The fusion of computed tomography (CT) and single-photon emission computerized tomography (SPECT) antibody images can enhance the information provided by either single modality by providing precise anatomical-functional correlation. Functional abnormalities seen on the low resolution SPECT antibody images can be precisely located with specific anatomic structures seen in the high resolution CT images. External fiducials located on each image modality aid in the automated registration, alignment and matching of CT and SPECT antibody images. The potential benefits of multi-modal fusion include: (A) the discrimination of more subtle activity peaks using anatomic organ segmentation, (B) temporal discrimination of recurrent disease, (C) assessment of residual activity post-surgery and (D) automated localization of significant focal activity. In addition, the correlation of function with anatomy may be used to establish the physiologic status of ambiguously identified objects in the anatomic image.

Metastatic malignancies in the gastrointestinal (GI) tract are of major concern in the United States and are a leading cause of cancer-related death. Incomplete preoperative identification of lesion sites coupled with low sensitivity in identifying sites of recurrence of metastatic disease post surgery constrain potential management care options. Radiolocalization of lesions plays a significant role in the care and management of clinical patients with malignancies. However, in cases where the sites are extensive and complex, protocols employing bone scans, liver scans, gallium scans and other similar techniques have proven to be of variable utility.

Imaging with radiolabeled monoclonal antibodies (Mab) provides tissue-specific localization of primary and metastatic disease. Functional studies with SPECT Mab suffer difficulties in complete diagnosis and identification of disease. Similar findings have been reported for computed tomography (CT) imaging alone. Birnbaum (1) indicated that although single-photon emission computerized tomography (SPECT) provides unique functional information it is limited by poor anatomic detail. Non-specific localization of the antibody makes image interpretation difficult, since contrast improvement offered by SPECT imaging increases both the tumor specific and non-specific background uptake.

The correlation of anatomic and functional modalities permits improved focal localization and at the same time enhances the identification of regions with non-specific activity. Swayne (2) reported that functional abnormalities detected with inherently low resolution functional imaging can be anatomically localized. The more accurate localization of focal activity can provide valuable information to physicians and surgeons for both surgical resection and for the guidance and assessment of radiation and chemotherapy. Collier (3) reported on a study with 111In CYT - 103 used in the detection of colorectal cancer that clinical management changed in 27% of the cases.

The primary role of correlation in chest and abdomen Kramer and Noz, (4) is to increase the specificity of SPECT and enhance the sensitivity of CT. Nonspecific antibody concentration, particularly in the blood pool or other normal anatomic structure is more easily recognized with SPECT, particularly in early Mab imaging. The identification of unsuspected tumor masses on CT (5,6) due to adjacent underfilled small bowel is aided by multimodal image correlation.

The fusion process depends on the accurate registration of the CT and SPECT images. Image registration requires the congruent matching of the SPECT and CT coordinates and simultaneously matching the size and voxel size of the SPECT and CT images.

There are two basic approaches to registration for tomographic images (7): internal landmarks and external landmarks or fiducials. Each of these are compromised by various imaging artifacts and subject movement. The use of internal landmarks is subject to errors due to the variable physiologic state of the functional image.

The use of external fiducials has several advantages for body imaging. External fiducials can be constructed with unique geometries which provide each slice with a unique external coordinate signature. External fiducials are unambiguously seen in each image modality and image and can be used to rapidly translate and rotate the target and object images.

Figure 1 shows an external fiducial marker system (Bio-Imaging Technologies Inc., West Trenton, N.J.) developed for use with monoclonal antibodies. The fiducials provide external markers which can be identified in the different modalities and provide the capability for semi automatic alignment. The fiducials provide a common set of body coordinates, which can be used both to locate images and to track local lesions in recurrent cases.

Figure 2 illustrates the use of an external fiducial. The left panel shows the initial fit of the CT (structural) and SPECT (functional) data sets. The right panel shows the correspondent of the two fiducials after translation and rotation. Figure 3 presents computed coronal and sagittal CT/SPECT fusion images post registration and clearly illustrate the data set correspondence.

