968 6. Budinger TF: Clinical and Research Quantitative Nuclear Medicine System, in: Medical Radioisotope Scintigraphy. Vienna, IAEA, 1973 7. Natarajan TK, Wagner HN Jr: A new image display and analysis system (IDA) for radionuclide imaging. Radiology 93:823, 1969 8. Brownell GL, Burnham CA: Numedics, a computer system for processing radioisotope data from multiple sources, USAEC Conf.710425. Oak Ridge, Tenn., USAEC, 1971, p 215 9. Brill AB, Erikson JJ, Lindahl CE: Digital systems for acquisition and storage, in Kenny PJ, Smith EM (eds): Quantitative Organ Visualization in Nuclear Medicine. Coral Gables, Fla., University of Miami Press, 1971, p 339 10. Mclntyre WJ, Bott RE, lshii Y, Pritchard WH: Atraumatic measurement of the distribution of myocardial blood flow using 43K and a scintillation camera, in: Medical Radioisotope Scintigraphy. Vienna, IAEA, 1973 11. Nagai T, Fukuda N, linuma TA: Computer focussing using an appropriate gaussian function. J Nucl Med 10:209, 1969 12. Brown DW: Computer processing of radioisotope scans using Fourier and other transformations. J Nucl Med 11:304, 1970 13. Cooley JW, Tukey JW: An algorithm for the machine calculation of complex Fourier series. Mathematics of Computations 19:297, 1965. 14. Becker J, Graul EH: Computer evaluation and colour display of conventional scinti-
H.I.
GLASS
grams. Compact News in Nuclear Medicine 3: 55, 1972 15. Kuhl DE, Edwards RQ, RicciAR, Reivich M: Quantitative section scanning using orthogonal tangent correction. J Nucl Med 13:447, 1972 16. Chesler DA: Three dimensional reconstruction technique. (to be published) 17. Hegde VM, Williams ED, Lewis SM, et al: Measurement of splenic red cell volume and visualization of the spleen with Technetium99m. ~to be published) 18. Reizenstein P: Clinical Whole Body Counting. Bristol, U.K., John Wright and Son, 1973 19. Cohn SH, Shukla KK, Dombrowskie CS, Fairchild RG: A total body neutron activation facility employing portable a-n sources for medical research, in: Proceedings of IAEA Expert Panel on ln-Vivo Activation Analysis. Vienra, IAEA (in press) 20. Brown DW, Kirch D, Groome D, et al: General purpose computers for processing of image data, in Kenny PJ, Smith EM (eds): Quantitative Organ Visualization in Nuclear Medicine. Coral Gables, Fla., University of Miami Press, 1971, p 431 21. Henson PW: A note of some aspects of skin contamination by certain radionuclides in common use. Br J Radiol 45:938, 1972 22. Cox JR: Economy of Scale and Specialization in Large Computing Systems. Monograph 93, Biomedical Computing Laboratory, Washington University, St. Louis, Mo., 1968