Int J Raduuron Oncu/ogy Bml. Phys.. Vol. 18. pp 22 I-23 I Pnnted in the U.S.A. All rights reserved.
??Technical Innovations
DIGITAL
0360.3016/90 $3.00 + .OO Copyright G 1990 Pergamon Press plc
and Notes
TUMOR FLUOROSCOPY (DTF)-A IN THE THERAPY PLANNING M.
‘Department
of Radiation
Oncology,
University
HERBST’
AND
of Erlangen-Ntirnberg;
NEW DIRECT IMAGING FOR BRAIN TUMORS
M.
SYSTEM
FR~DER~ and ‘Fraunhofer
Arbeitsgruppe
fir lntegrierte
Schaltungen
Digital Tumor Fluoroscopy is an expanded x-ray video chain optimized to iodine contrast with an extended Gy scale up to 64000 Gy values. Series of pictures are taken before and after injection of contrast medium. With the most recent unit, up to ten images can be taken and stored. The microprogrammable processor allows the subtraction of images recorded at any moment of the examination. Dynamic views of the distribution of contrast medium in the intravasal and extravasal spaces of brain and tumor tissue are gained by the subtraction of stored images. Tumors can be differentiated by studying the storage and drainage behavior of the contrast medium during the period of examination. Meningiomas store contrast medium very intensively during the whole time of investigation, whereas astrocytomas grade 2-3 pick it up less strongly at the beginning and release it within 2 min. Glioblastomas show a massive but delayed accumulation of contrast medium and a decreased flow-off-rate. In comparison with radiography and MR-imaging the most important advantage of Digital Tumor Fluoroscopy is that direct information on tumor localization is gained in relation to the skull-cap. This enables the radiotherapist to mark the treatment field directly on the skull. Therefore it is no longer necessary to calculate the tumor volume from several CT scans for localization. In Radiotherapy Digital Tumor Fluoroscopy a unit combined with a simulator can replace CT planning. This would help overcome the disadvantages arising from the lack of a collimating system, and the inaccuracies which result from completely different geometric relationships between a CT unit and a therapy machine. Brain tumor, Radiotherapy, Therapy planning.
the scans to a lateral skull view for exact localization by the simulator. This method suffers from many inaccuracies and does not permit a sufficiently precise tumor localization. 4. In the CT scout view the central beam is focused on a certain point of the skull, such as the sella turcica. For therapy planning, the central beam ofthe simulator is placed in the center of the tumor. Using bone reference points to transfer the tumor coordinates from the scout view to the simulator image might result in faulty localization with a deviation of up to 3 cm due to the totally different relationships in the case of noncentral tumor localization, (i.e. in the anterior fossa).
INTRODUCTION Radiography diagnosis
and MR-tomography
and their image
quality
are standard is excellent.
in tumor Therefore,
they are routinely used for diagnosis, therapy planning, and follow-up of brain tumors after surgery and/or ra-
diotherapy. In the case of therapy planning, these techniques have several limitations as far as the requirements of radio-oncologists are concerned: With a CT or MR scanner it is impossible to simulate the movements and the field collimation of a linear accelerator or a 6oCo machine. CT and MR systems cannot be used directly during surgery or radiotherapy planning because they are located outside the operating theater or the simulator room. For direct localization they should be available at all times, but this would be prohibitive for reasons of cost and space. It is difficult to estimate the treatment volume on the basis of several CT scans by imaginary 3-D synthesis. The margins of a tumor have to be transferred from
Reprint requests to: Prof. Dr. M. Herbst, Oncology, University of Erlangen-Nuernberg, 852 Erlangeno, West Germany.
In view of this fact, neurosurgeons and radiotherapists are interested in imaging systems that can be used during surgery or simulation of radiotherapy. Such systems would permit direct and precise localization of brain tumors with high spatial resolution in relation to the skull (3). To solve this problem, a new imaging method has been developed at the University of Erlangen-Niirnberg (1, 2, 5). It is based on pursuing the dynamics of tumor and
Dpt. of Radiation Universitgtsstr. 2
Accepted
221
for publication
15 June 1989.
I. J. Radiation
Oncology
0 Biology 0 Physics
January
1990, Volume
18. Number
I
Imlge
I I
Memory
Reoltlme PlTG2SSW L
I
EMbyte
1
I I
L______________--_J
Fig. I. (a) Overview of the Digital Tumor Fluoroscopy unit showing the patient with his head in the fluoroscopic device and the system behind. (b) Configuration of the above shown Digital Tumor Fluoroscopy.
normal tissue opacification during the infusion of radiopaque fluid ( 100 ml/5 min). This modality is called Digital Tumor Fluoroscopy (DTF). It allows the distribution of contrast material to be studied inside intravasal and extravasal spaces of tumors and healthy tissue. With DTF, it is possible to demarcate a tumor from normal brain tissue mainly by three criteria: (1) Hypervascularization; (2) Reduced vascularization (necrosis); (3) Impaired blood-brain barrier. This procedure provides information on the extension of the tumor itself and its infiltration into the surrounding structures. The principle is also applicable to extracranial tumor locations. We started to investigate brain tumors, as the head can be easily immobilized in a treatment mask (6). METHODS
AND
MATERIALS
The system is made up of an x-ray generator, an image intensifier, a non-invasive head fixation device, and an image processing system. The principle of imaging is similar to that in DSA (Digital Subtraction Angiography), but there are some differences: (a) The Gy scale is extended from 1000-2000 (DSA) to 64000 values and the noise is reduced by integration of 1000-1200 images. (b) The xray video system is optimized to iodine contrasts. It is capable of resolving contrasts of 1.3 per 1000 if the requirement of spatial invariance of the subject is satisfied. High sensitivity for iodine contrast results from an increasing photon yield of the videosection of the chain and from the use of x-ray energies optimized to iodine (4065 kV, 5 mA). (c) The infusion techniques are similar to those used in computed-tomography (without catheter, * Image processing
system, KONTRON
IPS.
