Twenty-five years ago, the diagnostic radiology department at the Peter Bent Brigham Hospital, Boston, had acquired a new piece of imaging technology, one of the early MRI scanners, which produced remarkable transverse views of body anatomy. Since we radiation oncologists had been using gantry-mounted linear accelerators for about 10 years, we found these views to be particularly helpful as part of the tumor localization process. I remember ordering a scan of a patient with esophagus cancer in the prone position using a flat plywood table insert that we had rigged up. Unfortunately, the scan was done in the supine position using the routine curved tabletop. When I asked the director if he would please repeat the study, he was reluctant to squander one of the limited slots on a unit that was 100% utilized. He asked what was so important about patient position. I explained that our treatment was no longer just from the front and back, and in the prone position the esophagus dropped away from the spine and spinal cord, which was the dose-limiting tissue. Further, deformations of the patient’s surface caused by the curvature of the table propagated themselves deeper into the tissues, making therapy less precise when the patient was transferred to the flat table for treatment.
The unit director was incredulous, but agreed to repeat the study. Two hours later, I received a page and was told to call the disgnostic radiologist “stat.” The radiologist asked me, “How did you know that the organs moved so much in the prone position. The tumor is a full 2 cm further away from the spine! We should write this up.” I cannot tell you how many studies I have had to repeat over the past 25 years because of a fundamental difference in how we and the diagnostic imagers understand the purposes of imaging.
A clue to this communication gap can be found in the actual name of the specialty: diagnostic imaging. Radiologists are accustomed to asking what-is questions: “What is that process? Is it inflammatory, neoplastic, infectious?” By the time we radiation oncologists get involved, the difficult diagnostic work is completed, usually thanks to collaboration between radiologists and surgeons. We, on the other hand, need to get the answer to where-is questions: “Where is that tumor and what is its relationship to the surrounding normal tissues?”
Thirty years ago, most radiation oncologists started out with at least some training in diagnostic radiology. The radiation simulator was a fluoroscopic unit on a gantry with the same geometry as a linear accelerator. By use of the various iodinated contrast agents, the tumor could usually be localized with respect to lung, bone, and soft tissue. When it could not, clips placed by the surgeon could be used as landmarks. There was no real need to interact with the diagnostic imager since the oncologist was also one. As radiation oncology matured, the radiation oncologist lost his or her specific competence as an imager and was required to consult with the diagnostician, like any other specialist. Unfortunately, we did not explain the type of information that we needed.
The revolution in imaging has not had the impact on treatment accuracy that perhaps it could because of this divide between the diagnostic and therapeutic radiologist. Paradoxically, the advent of three-dimensional treatment requires much more interaction between the two specialties. The radiation oncology community has defined a number of different volumes of interest to be specified as part of treatment planning. The gross treatment volume (GTV) is that portion of the body where the tumor can be delineated by imaging or clinical means. Surrounding this structure is a larger volume that contains presumed subclinical tumor extension via direct microscopic and lymphatic permeation. This volume is called the clinical tumor volume (CTV). Finally, a shell of tissue is added around the CTV to include both setup error (both random and systematic), and organ movement. The latter is called the planning tumor volume (PTV). When radiation oncologists and therapy physicists sit down to create three-dimensional tumor treatment programs, they now treat the PTV to the specified dose, accepting that the CTV and GTV within the PTV and the normal tissues without will invariably receive higher doses than had this exercise not been done.
It is my contention that a close collaboration between the diagnostic onco-imager and the radiation oncologist can shrink the shells produced by both clinical uncertainty and body movement to a minimum, allowing for higher doses to be delivered to the tumor and improving the chance for cure. Functional MRI, magnetic resonance spectroscopy (MRS), positron emission tomography (PET) scanning, and labeled antibody-to-tumor antigen scanning can all define the CTV much better than clinical estimation. Dynamic CT cines can elucidate the extent of body movement and shrink the PTV.
For this collaboration to be fruitful, however, both partners must understand the needs and limitations of the other. Of primary importance is the need to image and treat the patient in the same position. Things have not changed since my first misunderstanding with the diagnostician at the Brigham 25 years ago. Patients are now treated in rigid body molds to minimize movement. Polyacrylamide plastic stents are placed in the mouth and other orifices to push normal tissues away from the radiation beam. The apertures of the various scanners and the MRI body and/or surface coils must be modified to allow for the use of the therapeutic immobilization devices. Patients should be imaged with their stents in place and the stents should be of tissue-equivalent or density-null material so as not to perturb the images.
Second, the diagnostic phantoms used to calibrate the imaging modality must include geometric check devices to make sure that distances are being measured properly. If the actuators on the spiral CT table are inaccurate, there will be longitudinal distortion and tumors will be measured incorrectly. Likewise, if there is magnetic distortion at the edges of the scanning window for MRI, the tumor volume may be estimated incorrectly. Both diagnostic and therapeutic radiologists must understand the volumetric error present within the newer nuclear medicine (PET and radio-labeled antibody) detectors so that the CTV is neither over- or underestimated.
Third, both diagnostic imagers and radiation oncologists must encourage industry to rapidly incorporate the latest DICOM (Digital Imaging and Communications in Medicine) standards to allow for multimodal image registration. The various nuclear medicine modalities will not have the same degree of geometric accuracy as CT or MR in the near future and will not easily demonstrate the relationship of normal tissue structures to the tumor. Likewise, CT probably will not give us as good an understanding of the microscopic tumor extensions as would PET or radioimmune scanning. Finally, MR cannot show us the bones as well as CT for correlation with the planar information obtained at conventional fluoroscopic simulation. Clearly for the foreseeable future, we will need to correlate all of these modalities to build a picture of the virtual tumor in the virtual body for 3-D planning purposes. This task will be made immeasurably easier when image registration and correlation are automated on any and all of the imaging and therapy workstations that will be swapping digital data in the near future.
While the push for subspecialization is here to stay, it may be time once again to consider cross-rotation of oncology and diagnostic trainees. It might also be useful for community and academic radiologists and oncologists to spend a bit of time in each other’s departments to familiarize each with the other’s needs and capabilities. At a minimum we should be talking with each other before studies are ordered and interpreted.
Christopher M. Rose, MD, is associate and technical director, Radiation Oncology, Valley Radiotherapy Associates Medical Group, Los Angeles, and the president of the American Society for Therapeutic Radiology and Oncology, Fairfax, Va.