Digitial image data can be processed and resampled on demand using computer techniques. Computer programs provide automated pixel resampling to a common pixel size and match the coordinate systems of both imaging modalities. The pixel size and slice thickness of each image data set is resampled to a uniform and common pixel size in all three dimensions.

Pairs of images, anatomically located, can be generated on user demand. Multiple sets of image slices can be generated with any thickness or spacing for each data set. Paired images are generated at the same pixel size and thickness. The data can be presented as matched pairs, or it can be overlaid and graphically fused to highlight the anatomical features with the SPECT uptake. Orthogonal planes can be automatically displayed centered on specific anatomic coordinates and be used to generate referent views. In addition these non-transaxial views of the CT data set can be generated both as slices and as pseudo-x-rays as shown in Figure 4.

Image fusion is the process of image superposition using two different image types: anatomic (CT, MRI) and functional (SPECT, positron emission tomography [PET]). This process provides the functional (SPECT antibody concentration) information in an anatomic context provided by the CT image. Given two data sets registered in three orthogonal axes, comparative slices can be created through any arbitrary point and at any orientation.

Generally, a fixed set of body-aligned axes are chosen as the reference or base-line. Since the two image data sets are spatially registered, overlay of congruent planes will show both radiolabeled antibody concentration and underlying anatomy. The contrast of either or both image planes can be adjusted to highlight focal lesions or organ or body features. Figure 5 illustrates the fusion process. Fusion is provided by splitting the display palette into two distinct color regimes. The anatomic image is assigned a gray scale palette and forms the background or underlay of the combined image. The functional image can be displayed in an overlay fashion in a different contrasting color palette. Both the contrast and transparency of either or both images can be singly or simultaneously adjusted.

False color or gray-scale image representing the SPECT and CT images are blended for optimum visual contrast. The precise spatial location of image slices in the context of total body geometry can be simultaneously displayed, thus providing anatomic context to the localization of the metastatic lesions. The fusion image allows spatial localization of specific lesions for both single and repeat studies. Activity contours based on the SPECT antibody activity z-score can be used to determine the extent and invasiveness of the lesion. The z-scoring technique provides a direct analytic way of cross-study normalization. The fused images also allow the volumetric determination of lesion size.

The following case studies illustrate the application of fusion imaging. The first example is of a 50-year old female with ovarian carcinoma. The second example is of a 69-year old female with colorectal cancer. Both patients were evaluated using intravenously administered 111In CYT-103, (Cytogen Corp.) in a clinical trial protocol.

Planar CT and SPECT images were acquired. CT: Abdominal and pelvic scans were obtained at 1 cm intervals from the level of the diaphragm to perineum. Axial scans were obtained from above the diaphragm to below the pubic symphysis. SPECT: image data was acquired in digital format in a 64 x 64 matrix. The images were collected in a 360° orbit with 40 second sampling every 6°. Tomographic reconstruction was done in a standard filtered back projection.

Voxel dimensions were derived from field of view information provided on the images. Slice spacing was determined by the reconstruction. The pixel size for the CT images were derived by measuring 5 cm markers provided on the digital images. This measure was repeated five times on each slice to determine the average pixel value. Slice spacing for the raw CT images was 10 mm between planes.

Post-acquisition, the image data was processed to construct registered three-dimensional data sets on the MIRA© System (Loats Associates, Inc.). These patients were reviewed from an ongoing project in a post-hoc experiment and no external fiducials were included. Registration of the two data sets were accomplished by iterative comparison of internal landmarks.

Figure 6 presents a composite image showing the progression of analyzed images - CT, SPECT, fusion and z-score for three planes spanning the region with major lesions.