intravenously, 100 ml in about 30 set-5 min). (d) Algorithms for the manual and automatic shifting and rotating of the subtraction mask are necessary to eliminate motion artifacts of the patient in the case of protracted observation times. A first series of 1OOO- 1200 images being integrated to one picture is taken before applying radiopaque fluid for the purpose of obtaining a mask for subtraction. The second series is recorded after injection of iodine contrast medium (Omnipaque 300 mg/ml, 100 ml). The next series are recorded after 9 and 20 min. Two years ago, we started to investigate brain tumors by using DTF. Retrospectively. we can distinguish three stages of development: 1. In 1985 we started with a small computer with a storage capacity of 5 12 X 5 12 pixel of 8 bit. This technical prototype was able to integrate six images on the camera tube by beam blanking to an R-bit picture. Then 200 of these digital images were added by the microcomputer to a 16-bit image. To keep the computing time within acceptable limits, an image format 128 X 128 points was selected. A disadvantage of this system was that no images could be added digitally. 2. One year ago, we started to develop a clinical prototype with a new image-processing system, (Fig. la and 1b) “IPs.“* capable of storing 60 images of 5 12 X 5 12 pixels and of adding 8-bit images digitally in real time (25 per set) to a 16-bit image. This is achieved by using a special-real time processor. Subtractions of the mask from the following images are done by another microprogrammable microprocessor controlling the image storage. This device enables shifting and rotating of the mask images (including interpolations) very quickly
Digital tumor fluoroscopy 0 M.
to eliminate artifacts resulting from patient movements. It is now possible to integrate digitally 1024 images (format 5 12 X 5 12) into one, resulting in better image quality than that obtained by camera beam blanking. To improve the methods, radiopaque fluid is injected intravenously as a bolus over 3 min, but not over 5 min as we did before. Fluoroscopy is started during the last 15 set of the bolus injection, when the iodine concentration in the vessels is highest. Thus it is possible to observe an early phase of flow into the tumors with a high degree of vascularization. As before, four series of images (1024 integrated) are obtained. The first series is taken before injection of contrast material, the second starts 15 set before the end of bolus injection, and the third and fourth are taken after 9 and 15 min, respectively. Radiation exposure (skin dose) on the tube side of the patient’s head during a series was 1.36 R with the technical prototype. A dose reduction was achieved by using the clinical prototype (Table 1). The reduction of integrations cut exposure to a total dose of 0.64 R without any loss of information. Exposure is lower by a factor of 30 than with a roughly comparable spatial resolution in a reconstruct of a lateral or frontal view. 3. The method has been improved during the past 3 months by changing the procedure of contrast-medium injection. The fluid is now injected as a bolus within 30-45 sec. So the number of integrations can be re-
223
HERBST AND M. FRODER
duced from 1024 to 256 resulting in a 16-bit image without loss of detail. This curtails the dose further by a factor of 4 and yields a better temporal resolution. Therefore, the number of total images per examination was increased from 4 to 10. This facilitates better viewing of the opacification and, reduces the sensitivity to motion artifacts because an increased number of image combinations is possible for subtraction. In addition to the perfusion reaction being studied, the “depletion” characteristics of tissue are observed by subtracting images from the image recorded at the time of maximum iodine concentration in the blood. RESULTS
The results obtained from 48 patients investigated with brain tumors were analyzed. Two groups of patients are to be distinguished. Group I The first group of 19 patients was examined with the technical prototype for DTF. Seven of 19 patients were studied twice, once before radiotherapy was started and the second time for follow-up. All 19 patients had CT scanning before DTF. Most tumors were confirmed histologically. Nine patients suffered from astrozytoma grade 2-3, seven had glioblastomas, two meningiomas, and one patient was affected by metastases of squamous-cell carcinoma. With the exception of one (motion artifacts), all
Table I. Course of digital fluoroscopy concerning timing, contrast-medium number of images, duration, and radiation absorbed dose
Method Technical prototype
(method I) Clinical prototype
(method II)
Time of imaging
Duration of CMapplication
Exposure time (one 16-bit image)
Number of integrations to one image
application,
Skin dose (energy: 44-65 KV/SmA)
Omnipaque 5 min 10 min 20 min
300 ml
5 min
56 56 56 56
set set set set
6 6 6 6
X x X x
200 200 200 200
0.34 0.34 0.34 0.34 Total = 1.36
Omnipaque 3 min 9 min 15 min
300 ml
3 min
45 45 45 45
set set set set
4 4 4 4
X X X X
256 256 256 256
0.268 0.268 0.268 0.268 Total = 1.072
Omnipaque
300 ml 1X 1X 1X 1X 1X 1X 1X 1X 1X
256 256 256 256 256 256 256 256 256
0.064 0.064 0.064 0.064 0.064 0.064 0.064 0.064 0.064 Total = 0.064
(method III)
30-45 set 30 set 1 min 2 min 3 min 5 min 7 min 10 min 13 min
11 set 1 I set 11 set 11 set
11 set 11 11 11 11
set set set set
R (field 6 X 6 cm) R R R R R (field 15 X 20 cm) R R R R (field 15 X 20 cm)
224
I. J. Radiation Oncology 0 Biology 0 Physics
gliomas were detected and displayed. In one case. it was difficult to detect the tumor by CT as it was isodense with the surrounding edema but there was no problem in viewing the tumor fluoroscopically. CUSP I (technical prototype, method I). Figure 2 shows a survey radiograph of the skull with a grid superimposed for localization. The image gained by DTF is superimposed on the scout view. The small section of 6 X 6 cm represents a meningioma studied by tumor fluoroscopy. The small size of the tumorfluoroscopic image is due to the fact that the prototype could not calculate larger sections. The associated computer tomogram showing the meningioma in the left cerebral hemisphere is reproduced in Figure 3. The results of the dynamic studies are shown in Figure 4. The individual images are characterized by the adjacent diagram on the right side. The first horizontal fields contain the images 1. l- 1.4. Image 1.1 was obtained without radiopaque fluid; the views 1.2- 1.4 were taken with contrast medium 5, 10, and 20 min after injection of the radiopaque agent. The views in the second, third and fourth row were taken after subtraction. The images 4.3, 3.3, and 2.3 are the result ofsubtraction ofthe images taken after 5, 10, and 20 min from the mask (1.1). The tumor shows strong opacification directly after infusion which remains nearly unchanged during the time of examination. This is a typical flow characteristic of meningiomas during fluoroscopy.
