ht. J. Radiafion Chcology
l
Biol. Phy~.,~1979, Vol. 5, pp. 445447.
Pergamon Press Ltd.
Printed in the U.S.A.
Editorial
COMPUTED
TOMOGRAPHY
IN PLANNING
RADIATION
THERAPY
MICHAEL GOITEIN, Ph.D. Division of Radiation Biophysics, Department of Radiation Medicine, Massachusetts General Hospital, and Harvard Medical School, Boston, MA 02114, U.S.A. Computed
tomography,
Treatment
planning.
The delineation of tumors and adjacent structures has traditionally been a weak link in radiation therapy. Several studies suggest that computed tomography can make a major impact in this area. Munzenrider et a/.,’ for example, found that in 15 of 75 patients (20%), computed tomography (CT) scans revealed areas of tumor which would otherwise not have been included in the radiation field. Coverage would have been only marginal in a further 20 patients (27%). These and other considerations led to the judgement that CT scans were “essential” for 55% of the patients. These impressive conclusions have been supported by a number of other recent studies.‘*2A*6” Despite the contributions of CT scans to radiation therapy, few hospitals treat enough patients to justify the purchase of a CT scanner dedicated solely to radiation therapy. Rather, scanners bought primarily for diagnostic purposes are used secondarily for planning therapy. Under these circumstances it is reasonable to propose that therapy departments play some role in defining specifications for CT scanners bought by their institutions. This is desirable since there are additional requirements in a scanner used for radiation therapy. This question has recently been considered by a subcommittee of the Committee on Radiation Oncology Studies.” The special needs of radiation therapy stem from: (1) the necessity of fully delineating tumors and adjacent normal structures; (2) the necessity of scanning the patient in a position which is appropriate and can be reproduced during treatment; and (3) the desirability of using CT scans to assist in calculations of dose.
This idea arises from the fact that only modest resolutions are necessary for dosimetric calculations.3 However, tumor and normal tissue localization is an essential component of treatment planning. Whereas the primary problem which a diagnostician addresses is whether tumor is present, the therapist must determine the extent of the lesion in all directions; the position and margins of adjacent structures; and the relationship of these to external landmarks which can be used to position the patient for treatment. In this light, the requirements for CT scanners used for treatment planning are at least as demanding, and perhaps more so, than those for machines used only for diagnostic purposes. At present, the absorption coefficient resolutions obtained with state-of-the-art scanners are marginally adequate to differentiate many tissues of interest. In particular, there may be systematic errors which lead to uncertainties in the absorption coefficient associated with each tissue. Such errors can probably be reduced by the use of a CT scanner equipped with an iterative beam hardening correction.7 Clearly, there is no justification for relaxation of the absorption coefficient resolution demanded of a CT scanner used in therapy. Perhaps the most obvious difference between scans used for treatment planning and those used for diagnostic purposes is the necessity for the former to include the entire volume of interest rather than samples at a limited number of levels. Such a volumetric scan is required to delineate the tumor fully in three dimensions and avoid missing small extensions. Contiguity of both normal and abnormal structures in adjacent scans is important in defining disease and can be lost if scans are too far apart. Moreover, multiple closely spaced scans permit coronal and sagittal reconstructions and, indeed, reconstructions in arbitrary planes, thus significantly improving the interpretation of the scan. Coronal sections of the head, for example, have been of great benefit,” and
Tumor and normal tissue localization
There is a widespread misconception that in treatment planning the resolution of both spatial and absorption coefficient measurements need be less accurate than that necessary for diagnostic purposes. 445
446
Radiation
Oncology
0 Biology
0 Physics
one can anticipate similar dividends elsewhere in the body. Another benefit of closely spaced scans is that spatial resolution in the plane of the scan will be better in thinner sections. It seems likely that the resolutions of recently developed scanners, which are approaching 7 line pairs per centimeter, are well matched to present needs. HOW closely spaced must scanning be for good treatment planning? The answer depends in part upon the nature of the disease and on the region of interest. The CT scanner at best delineates bulk disease and, as always, the therapist must determine final margins on the basis of predictions of microscopic extension. He must also consider the margins of nearby normal structures. Sampling theory suggests that the distance between slices should be one-half the desired spatial accuracy. In general, it seems reasonable to define field margins with an accuracy of 1 cm, which implies a slice spacing of 5 mm. Even finer samplings would be desirable for the more critical applications. It should be noted that the radiation dose received by the patient in scanning multiple thin sections is of the same order as that received in scanning less numerous thick sections. Where noise is a problem, adjacent thin sections (whether transverse or coronal or sagittal) can be averaged together. Since the full volume of suspected disease and some extension beyond it must be scanned in one continuous maneuver, the area of interest may easily be as large as 30 cm along the patient axis. If scans are taken 5 mm apart, then as many as 60 slices may be required. The need for close spacing has been underscored by our experience4 with a scanner capable of taking only about 15 slices during the course of an hour or more. The coarseness of such sampling, with separations of perhaps 2 cm between scans, often made it impossible to choose superior and inferior margins with sufficient accuracy. Moreover, the length of time required for the procedure made adequate patient immobilization difficult. Patient positioning Reproducibility of the patient position depends on a multitude of factors, the most important of which is the training and interest of the CT technicians. Without the full and active cooperation of the CT staff it is impossible to realize the full benefits of CT scanning. The capacity to simulate AP and lateral (and perhaps oblique) “plain films” by moving the patient past the stationary X-ray tube and detectors (called a “scout view” by one manufacturer) is a valuable aid in the proper indexing and registration of the patient and should be considered an essential component in a scanner used for therapy. Anyone who has used CT scans in planning radiation therapy has spent many
March
1979, Volume
5, No.