CT Evaluation. From the CT examination, a single large bi-lobed or two large complex masses were observed. Two large pelvic masses were also identified. There appear to be two masses, a superior mass and an inferior mass. Both masses appeared to contain multiple large homogeneous low attenuation masses which are probably fluid filled. The masses most likely represent ovarian neoplasm. The superior mass measured approximately 19 x 10 cm axially by 10 cm in the craniocaudal orientation. It is located anterior to the uterus, just cephalad to the urinary bladder. It displaced the bowel upward and laterally. The superior extent of the mass was just below the renal vessels. The inferior mass measured 10 x 10 cm by 8 cm craniocaudal. The inferior mass was located posterior to the uterus. It displaced the rectum posteriorly and the bladder in an anterior direction. The right and left kidneys, the liver and the spleen appeared to be radiographically normal. Discrete pelvic lymph nodes were identified 2.0, 1.8 and 1.5 cm in diameter.

Antibody Study. Approximately 5 mCi of 111In CYT-103 radiolabeled monoclonal were used in the antibody study. Both static planar images and tomographic cuts were obtained in the transaxial, coronal and sagittal planes in the thorax and in the abdomen from the upper abdomen to the pelvis. In the antibody study, two large relatively homogeneous regions were evident in the pelvis and lower abdomen. The first appears to be posterior to the low pelvis lying anterior to the sacrum. It is primarily central and a little to the left of the midline extending superior to the pelvic rim, Figure 6, Column 2.

The second is higher and more anteriorally located in the central pelvis. It exhibits a more irregular uptake pattern with three or four 3-4 cm zones of increased activity. The region extends from above the urinary bladder to above the level of the umbilicus. On the sagittal slices the mass extends to the anterior abdominal wall. The irregular pattern of uptake may indicate omental node involvement. There was also some uptake in the mediastinum and left supraclavicular region which may be due to adenopathy.

Fusion Images. Fusion images were constructed for each of the registered slices indicated in the scout images. In the fusion images, Figure 6, column 3, the background (in grey scale) is the CT and the foreground (in color scale) is the corresponding SPECT antibody image. Each of the slices has a known spatial location. Therefore, centroid locations for all hot spot/cold spot foci can be measured. Prominent activity foci are evident in the images. These foci correspond to anomalous objects in the CT images.

Based upon the total activity in the twelve images slices, standardized Zscore images were created and are shown in Figure 6, column 4.

Figure 7 presents a composite image showing the progression of analyzed images - CT, SPECT, fusion and z-score for three planes spanning the region with major lesions. Upon examination of CT slices, one mass was identified, a cecal mass with an estimated volume of 267 cm3. No definite retroperitoneal adenopathy was seen. There was no evidence of metastatic disease within the liver. No inguinal adenopathy was identified. Planar SPECT images showed distinct and broad uptake in the left upper quadrant. The tumor is probably large enough to have invaded the pericolic fat (Duke’s stage C). SPECT images, (Fig. 7, column 2) were developed for the region extending from L5 lumbar to 66 mm below L5 lumbar, at 6 mm spacing.

Fusion Images. Fusion images (Fig. 7, column 3) were constructed for each of the registered slices. Prominent activity is visible in the area of the cecum and, to a lesser extent, in the area of the descending colon. Based upon the total activity in the 12 image slices, standardized z-score images were created as shown in Figure 7, column 4.

With the increasing use of Mab as imaging agents, the need exists to quantify and locate uptake phenomena anatomically. The primary role of multimodal image registration and fusion is to increase the specificity of the SPECT study performed and to enhance the sensitivity of CT for the identification in colon and ovarian carcinoma. The lack of anatomical landmarks in SPECT studies has increased interest in combining SPECT with studies that give anatomic detail such as MRI or CT imagery. When used with registration and spatial frequency preserving resampling, multiple imaging modalities provide an additional useful and necessary anatomic component for clinical diagnosis. The accuracy of both CT and SPECT monoclonal-based imaging is significantly improved.

Precise stereotactic localization of antibody images in repeat imaging protocols is generally not possible without registration. Changes in a specific anatomic region or organs cannot be reliably identified as to location or activity. With registration and fusion, it becomes possible to follow changes in tumor state and extent over time. Additionally, recurrence and new tumor sets can be isolated. SPECT imagery that have been registered to the CT slices can be compared directly, without fearing that they represent a different area of the body. Two images can be subtracted to identify subtle changes in state or condition.