Group 2 The second group was investigated after improving the system by extending the image format from 128 X 128 to 5 12 X 5 12 points. The microprogrammable image processing system enables digital real time integration (method 2 and 3). This reduces the time of exposure and
January 1990, Volume 18, Number I
allows ten images to be recorded instead of four during the study. It also yields more information on the turnover of contrast medium in the tumor and its surrounding structures. Twenty-nine patients with different brain tumors were examined. Eleven of them had glioblastoma, seven astrocytoma grade 2-3, five metastases (from an ovarian, one gastric, and two from testicular, one mamma carcinoma) another four had a meningioma. One patient suffered from a carcinoma infiltrating the base of the skull and the brain. Another patient had a ganglioghoma. The diagnosis was histologically confirmed for 20 patients. The images of the next case were recorded with the new system. Cuss 2 (clinical prototype, method 2-Imqing M’as
started ,ftiNoMing injection now being reduced to 3 min). Figure 5a shows the CT scan of a female patient with a metastasis of an ovarian cancer in the right hemisphere and Figure 5b-d shows the views of the dynamic study with DTF after 5,9. and 15 min. The dynamic study (5b5d) reveals pronounced opacification during the investigation. The tumor shows heavy blushing already in the first image (5b) which was recorded I5 set before the end of injection. The following images 5c (after 9 min) and 5d (after 15 min) reveal increased opacification and extension of the dark area, although the intravasal concentration of contrast medium decreased. This does not mean that contrast material was actively stored. In fact, only the ratio between the amount of iodine slowly leaving the tumorous tissue and the rapidly falling concentration in healthy tissue is increased. The long duration of storage indicates that the blood-brain barrier is impaired. Cusr 3 (clinicul protot?‘pe. method S-Injection in about 30 .YK and earl~~imuggingduring injection). Figure 6a shows
Fig. 2. (Case I) Plain radiograph ofthe skull with the superimposed raster for radiotherapy planning. In the middle, an image gained by tumor fluoroscopy is faded in. The small section is 6 X 6 cm (128 X 128 pixels). The technical prototype of DTF was only able to cope with sections of this size.
Digital tumor fluoroscopy
Fig. 3. (Case 2) CT-scan
of the meningioma
the computertomogram of a patient with a astrocytoma grade 3-4 verified at biopsy. Mainly two areas of tumorous tissue (medial and lateral) can be identified. The location of both structures is transferred from the CT scans to an overview of the skull (Fig. 6b). The extension of the tumor from transversal CT scans can be compared with that of DTF in lateral projection. The images 6c-e show increasing opacification over time. The regions with an impaired blood-brain barrier are enhanced (frontal ring structures, arrows), although the iodine concentration in the blood vessels decreases. In this case, more information ofthe real
Fig. 4. (Case without and the result of after 20 min
0 M. HERBST AND M. FRODER
225
shown in Figure 5
extension of the pathological process were gained from observing the washout characteristics (6f-h). In the early image, 6f (washout of iodine between 30 and 60 set after infusion), only a small area of the tumor appears in the center of the skull, indicating a decrease in opacification during this special time period (30-60 set, dark). The tumor area appears extended in the following image 6g (washout from 30 set to 2 min), whereas in the frontal ring structure the iodine concentration seems to rise further compared with normal tissue (white arrows). The highly delayed flow-off from this large area indicates a
1) Dynamic study of a meningioma with DTF. (See numbers on the adjacent diagram). Section 1.1 section 1.2-1.4 with contrast agent 8, 15 and 20 min after radiopaque infusion. Section 4.3 shows subtraction of 1.2 from 1.1. Section 3.3 is taken after 8 min (section 1.3 from 1.1) and section 2.3 (section 1.4 from 1.1).
Fig. 5. (Case 2) (a) C T scan of a patient with a cerebral metastasis of an ovarian cancer in the right hemisphere; study of a patient with a view of a metastasis from an ovarian cancer at 15 sec. (5b), 9 min (b-c j) dynamid kF (5c) and 15 min (5d). Opacification seems to become more intense with time.
Digital tumor
lluoroscopy
0 M. HERBST AND M. FRODER
Fig. 5.