3
hours trying with varying degrees of success to overlay the scan planes on AP and lateral films. A scan sequence should be completed in perhaps lo-15 min in order to ensure stable reproducible patient positioning (although not all scans need be reconstructed within that time). Thus, for a sequence of 60 scans, scans must be taken every 10 set or so. This repetition rate places stringent requirements on the X-ray tube which must have high anode and housing cooling rates. In addition, efficient utilization of the emitted photons is necessary. This implies good geometric and conversion efficiencies for the detection of photons. Available scanners differ quite widely in these factors. A large patient aperture greatly facilitates patient positioning. An aperature diameter of 60 cm, now available in many machines, should probably be considered the minimum acceptable size. It is highly desirable to be able to see lateral skin marks. When scanning in the pelvis, for example, scan diameters of 40 cm or less require exquisite attention to patient alignment if such marks are not to be lost. A large reconstruction diameter is desirable from this point of view. Absence of bolus material is also an important requirement for appropriate and reproducible patient positioning. A couch with a flat top is relatively unimportant, since a flat filler can readily be designed for non-flat couches. Calculation of dose Several commercial companies are presently developing programs which overlay conventional treatment planning techniques onto CT scans. I am in sympathy with recent comments in these pages,’ expressing reservations about the value of automated implementation of such schemes at the present time. The goal of the treatment planning process is to assist in the choice of the configuration of radiation beams (or sources) and, most importantly, to design the field shape (and, if needed, a compensating filter) for each field. This should involve three dimensional codes which: permit viewing the target volume and selected normal structures from the perspective of a therapy unit; permit the delineation of a field outline from such a display; and, perhaps, calculate dose in three dimensions in the face of inhomogeneities. We are far from having these capabilities now. In summary, the features of a CT scarier used for radiation therapy should include: (1) excellent spatial and absorption coefficient resolution, (2) capability of scanning a large number (about 60) of relatively thin (5 mm or less) sections continuously at a rate of one section per 10 set, (3) capacity to generate an AP and lateral “plain film” view referenced to scan sections,
*
Computed tomography
in planning radiation therapy 0 M. GOITEIN
(4) large patient aperture and large diameter of reconstruction, (5) scan without bolus material, (6) high anode and X-ray tube housing cooling rates and high geometric and photon conversion efficiencies of the detectors, (7) sagittal and coronal display capability. Clearly, not all of these features are available as standard options on available equipment. In particular, item (2) is the most difficult specification to meet, and perhaps the most contentious. However, few would argue that it is not desirable. I am confident that several manufacturers are able at this time to meet this and the other specifications with
447
little additional cost. The above specifications are closer to those recommended by the CROS Subcommittee” for a dedicated radiotherapy scanner than for a “diagnostic scanner with therapy options”. This arises both because of an increased awareness of the importance of these requirements for treatment planning and, because the state-of-the-art in computed tomography has made these requirements quite realistic, both technically and economically, even for scanners used primarily for diagnostic purposes. These capabilities can be expected to enhance the diagnostic value of such scanners, as well as increase the benefits to be gained from an already useful tool in radiation therapy.
