is the Director of the Division of Nuclear Medicine in the
Department of Radiology at Albert Einstein Medical Center; an
Assistant Professor of Radiology at Jefferson Medical College;
and the Director of Nuclear Radiology at Germantown MRI & PET
Center, Philadelphia, PA.
Fusion of anatomic (CT, MRI) with metabolic (PET, SPECT)
scanning has improved the diagnostic accuracy of tumor imaging for
; however, it was practiced at only a few centers due to its
difficulty. Recent advances in imaging hardware and computer
software have made this exciting technique significantly easier to
adopt. This article will review fusion methods and discuss the
benefits of software fusion.
With mental fusion, the anatomic and metabolic studies are
placed side-by-side on the viewbox, and the physician attempts to
correlate the two. There is a relatively high degree of
interobserver variability in interpretation, depending on the skill
and experience of the physician.
Several independent software products, including Medical Image
Merge (Zalen, Cleveland, OH), Hermes Multi-Modality Image Fusion
& Review (Nuclear Diagnostics, Stockholm, Sweden), and software
from the PET camera vendors, allow importation of DICOM-compliant
data from multiple modalities such as PET, SPECT, CT, or MRI in
order to perform software fusion. Fusion can be rigid, where the
organs remain fixed in size; or deformable, where the organs can be
warped and stretched in order to attempt to account for changes in
relative organ position and shape between scans. This is especially
important when scans are taken more than a few minutes apart, as
the patient's lungs, gastrointestinal tract, and bladder are likely
to change in size. Without positioning aids, the patient may also
bend the neck or torso, and the extremities are commonly positioned
differently. A CT scan of the chest is usually done with the arms
up over the head, while a PET scan is usually done with the arms
down by the chest.
Software fusion can be used to fuse OctreoScan (Indium-111
Pentetreo-tide, Mallinkrodt, St. Louis, MO) SPECT with CT (Figure
1). In this example, the initial CT scan was not diagnostic. On a
subsequent day, the SPECT scan was positive, but the location of
the lesion was uncertain. The DICOM-compliant CT and SPECT scans
were imported into a MiraView/Reveal MVS dual-monitor workstation
(Mirada Solutions Ltd., Oxford, UK, and CTI Medical Imaging Inc.,
Knoxville, TN). Automatic rigid fusion was applied, the results
were inspected, manual correction was performed by translating and
rotating the PET data, then automatic deformable fusion was
performed. At each step, the fusion was inspected visually along
the transverse, sagittal, and coronal planes by changing the
transparency of the pseudocolored overlay SPECT and adjusting the
overlay position until the liver, spleen, and kidneys matched up
with their locations on the CT scan. Displaying the data on two
large monitors was very helpful. Then, simultaneous SPECT and CT
crosshairs were turned on, pinpointing an insulinoma at the
junction of the body and tail of the pancreas on the CT scan.
Software fusion works better if there are organs concentrating
physiologic radiopharmaceuticals near the region of interest; these
can be used as landmarks to check the accuracy of the registration.
In addition, a more similar body position between scans will
improve the success rate of the registration.
Figure 2 presents an example of
F-FDG-PET/CT software fusion in the chest and was done during
preoperative staging of a head and neck squamous cell carcinoma.
The initial CT scan was done in a breath-hold at maximum
inspiration, while the PET scan was done during quiet breathing.
Automatic rigid fusion was done. Then direct manual translation and
rotation of the PET data was performed using internal landmarks to
adjust the PET to the CT, superimposing the FDG in vertebral body
red marrow and in mediastinal blood pool in the ascending and
descending aorta, right and left pulmonary arteries, and superior
vena cava to their locations on the CT. Then automatic deformable
fusion was applied to compensate for differences in lung volumes
between the PET and CT scans. The PET image is converted to a
semitransparent orange pseudo-color, and this is overlaid onto the
CT image. The PET overlay transparency is varied dynamically with
the computer mouse, localizing the tumor to a normal-size, right
hilar node on the CT scan. Note that the arms are positioned
farther away from the chest walls on the CT scan, whereas they are
touching the chest on the PET scan; after fusion, the
semi-transparent orange FDG uptake in arm muscles is projected over
empty space between the chest walls and the arms on the CT scan. If
this were the region of clinical interest, additional processing
with registration using manually placed landmarks would be
necessary to overlay the arm on the PET scan to the arm on the CT
scan. Fiducial markers can be placed externally
to help guide the fusion.
Software fusion with positioning devices
By ensuring the patient is positioned the same way during each
scanning session, fixation devices allow accurate fusion.
This is especially important for radiation therapy planning.
Radiation treatment sessions take place over weeks, so radiation
oncologists use a variety of devices to keep the patient
immobilized in order to treat tumors without over-radiating normal
In another example, following diagnosis of squamous cell cancer
of the head and neck, a CT scan was done in a custom-molded mask
for radiation treatment planning (Figure 3). The patient was to
undergo radiation treatments for several weeks while wearing the
same mask. On a different day at a different location, a FDG-PET
scan was done using the same mask. Automatic rigid fusion was
applied with a single command. Because the patient positioning was
similar, the software was able to pinpoint an unsuspected
metastasis, changing the radiation treatment plan. Before
performing studies in positioning devices, both CT and PET
technologists should be trained in the proper application of these
Despite new artifacts that need to be accounted for,
hardware fusion with combined PET/CT scanners has been shown to
significantly improve diagnostic accuracy over mental fusion.
The scans are taken on the same table only a few minutes apart, and
the patient should stay in the same position. The study is much
faster because the CT scan is also used for attenuation correction
of the emission PET data. There is still a time gap of several
minutes between scans, so there may still be patient motion and
artifact from differences in stages of respiration between PET and
CT, bladder filling, and peristalsis.
Good software fusion and hardware fusion are synergistic, and
having both will lead to a more accurate, confident diagnosis and
localization. Hardware fusion gives better data for software fusion
to work with, and software fusion can complement hardware fusion by
tweaking the registration more precisely.
Despite these proven benefits, there remains significant
difficulty in obtaining adequate reimbursement for the additional
cost of the integrated scanner and the increased complexity of
acquisition and interpretation. Not every facility will be able to
afford combined PET/CT. If a PET/CT scanner is not available,
software fusion still conveys significant clinical benefits by
adding anatomic information to metabolic imaging.