much larger pathological area than is marked in the “CT survey view” (6b) by coordinates out of the scans. DISCUSSION Radiology and MR-imaging are excellent tools for displaying brain tumors. However, they furnish only indirect information on the tumor localization for the therapy to be instituted by neurosurgeons or radio-oncologists. This is because the tumor location must be reconstructed from several scans and then transferred to a lateral view of the skull. The accuracy of this procedure is impaired for therapy planning by the different geometries of the centrically focussed CT scanner and the eccentrically focussed simulator. Faulty tumor location occurs whenever the lesion is located eccentrically in the head which is quite often the case as most of the tumors are located eccentrically and they are the only ones to be treated by sparing normal brain tissue. So these different geometries cause faulty tumor location for therapy. Even if these machines were located in the operating theatre or in the radiotherapy department it would not help much, because a scan can only furnish 2-dimensional information and the tumor is not visible by fluoroscopy during the action of operation or therapy simulation. So one has to take several scans to cover the complete outline of the tumor and to get enough bone reference points for the indirect reconstruction of the tumor localization on a lateral survey view. All this tends to produce errors, is expensive and time consuming. On the other hand CT units and MR scanners have no collimation system to confine the volume for treatment. In view of these shortcomings, both neurosurgeons and radiotherapists are interested in imaging systems capable of fulfilling their demands. Digital Tumor Fluoroscopy is
221
(Con&)
a new imaging method which comes close to a realization of these demands. It is capable of outlining brain tumors directly in relation to the skull in almost every projection. So the neurosurgeon can estimate during the intervention an optimal approach for stereotactical biopsies and implantation of radionuclids. For percutaneous radiotherapy the system allows the treatment field to be directly localized and marked on the skin or on the treatment mask. DTF differs from CT and MR imaging because it furnishes information not only on the location but also on the dynamics of the blood perfusion of tumors. The sensitivity of DTF to iodine is higher than that of computed tomography because the x-ray energy is optimized to iodine. An increased photon yield of the video system is possible, because the x-ray tube is operated at 40-65 KV/ 5mA. This gives a resolution of iodine from 1.3 per thousand and demonstrates different opacifications, especially in the areas surrounding the tumor. Low iodine concentrations are hardly detectable with computed tomography because the used energy of about 12.5 KV is not favorable to the absorption characteristic of iodine. The contrast medium is administered in less than 1 min. Formerly it was given as an infusion over 5 min and the first images were taken after this time. So a very low concentration of iodine remained in the tissue, because a large amount was excreted by the kidneys during this time. There was no information about the distribution of contrast medium in the tumor area immediately after the start of bolus application. Now imaging starts after 30 seconds. The dynamics of opacification in the tumor region can be studied in relation to the normal vessels. In the case of a highly vascularized tumor, intense opacification appears to decrease in relation to the vessels. If the blood-brain barrier is impaired, the flow-off characteristics are different.
228
1. J. Radiation Oncology 0 Biology 0 Physics
Brain tumors show special characteristics of opacification ( 12). This is confirmed by our studies (64 patients) using DTF. Low grade astrocytomas are marked by opacification at the beginning of the study and by a decrease in relation to that of the blood vessels. Glioblastomas are opacified in the second phase of examination without depletion at the end. Meningiomas show intense blushing at the beginning without change of opacification until the end of the examination. This provides infor-
January
1990, Volume
18, Number
I
mation on the possible histological typing of a tumor, but cannot replace histology. In case of uncertain histology the neurosurgeon can decide interoperatively to take biopsies from areas with different opacification characteristics revealed by DTF. This is not possible with CT because the scanner is located outside the operating theater, and does not give information about the dynamic course of opacification of the whole tumor at the same time.
Fig. 6. (Case 3) (a) CT scan of a patient with a astrocytoma grade 3-4. Two areas of tumor tissue are opacified in the frontal midline and left parietal region; (b) Translation of the astrocytoma from transversal CT scans (one shown in Fig. 6a) to a lateral view of a plain skull radiograph. Scanning from the base of the skull to the top the frontal lesion (see Fig. 6a) is extending to the large, dash-lined area: (c-e) DTF study of an astrocytoma, grade 34. Images taken at 30 set (7c), 2 min (7d) and 13 min (7e). The contrast medium is stored (with more arrows at the beginning of the study and released partially over time; (f-h) DTF study of tissue “destoring” contrast medium (dark structures) between 30 set and 1 min. (6f), between 30 set and 2 min (6g), and between 30 set and 10 min (6h).
Digital tumor
fluoroscopy
0 M. HERBST AND M. FRODER
Fig. 6. (Contd)
229
230
I. J. Radiation
Oncology
0 Biology 0 Physics
January
1990, Volume
18. Number
1
Fig. 6. (Con@
The importance of blood perfusion of tumors gave rise to dynamic CT studies in the early years of computed tomography (7). They furnished a lot of fundamental diagnostic findings. Unfortunately, the time-consuming procedure is not practicable in clinical routine, because only one slice can be studied at a time. Studying the whole tumor region slice by slice would push the expenditure of time and dose to unacceptable levels. The highest dose in DTF (skin dose on the entry side of the skull) was 0.4 R per series with the technical prototype (3,4) and is now 0.064 R with the improved clinical system. Four series of images taken with the technical prototype resulted in a dose of 1.6 R and of 0,64 R with the clinical type after 10 series. Any attempt to reconstruct
sagittal slices with high resolution by CT would inevitably lead to a dose escalation. Overlapping scans with high spatial resolution would give a high integral dose as a result of added scattered radiation fields. The cost of the equipment is determined by the following two system components: (1) The x-ray video chain, which is standard equipment of any operating theatre or simulator room: and (2) The image-processing system is commercially available throughout the world and of standard design. Consequently the acquisition cost of a DTF setup is in the range of normal clinical equipment, and is expected to be negligible in relation to that for CT, because of falling prices for electronic circuits and storage systems in the future.
Digital tumor fluoroscopy 0
M. HERBST AND M. FRODER
231
REFERENCES Froder, M.; Herbst, M. Dynamic studies of brain tumors by the use of digital fluoroscopy. IEEE Trans. MI M l-4: 104-l 13; 1985. Froder, M.; Herbst, M. Darstellung von hirntumoren mit computeruntersttitzter rontgenbildverstarkertechnik. Fortschr. Rontgenstr. 146:66-72; 1987. Friider, M.; Seitzer, D.; Buren, G.; Dieckmann, G. Digitale rontgenvideobildverarbeitung filr die zielpunktberechnung in der stereotaktischen neurochirurgie. Fortschr. Rbntgenstr. 140:84-86; 1984. Greitz, T. Diagnostic and therapeutic strategies in neuroradiology. But. f. Radio]. 58:1 148-l 163; 1985.
5. Herbst, M.; Froder, M.; During, diagnostik und therapieplanung lentherapie (In press).
A. Tumorfluoroskopievon hirntumoren. Strah-
Herbst, M.; Wuermeling, M.; Sauer, R. Fixierung von patienten mit tumoren im Kopf-Hals-Bereich durch individuelle bestrahlungsmasken. In: Wannenmacher, M., ed. Naso-pharynx-tumore. Mtinchen-Wien-Baltimore: Urban & Schwarzenberg; 1983:243-245. Norman, D.; Stevens, E. A.; Wing, S. D.; Lewin, V.; Newton, T. H. Quantitative aspects of contrast enhancement in cranial computed tomography. Radiology 129:683-688; 1978.