REFERENCES 1. Brizel, H.E., Livingston, P.A., Grayson, E.V.: RadioJ.S.: Radiotherapy treatment planning: initial clinical therapeutic applications of pelvic computerized tomogexperience using a whole body CT scanner directly raphy, presented at the 63rd Scientific Assembly and linked to a treatment planning computer. J. Comput. Annual Meeting of the Radiological Society of North Assist Tomogr. 2: 525, 1978. America, 27 Nov.-2 Dec. 1977. 7. Joseph, P.M., Spital, R.D.: A method for correcting 2. Emami, B., Melo, A., Carter, B.L., Munzenrider, J.E. bone induced artifacts in computed tomography scanand Piro, A.J.: Value of computed tomography in ners. J. Comput. Assist. Tomogr. 2: 100-108, 1978. radiotherapy of lung cancer. Am. J. Roentgenol. 131: 8. Munzenrider, J.E., Pilepich, M., Rene-Ferrero, J.B., 63-67, 1978. Tchakarova, I., Carter, B.L.: Use of body scanner in 3. Geise, R.A., McCullough, E.C.: The use of CT scanners radiotherapy treatment planning. Cancer 40: 170-179, megavoltage photon-beam therapy planning. 1977. :adiology 124: 133-141, 1977. 4. Goitein, M., Wittenberg, J., Doucette,
9. Ragan, D.P., Perez, J., Friedberg,
C.,
Mendiondo, M., Gunderson, L., Lingood, R., Shipley, W.U., Fineberg, M.V.: The value of CT in radiotherapy treatment planning. J. Comput. Assist. Tomogr. 2: 524 525, 1978.
5. Hendee, W.R.: Computed tomography in radiation therapy treatment planning. Znt. J. Radiat. Oncol. Biol. Phys. 4: 539-540, 1978. 6. Hobday, P.A., Husband, J.E., Parker, R.P., Macdonald,
C.A.: Efficacy
of CT-assisted
two-
dimensional treatment planning: Analysis of 45 patients. Am. J. Roentgenol. 131: 75-79, 1978. 10. Stewart, J.R., Hicks, J.A., Boone, M.L.M., Simpson, L.D.: Computed tomography in radiation therapy. Znt. J. Radiat. Oncol. Biol. Phys. 4: 313-324, 1978. 11. Tadmor, R., New, P.F.J.: Computed tomography of the orbit with special emphasis on corona1 sections-II. Pathalogical anatomy. J. Comput. Assist. Tomogr. 2: 34-44,
1978.
l
Biol. Phy~.,~1979, Vol. 5, pp. 445447.
Pergamon Press Ltd.
Printed in the U.S.A.
Editorial
COMPUTED
TOMOGRAPHY
IN PLANNING
RADIATION
THERAPY
MICHAEL GOITEIN, Ph.D. Division of Radiation Biophysics, Department of Radiation Medicine, Massachusetts General Hospital, and Harvard Medical School, Boston, MA 02114, U.S.A. Computed
tomography,
Treatment
planning.
The delineation of tumors and adjacent structures has traditionally been a weak link in radiation therapy. Several studies suggest that computed tomography can make a major impact in this area. Munzenrider et a/.,’ for example, found that in 15 of 75 patients (20%), computed tomography (CT) scans revealed areas of tumor which would otherwise not have been included in the radiation field. Coverage would have been only marginal in a further 20 patients (27%). These and other considerations led to the judgement that CT scans were “essential” for 55% of the patients. These impressive conclusions have been supported by a number of other recent studies.‘*2A*6” Despite the contributions of CT scans to radiation therapy, few hospitals treat enough patients to justify the purchase of a CT scanner dedicated solely to radiation therapy. Rather, scanners bought primarily for diagnostic purposes are used secondarily for planning therapy. Under these circumstances it is reasonable to propose that therapy departments play some role in defining specifications for CT scanners bought by their institutions. This is desirable since there are additional requirements in a scanner used for radiation therapy. This question has recently been considered by a subcommittee of the Committee on Radiation Oncology Studies.” The special needs of radiation therapy stem from: (1) the necessity of fully delineating tumors and adjacent normal structures; (2) the necessity of scanning the patient in a position which is appropriate and can be reproduced during treatment; and (3) the desirability of using CT scans to assist in calculations of dose.