??Technical Innovations
DIGITAL
0360.3016/90 $3.00 + .OO Copyright G 1990 Pergamon Press plc
and Notes
TUMOR FLUOROSCOPY (DTF)-A IN THE THERAPY PLANNING M.
‘Department
of Radiation
Oncology,
University
HERBST’
AND
of Erlangen-Ntirnberg;
NEW DIRECT IMAGING FOR BRAIN TUMORS
M.
SYSTEM
FR~DER~ and ‘Fraunhofer
Arbeitsgruppe
fir lntegrierte
Schaltungen
Digital Tumor Fluoroscopy is an expanded x-ray video chain optimized to iodine contrast with an extended Gy scale up to 64000 Gy values. Series of pictures are taken before and after injection of contrast medium. With the most recent unit, up to ten images can be taken and stored. The microprogrammable processor allows the subtraction of images recorded at any moment of the examination. Dynamic views of the distribution of contrast medium in the intravasal and extravasal spaces of brain and tumor tissue are gained by the subtraction of stored images. Tumors can be differentiated by studying the storage and drainage behavior of the contrast medium during the period of examination. Meningiomas store contrast medium very intensively during the whole time of investigation, whereas astrocytomas grade 2-3 pick it up less strongly at the beginning and release it within 2 min. Glioblastomas show a massive but delayed accumulation of contrast medium and a decreased flow-off-rate. In comparison with radiography and MR-imaging the most important advantage of Digital Tumor Fluoroscopy is that direct information on tumor localization is gained in relation to the skull-cap. This enables the radiotherapist to mark the treatment field directly on the skull. Therefore it is no longer necessary to calculate the tumor volume from several CT scans for localization. In Radiotherapy Digital Tumor Fluoroscopy a unit combined with a simulator can replace CT planning. This would help overcome the disadvantages arising from the lack of a collimating system, and the inaccuracies which result from completely different geometric relationships between a CT unit and a therapy machine. Brain tumor, Radiotherapy, Therapy planning.
the scans to a lateral skull view for exact localization by the simulator. This method suffers from many inaccuracies and does not permit a sufficiently precise tumor localization. 4. In the CT scout view the central beam is focused on a certain point of the skull, such as the sella turcica. For therapy planning, the central beam ofthe simulator is placed in the center of the tumor. Using bone reference points to transfer the tumor coordinates from the scout view to the simulator image might result in faulty localization with a deviation of up to 3 cm due to the totally different relationships in the case of noncentral tumor localization, (i.e. in the anterior fossa).
INTRODUCTION Radiography diagnosis
and MR-tomography
and their image
quality
are standard is excellent.
in tumor Therefore,
they are routinely used for diagnosis, therapy planning, and follow-up of brain tumors after surgery and/or ra-
diotherapy. In the case of therapy planning, these techniques have several limitations as far as the requirements of radio-oncologists are concerned: With a CT or MR scanner it is impossible to simulate the movements and the field collimation of a linear accelerator or a 6oCo machine. CT and MR systems cannot be used directly during surgery or radiotherapy planning because they are located outside the operating theater or the simulator room. For direct localization they should be available at all times, but this would be prohibitive for reasons of cost and space. It is difficult to estimate the treatment volume on the basis of several CT scans by imaginary 3-D synthesis. The margins of a tumor have to be transferred from
Reprint requests to: Prof. Dr. M. Herbst, Oncology, University of Erlangen-Nuernberg, 852 Erlangeno, West Germany.
In view of this fact, neurosurgeons and radiotherapists are interested in imaging systems that can be used during surgery or simulation of radiotherapy. Such systems would permit direct and precise localization of brain tumors with high spatial resolution in relation to the skull (3). To solve this problem, a new imaging method has been developed at the University of Erlangen-Niirnberg (1, 2, 5). It is based on pursuing the dynamics of tumor and
Dpt. of Radiation Universitgtsstr. 2
Accepted
221
for publication
15 June 1989.
I. J. Radiation
Oncology
0 Biology 0 Physics
January
1990, Volume
18. Number
I
Imlge
I I
Memory
Reoltlme PlTG2SSW L
I
EMbyte
1
I I
L______________--_J
Fig. I. (a) Overview of the Digital Tumor Fluoroscopy unit showing the patient with his head in the fluoroscopic device and the system behind. (b) Configuration of the above shown Digital Tumor Fluoroscopy.
normal tissue opacification during the infusion of radiopaque fluid ( 100 ml/5 min). This modality is called Digital Tumor Fluoroscopy (DTF). It allows the distribution of contrast material to be studied inside intravasal and extravasal spaces of tumors and healthy tissue. With DTF, it is possible to demarcate a tumor from normal brain tissue mainly by three criteria: (1) Hypervascularization; (2) Reduced vascularization (necrosis); (3) Impaired blood-brain barrier. This procedure provides information on the extension of the tumor itself and its infiltration into the surrounding structures. The principle is also applicable to extracranial tumor locations. We started to investigate brain tumors, as the head can be easily immobilized in a treatment mask (6). METHODS
AND
MATERIALS
The system is made up of an x-ray generator, an image intensifier, a non-invasive head fixation device, and an image processing system. The principle of imaging is similar to that in DSA (Digital Subtraction Angiography), but there are some differences: (a) The Gy scale is extended from 1000-2000 (DSA) to 64000 values and the noise is reduced by integration of 1000-1200 images. (b) The xray video system is optimized to iodine contrasts. It is capable of resolving contrasts of 1.3 per 1000 if the requirement of spatial invariance of the subject is satisfied. High sensitivity for iodine contrast results from an increasing photon yield of the videosection of the chain and from the use of x-ray energies optimized to iodine (4065 kV, 5 mA). (c) The infusion techniques are similar to those used in computed-tomography (without catheter, * Image processing
system, KONTRON
IPS.