This idea arises from the fact that only modest resolutions are necessary for dosimetric calculations.3 However, tumor and normal tissue localization is an essential component of treatment planning. Whereas the primary problem which a diagnostician addresses is whether tumor is present, the therapist must determine the extent of the lesion in all directions; the position and margins of adjacent structures; and the relationship of these to external landmarks which can be used to position the patient for treatment. In this light, the requirements for CT scanners used for treatment planning are at least as demanding, and perhaps more so, than those for machines used only for diagnostic purposes. At present, the absorption coefficient resolutions obtained with state-of-the-art scanners are marginally adequate to differentiate many tissues of interest. In particular, there may be systematic errors which lead to uncertainties in the absorption coefficient associated with each tissue. Such errors can probably be reduced by the use of a CT scanner equipped with an iterative beam hardening correction.7 Clearly, there is no justification for relaxation of the absorption coefficient resolution demanded of a CT scanner used in therapy. Perhaps the most obvious difference between scans used for treatment planning and those used for diagnostic purposes is the necessity for the former to include the entire volume of interest rather than samples at a limited number of levels. Such a volumetric scan is required to delineate the tumor fully in three dimensions and avoid missing small extensions. Contiguity of both normal and abnormal structures in adjacent scans is important in defining disease and can be lost if scans are too far apart. Moreover, multiple closely spaced scans permit coronal and sagittal reconstructions and, indeed, reconstructions in arbitrary planes, thus significantly improving the interpretation of the scan. Coronal sections of the head, for example, have been of great benefit,” and
Tumor and normal tissue localization
There is a widespread misconception that in treatment planning the resolution of both spatial and absorption coefficient measurements need be less accurate than that necessary for diagnostic purposes. 445
446
Radiation
Oncology
0 Biology
0 Physics
one can anticipate similar dividends elsewhere in the body. Another benefit of closely spaced scans is that spatial resolution in the plane of the scan will be better in thinner sections. It seems likely that the resolutions of recently developed scanners, which are approaching 7 line pairs per centimeter, are well matched to present needs. HOW closely spaced must scanning be for good treatment planning? The answer depends in part upon the nature of the disease and on the region of interest. The CT scanner at best delineates bulk disease and, as always, the therapist must determine final margins on the basis of predictions of microscopic extension. He must also consider the margins of nearby normal structures. Sampling theory suggests that the distance between slices should be one-half the desired spatial accuracy. In general, it seems reasonable to define field margins with an accuracy of 1 cm, which implies a slice spacing of 5 mm. Even finer samplings would be desirable for the more critical applications. It should be noted that the radiation dose received by the patient in scanning multiple thin sections is of the same order as that received in scanning less numerous thick sections. Where noise is a problem, adjacent thin sections (whether transverse or coronal or sagittal) can be averaged together. Since the full volume of suspected disease and some extension beyond it must be scanned in one continuous maneuver, the area of interest may easily be as large as 30 cm along the patient axis. If scans are taken 5 mm apart, then as many as 60 slices may be required. The need for close spacing has been underscored by our experience4 with a scanner capable of taking only about 15 slices during the course of an hour or more. The coarseness of such sampling, with separations of perhaps 2 cm between scans, often made it impossible to choose superior and inferior margins with sufficient accuracy. Moreover, the length of time required for the procedure made adequate patient immobilization difficult. Patient positioning Reproducibility of the patient position depends on a multitude of factors, the most important of which is the training and interest of the CT technicians. Without the full and active cooperation of the CT staff it is impossible to realize the full benefits of CT scanning. The capacity to simulate AP and lateral (and perhaps oblique) “plain films” by moving the patient past the stationary X-ray tube and detectors (called a “scout view” by one manufacturer) is a valuable aid in the proper indexing and registration of the patient and should be considered an essential component in a scanner used for therapy. Anyone who has used CT scans in planning radiation therapy has spent many
March
1979, Volume
5, No.
3
hours trying with varying degrees of success to overlay the scan planes on AP and lateral films. A scan sequence should be completed in perhaps lo-15 min in order to ensure stable reproducible patient positioning (although not all scans need be reconstructed within that time). Thus, for a sequence of 60 scans, scans must be taken every 10 set or so. This repetition rate places stringent requirements on the X-ray tube which must have high anode and housing cooling rates. In addition, efficient utilization of the emitted photons is necessary. This implies good geometric and conversion efficiencies for the detection of photons. Available scanners differ quite widely in these factors. A large patient aperture greatly facilitates patient positioning. An aperature diameter of 60 cm, now available in many machines, should probably be considered the minimum acceptable size. It is highly desirable to be able to see lateral skin marks. When scanning in the pelvis, for example, scan diameters of 40 cm or less require exquisite attention to patient alignment if such marks are not to be lost. A large reconstruction diameter is desirable from this point of view. Absence of bolus material is also an important requirement for appropriate and reproducible patient positioning. A couch with a flat top is relatively unimportant, since a flat filler can readily be designed for non-flat couches. Calculation of dose Several commercial companies are presently developing programs which overlay conventional treatment planning techniques onto CT scans. I am in sympathy with recent comments in these pages,’ expressing reservations about the value of automated implementation of such schemes at the present time. The goal of the treatment planning process is to assist in the choice of the configuration of radiation beams (or sources) and, most importantly, to design the field shape (and, if needed, a compensating filter) for each field. This should involve three dimensional codes which: permit viewing the target volume and selected normal structures from the perspective of a therapy unit; permit the delineation of a field outline from such a display; and, perhaps, calculate dose in three dimensions in the face of inhomogeneities. We are far from having these capabilities now. In summary, the features of a CT scarier used for radiation therapy should include: (1) excellent spatial and absorption coefficient resolution, (2) capability of scanning a large number (about 60) of relatively thin (5 mm or less) sections continuously at a rate of one section per 10 set, (3) capacity to generate an AP and lateral “plain film” view referenced to scan sections,
*
Computed tomography
in planning radiation therapy 0 M. GOITEIN
(4) large patient aperture and large diameter of reconstruction, (5) scan without bolus material, (6) high anode and X-ray tube housing cooling rates and high geometric and photon conversion efficiencies of the detectors, (7) sagittal and coronal display capability. Clearly, not all of these features are available as standard options on available equipment. In particular, item (2) is the most difficult specification to meet, and perhaps the most contentious. However, few would argue that it is not desirable. I am confident that several manufacturers are able at this time to meet this and the other specifications with
447
little additional cost. The above specifications are closer to those recommended by the CROS Subcommittee” for a dedicated radiotherapy scanner than for a “diagnostic scanner with therapy options”. This arises both because of an increased awareness of the importance of these requirements for treatment planning and, because the state-of-the-art in computed tomography has made these requirements quite realistic, both technically and economically, even for scanners used primarily for diagnostic purposes. These capabilities can be expected to enhance the diagnostic value of such scanners, as well as increase the benefits to be gained from an already useful tool in radiation therapy.
REFERENCES 1. Brizel, H.E., Livingston, P.A., Grayson, E.V.: RadioJ.S.: Radiotherapy treatment planning: initial clinical therapeutic applications of pelvic computerized tomogexperience using a whole body CT scanner directly raphy, presented at the 63rd Scientific Assembly and linked to a treatment planning computer. J. Comput. Annual Meeting of the Radiological Society of North Assist Tomogr. 2: 525, 1978. America, 27 Nov.-2 Dec. 1977. 7. Joseph, P.M., Spital, R.D.: A method for correcting 2. Emami, B., Melo, A., Carter, B.L., Munzenrider, J.E. bone induced artifacts in computed tomography scanand Piro, A.J.: Value of computed tomography in ners. J. Comput. Assist. Tomogr. 2: 100-108, 1978. radiotherapy of lung cancer. Am. J. Roentgenol. 131: 8. Munzenrider, J.E., Pilepich, M., Rene-Ferrero, J.B., 63-67, 1978. Tchakarova, I., Carter, B.L.: Use of body scanner in 3. Geise, R.A., McCullough, E.C.: The use of CT scanners radiotherapy treatment planning. Cancer 40: 170-179, megavoltage photon-beam therapy planning. 1977. :adiology 124: 133-141, 1977. 4. Goitein, M., Wittenberg, J., Doucette,
9. Ragan, D.P., Perez, J., Friedberg,
C.,
Mendiondo, M., Gunderson, L., Lingood, R., Shipley, W.U., Fineberg, M.V.: The value of CT in radiotherapy treatment planning. J. Comput. Assist. Tomogr. 2: 524 525, 1978.
5. Hendee, W.R.: Computed tomography in radiation therapy treatment planning. Znt. J. Radiat. Oncol. Biol. Phys. 4: 539-540, 1978. 6. Hobday, P.A., Husband, J.E., Parker, R.P., Macdonald,
C.A.: Efficacy
of CT-assisted
two-
dimensional treatment planning: Analysis of 45 patients. Am. J. Roentgenol. 131: 75-79, 1978. 10. Stewart, J.R., Hicks, J.A., Boone, M.L.M., Simpson, L.D.: Computed tomography in radiation therapy. Znt. J. Radiat. Oncol. Biol. Phys. 4: 313-324, 1978. 11. Tadmor, R., New, P.F.J.: Computed tomography of the orbit with special emphasis on corona1 sections-II. Pathalogical anatomy. J. Comput. Assist. Tomogr. 2: 34-44,
1978.