intravenously, 100 ml in about 30 set-5 min). (d) Algorithms for the manual and automatic shifting and rotating of the subtraction mask are necessary to eliminate motion artifacts of the patient in the case of protracted observation times. A first series of 1OOO- 1200 images being integrated to one picture is taken before applying radiopaque fluid for the purpose of obtaining a mask for subtraction. The second series is recorded after injection of iodine contrast medium (Omnipaque 300 mg/ml, 100 ml). The next series are recorded after 9 and 20 min. Two years ago, we started to investigate brain tumors by using DTF. Retrospectively. we can distinguish three stages of development: 1. In 1985 we started with a small computer with a storage capacity of 5 12 X 5 12 pixel of 8 bit. This technical prototype was able to integrate six images on the camera tube by beam blanking to an R-bit picture. Then 200 of these digital images were added by the microcomputer to a 16-bit image. To keep the computing time within acceptable limits, an image format 128 X 128 points was selected. A disadvantage of this system was that no images could be added digitally. 2. One year ago, we started to develop a clinical prototype with a new image-processing system, (Fig. la and 1b) “IPs.“* capable of storing 60 images of 5 12 X 5 12 pixels and of adding 8-bit images digitally in real time (25 per set) to a 16-bit image. This is achieved by using a special-real time processor. Subtractions of the mask from the following images are done by another microprogrammable microprocessor controlling the image storage. This device enables shifting and rotating of the mask images (including interpolations) very quickly
Digital tumor fluoroscopy 0 M.
to eliminate artifacts resulting from patient movements. It is now possible to integrate digitally 1024 images (format 5 12 X 5 12) into one, resulting in better image quality than that obtained by camera beam blanking. To improve the methods, radiopaque fluid is injected intravenously as a bolus over 3 min, but not over 5 min as we did before. Fluoroscopy is started during the last 15 set of the bolus injection, when the iodine concentration in the vessels is highest. Thus it is possible to observe an early phase of flow into the tumors with a high degree of vascularization. As before, four series of images (1024 integrated) are obtained. The first series is taken before injection of contrast material, the second starts 15 set before the end of bolus injection, and the third and fourth are taken after 9 and 15 min, respectively. Radiation exposure (skin dose) on the tube side of the patient’s head during a series was 1.36 R with the technical prototype. A dose reduction was achieved by using the clinical prototype (Table 1). The reduction of integrations cut exposure to a total dose of 0.64 R without any loss of information. Exposure is lower by a factor of 30 than with a roughly comparable spatial resolution in a reconstruct of a lateral or frontal view. 3. The method has been improved during the past 3 months by changing the procedure of contrast-medium injection. The fluid is now injected as a bolus within 30-45 sec. So the number of integrations can be re-
223
HERBST AND M. FRODER
duced from 1024 to 256 resulting in a 16-bit image without loss of detail. This curtails the dose further by a factor of 4 and yields a better temporal resolution. Therefore, the number of total images per examination was increased from 4 to 10. This facilitates better viewing of the opacification and, reduces the sensitivity to motion artifacts because an increased number of image combinations is possible for subtraction. In addition to the perfusion reaction being studied, the “depletion” characteristics of tissue are observed by subtracting images from the image recorded at the time of maximum iodine concentration in the blood. RESULTS
The results obtained from 48 patients investigated with brain tumors were analyzed. Two groups of patients are to be distinguished. Group I The first group of 19 patients was examined with the technical prototype for DTF. Seven of 19 patients were studied twice, once before radiotherapy was started and the second time for follow-up. All 19 patients had CT scanning before DTF. Most tumors were confirmed histologically. Nine patients suffered from astrozytoma grade 2-3, seven had glioblastomas, two meningiomas, and one patient was affected by metastases of squamous-cell carcinoma. With the exception of one (motion artifacts), all
Table I. Course of digital fluoroscopy concerning timing, contrast-medium number of images, duration, and radiation absorbed dose
Method Technical prototype
(method I) Clinical prototype
(method II)
Time of imaging
Duration of CMapplication
Exposure time (one 16-bit image)
Number of integrations to one image
application,
Skin dose (energy: 44-65 KV/SmA)
Omnipaque 5 min 10 min 20 min
300 ml
5 min
56 56 56 56
set set set set
6 6 6 6
X x X x
200 200 200 200
0.34 0.34 0.34 0.34 Total = 1.36
Omnipaque 3 min 9 min 15 min
300 ml
3 min
45 45 45 45
set set set set
4 4 4 4
X X X X
256 256 256 256
0.268 0.268 0.268 0.268 Total = 1.072
Omnipaque
300 ml 1X 1X 1X 1X 1X 1X 1X 1X 1X
256 256 256 256 256 256 256 256 256
0.064 0.064 0.064 0.064 0.064 0.064 0.064 0.064 0.064 Total = 0.064
(method III)
30-45 set 30 set 1 min 2 min 3 min 5 min 7 min 10 min 13 min
11 set 1 I set 11 set 11 set
11 set 11 11 11 11
set set set set
R (field 6 X 6 cm) R R R R R (field 15 X 20 cm) R R R R (field 15 X 20 cm)
224
I. J. Radiation Oncology 0 Biology 0 Physics
gliomas were detected and displayed. In one case. it was difficult to detect the tumor by CT as it was isodense with the surrounding edema but there was no problem in viewing the tumor fluoroscopically. CUSP I (technical prototype, method I). Figure 2 shows a survey radiograph of the skull with a grid superimposed for localization. The image gained by DTF is superimposed on the scout view. The small section of 6 X 6 cm represents a meningioma studied by tumor fluoroscopy. The small size of the tumorfluoroscopic image is due to the fact that the prototype could not calculate larger sections. The associated computer tomogram showing the meningioma in the left cerebral hemisphere is reproduced in Figure 3. The results of the dynamic studies are shown in Figure 4. The individual images are characterized by the adjacent diagram on the right side. The first horizontal fields contain the images 1. l- 1.4. Image 1.1 was obtained without radiopaque fluid; the views 1.2- 1.4 were taken with contrast medium 5, 10, and 20 min after injection of the radiopaque agent. The views in the second, third and fourth row were taken after subtraction. The images 4.3, 3.3, and 2.3 are the result ofsubtraction ofthe images taken after 5, 10, and 20 min from the mask (1.1). The tumor shows strong opacification directly after infusion which remains nearly unchanged during the time of examination. This is a typical flow characteristic of meningiomas during fluoroscopy.
Group 2 The second group was investigated after improving the system by extending the image format from 128 X 128 to 5 12 X 5 12 points. The microprogrammable image processing system enables digital real time integration (method 2 and 3). This reduces the time of exposure and
January 1990, Volume 18, Number I
allows ten images to be recorded instead of four during the study. It also yields more information on the turnover of contrast medium in the tumor and its surrounding structures. Twenty-nine patients with different brain tumors were examined. Eleven of them had glioblastoma, seven astrocytoma grade 2-3, five metastases (from an ovarian, one gastric, and two from testicular, one mamma carcinoma) another four had a meningioma. One patient suffered from a carcinoma infiltrating the base of the skull and the brain. Another patient had a ganglioghoma. The diagnosis was histologically confirmed for 20 patients. The images of the next case were recorded with the new system. Cuss 2 (clinical prototype, method 2-Imqing M’as
started ,ftiNoMing injection now being reduced to 3 min). Figure 5a shows the CT scan of a female patient with a metastasis of an ovarian cancer in the right hemisphere and Figure 5b-d shows the views of the dynamic study with DTF after 5,9. and 15 min. The dynamic study (5b5d) reveals pronounced opacification during the investigation. The tumor shows heavy blushing already in the first image (5b) which was recorded I5 set before the end of injection. The following images 5c (after 9 min) and 5d (after 15 min) reveal increased opacification and extension of the dark area, although the intravasal concentration of contrast medium decreased. This does not mean that contrast material was actively stored. In fact, only the ratio between the amount of iodine slowly leaving the tumorous tissue and the rapidly falling concentration in healthy tissue is increased. The long duration of storage indicates that the blood-brain barrier is impaired. Cusr 3 (clinicul protot?‘pe. method S-Injection in about 30 .YK and earl~~imuggingduring injection). Figure 6a shows
Fig. 2. (Case I) Plain radiograph ofthe skull with the superimposed raster for radiotherapy planning. In the middle, an image gained by tumor fluoroscopy is faded in. The small section is 6 X 6 cm (128 X 128 pixels). The technical prototype of DTF was only able to cope with sections of this size.
Digital tumor fluoroscopy
Fig. 3. (Case 2) CT-scan
of the meningioma
the computertomogram of a patient with a astrocytoma grade 3-4 verified at biopsy. Mainly two areas of tumorous tissue (medial and lateral) can be identified. The location of both structures is transferred from the CT scans to an overview of the skull (Fig. 6b). The extension of the tumor from transversal CT scans can be compared with that of DTF in lateral projection. The images 6c-e show increasing opacification over time. The regions with an impaired blood-brain barrier are enhanced (frontal ring structures, arrows), although the iodine concentration in the blood vessels decreases. In this case, more information ofthe real
Fig. 4. (Case without and the result of after 20 min
0 M. HERBST AND M. FRODER
225
shown in Figure 5
extension of the pathological process were gained from observing the washout characteristics (6f-h). In the early image, 6f (washout of iodine between 30 and 60 set after infusion), only a small area of the tumor appears in the center of the skull, indicating a decrease in opacification during this special time period (30-60 set, dark). The tumor area appears extended in the following image 6g (washout from 30 set to 2 min), whereas in the frontal ring structure the iodine concentration seems to rise further compared with normal tissue (white arrows). The highly delayed flow-off from this large area indicates a
1) Dynamic study of a meningioma with DTF. (See numbers on the adjacent diagram). Section 1.1 section 1.2-1.4 with contrast agent 8, 15 and 20 min after radiopaque infusion. Section 4.3 shows subtraction of 1.2 from 1.1. Section 3.3 is taken after 8 min (section 1.3 from 1.1) and section 2.3 (section 1.4 from 1.1).
Fig. 5. (Case 2) (a) C T scan of a patient with a cerebral metastasis of an ovarian cancer in the right hemisphere; study of a patient with a view of a metastasis from an ovarian cancer at 15 sec. (5b), 9 min (b-c j) dynamid kF (5c) and 15 min (5d). Opacification seems to become more intense with time.
Digital tumor
lluoroscopy
0 M. HERBST AND M. FRODER
Fig. 5.
much larger pathological area than is marked in the “CT survey view” (6b) by coordinates out of the scans. DISCUSSION Radiology and MR-imaging are excellent tools for displaying brain tumors. However, they furnish only indirect information on the tumor localization for the therapy to be instituted by neurosurgeons or radio-oncologists. This is because the tumor location must be reconstructed from several scans and then transferred to a lateral view of the skull. The accuracy of this procedure is impaired for therapy planning by the different geometries of the centrically focussed CT scanner and the eccentrically focussed simulator. Faulty tumor location occurs whenever the lesion is located eccentrically in the head which is quite often the case as most of the tumors are located eccentrically and they are the only ones to be treated by sparing normal brain tissue. So these different geometries cause faulty tumor location for therapy. Even if these machines were located in the operating theatre or in the radiotherapy department it would not help much, because a scan can only furnish 2-dimensional information and the tumor is not visible by fluoroscopy during the action of operation or therapy simulation. So one has to take several scans to cover the complete outline of the tumor and to get enough bone reference points for the indirect reconstruction of the tumor localization on a lateral survey view. All this tends to produce errors, is expensive and time consuming. On the other hand CT units and MR scanners have no collimation system to confine the volume for treatment. In view of these shortcomings, both neurosurgeons and radiotherapists are interested in imaging systems capable of fulfilling their demands. Digital Tumor Fluoroscopy is
221
(Con&)
a new imaging method which comes close to a realization of these demands. It is capable of outlining brain tumors directly in relation to the skull in almost every projection. So the neurosurgeon can estimate during the intervention an optimal approach for stereotactical biopsies and implantation of radionuclids. For percutaneous radiotherapy the system allows the treatment field to be directly localized and marked on the skin or on the treatment mask. DTF differs from CT and MR imaging because it furnishes information not only on the location but also on the dynamics of the blood perfusion of tumors. The sensitivity of DTF to iodine is higher than that of computed tomography because the x-ray energy is optimized to iodine. An increased photon yield of the video system is possible, because the x-ray tube is operated at 40-65 KV/ 5mA. This gives a resolution of iodine from 1.3 per thousand and demonstrates different opacifications, especially in the areas surrounding the tumor. Low iodine concentrations are hardly detectable with computed tomography because the used energy of about 12.5 KV is not favorable to the absorption characteristic of iodine. The contrast medium is administered in less than 1 min. Formerly it was given as an infusion over 5 min and the first images were taken after this time. So a very low concentration of iodine remained in the tissue, because a large amount was excreted by the kidneys during this time. There was no information about the distribution of contrast medium in the tumor area immediately after the start of bolus application. Now imaging starts after 30 seconds. The dynamics of opacification in the tumor region can be studied in relation to the normal vessels. In the case of a highly vascularized tumor, intense opacification appears to decrease in relation to the vessels. If the blood-brain barrier is impaired, the flow-off characteristics are different.
228
1. J. Radiation Oncology 0 Biology 0 Physics
Brain tumors show special characteristics of opacification ( 12). This is confirmed by our studies (64 patients) using DTF. Low grade astrocytomas are marked by opacification at the beginning of the study and by a decrease in relation to that of the blood vessels. Glioblastomas are opacified in the second phase of examination without depletion at the end. Meningiomas show intense blushing at the beginning without change of opacification until the end of the examination. This provides infor-
January
1990, Volume
18, Number
I
mation on the possible histological typing of a tumor, but cannot replace histology. In case of uncertain histology the neurosurgeon can decide interoperatively to take biopsies from areas with different opacification characteristics revealed by DTF. This is not possible with CT because the scanner is located outside the operating theater, and does not give information about the dynamic course of opacification of the whole tumor at the same time.
Fig. 6. (Case 3) (a) CT scan of a patient with a astrocytoma grade 3-4. Two areas of tumor tissue are opacified in the frontal midline and left parietal region; (b) Translation of the astrocytoma from transversal CT scans (one shown in Fig. 6a) to a lateral view of a plain skull radiograph. Scanning from the base of the skull to the top the frontal lesion (see Fig. 6a) is extending to the large, dash-lined area: (c-e) DTF study of an astrocytoma, grade 34. Images taken at 30 set (7c), 2 min (7d) and 13 min (7e). The contrast medium is stored (with more arrows at the beginning of the study and released partially over time; (f-h) DTF study of tissue “destoring” contrast medium (dark structures) between 30 set and 1 min. (6f), between 30 set and 2 min (6g), and between 30 set and 10 min (6h).
Digital tumor
fluoroscopy
0 M. HERBST AND M. FRODER
Fig. 6. (Contd)
229
230
I. J. Radiation
Oncology
0 Biology 0 Physics
January
1990, Volume
18. Number
1
Fig. 6. (Con@
The importance of blood perfusion of tumors gave rise to dynamic CT studies in the early years of computed tomography (7). They furnished a lot of fundamental diagnostic findings. Unfortunately, the time-consuming procedure is not practicable in clinical routine, because only one slice can be studied at a time. Studying the whole tumor region slice by slice would push the expenditure of time and dose to unacceptable levels. The highest dose in DTF (skin dose on the entry side of the skull) was 0.4 R per series with the technical prototype (3,4) and is now 0.064 R with the improved clinical system. Four series of images taken with the technical prototype resulted in a dose of 1.6 R and of 0,64 R with the clinical type after 10 series. Any attempt to reconstruct
sagittal slices with high resolution by CT would inevitably lead to a dose escalation. Overlapping scans with high spatial resolution would give a high integral dose as a result of added scattered radiation fields. The cost of the equipment is determined by the following two system components: (1) The x-ray video chain, which is standard equipment of any operating theatre or simulator room: and (2) The image-processing system is commercially available throughout the world and of standard design. Consequently the acquisition cost of a DTF setup is in the range of normal clinical equipment, and is expected to be negligible in relation to that for CT, because of falling prices for electronic circuits and storage systems in the future.
Digital tumor fluoroscopy 0
M. HERBST AND M. FRODER
231
REFERENCES Froder, M.; Herbst, M. Dynamic studies of brain tumors by the use of digital fluoroscopy. IEEE Trans. MI M l-4: 104-l 13; 1985. Froder, M.; Herbst, M. Darstellung von hirntumoren mit computeruntersttitzter rontgenbildverstarkertechnik. Fortschr. Rontgenstr. 146:66-72; 1987. Friider, M.; Seitzer, D.; Buren, G.; Dieckmann, G. Digitale rontgenvideobildverarbeitung filr die zielpunktberechnung in der stereotaktischen neurochirurgie. Fortschr. Rbntgenstr. 140:84-86; 1984. Greitz, T. Diagnostic and therapeutic strategies in neuroradiology. But. f. Radio]. 58:1 148-l 163; 1985.
5. Herbst, M.; Froder, M.; During, diagnostik und therapieplanung lentherapie (In press).
A. Tumorfluoroskopievon hirntumoren. Strah-
Herbst, M.; Wuermeling, M.; Sauer, R. Fixierung von patienten mit tumoren im Kopf-Hals-Bereich durch individuelle bestrahlungsmasken. In: Wannenmacher, M., ed. Naso-pharynx-tumore. Mtinchen-Wien-Baltimore: Urban & Schwarzenberg; 1983:243-245. Norman, D.; Stevens, E. A.; Wing, S. D.; Lewin, V.; Newton, T. H. Quantitative aspects of contrast enhancement in cranial computed tomography. Radiology 129:683-688; 1978.