Using volume data sets, orthogonal (sagittal and coronal) CT images can be generated. Sufficient anatomic features are often available on the sagittal and coronal CT views to identify and localize small features with antibody uptake above background, like lymph nodes, that are difficult to isolate in the SPECT images alone. Precise localization in the x-y plane is accommodated by generating the appropriate transaxial matching CT slices. The matching SPECT images can now be displayed either singly or as a fused image.

The studies done to date indicate that Mab fusion imaging is successful at isolating early disease processes which are not observed in the CT images. In 1984, Delaloye et al. (8) reporting on the utility of SPECT with 123I-Labeled F(ab')2 from monoclonal anti-CEA antibodies indicated that the images were of particular value in patients with rising CEA but who exhibited negative radiologic images.

Kramer et. al. (9) performed a comprehensive study to correlate CT-SPECT antibody fusion images based on confirmation with surgical findings. They concluded that nonspecific antibody concentration in the blood pool or other normal anatomic structures is more easily recognized with CT-SPECT fusion due to the spatial overlay on structure. He also indicated that CT-SPECT fusion was also useful in identifying tumor masses that were undetected on the CT alone due to adjacent underfilled small bowels.

SPECT imaging without comparison to complementary (side-by-side) or overlayed and fused anatomical imaging can give rise to false-positives related to antibody uptake. False-positives in the SPECT images have been observed related to activity in the blood pool and in anomalous foci in the liver. For the most part, however, fusion imaging can resolve or alert both areas with false-positives and false-negatives in the SPECT image alone.

The advantages of image fusion over visual comparison of multimodalitiy are: (a) the fusion technique is useful to correct for variability in orientation, position and dimension; (b) it allows precise anatomic-physiologic correlation; and (c) it permits regional quantitation.

Tumor volumes defined on the basis of antibody concentration can be generated from the SPECT images. The distribution of antibody uptake in specific anatomic regions or non-specific background regions can also be evaluated. Three-dimensional volumes of tumors of sufficient size can be generated and viewed from different spatial perspectives.

Another important use of the registration-fusion technique is in follow-up studies where the primary purpose is to identify recurrence of the tumors or the spreading of metastatic disease. The major consequences of the registration process is that the sites of suspected activity can be revisited post-surgery to determine the extent of recovery or recurrence of disease.

The author would like to thank Dr. Robert Maquire and his staff at CYTOGEN for both the image data sets and for useful discussions and suggestions. The author would also like to thank Andrew Loats from Bio-Imaging Technologies, Inc., West Trenton, N.J. for information on the external fiducial. Also, the staff at LAI particularly Peter Loats, Sandra Loats, Teresa Rippeon, Maureen McIver and Donna Heartley. This work was supported by a presentation grant from CYTOGEN and in part by grants from National Institutes of Health, Multi Image Brain Quantitation SBIR NS-26548-02 and Multi Image Brain Quantitation SBIR NS-26548-01.

REFERENCES

1. Birnbaum BA, Noz ME, Chapnick J, Sanger JJ, Megibow AJ, Maguire GQ, Weinreb JC, Kaminer EM, Kramer EL. Hepatic Hemangiomas: (1991) Diagnosis with Fusion of MR, CT, and TC-99m-Labeled Red Blood Cell SPECT Images. Radiology 1991;181:469-474

2. Swayne LC, Kaplan IL. Image Fusion in Medicine: An Overview using the CT-SPECT Model. J Nucl Med 1989;17:31-35

3. Collier BD, Abdel-Nabi H, Doerr R, Harwood S, Holmes RA. (1990) Detection of colorectal cancer with 111In-labeled CYT-103 (abstr). Radiology 1990;181(suppl P):258

4. Kramer EL, Noz ME. (1991) CT/SPECT fusion for analysis of radiolabled antibodies: applications in gastrointestinal and lung cancinoma. Int J Rad Appl Instrum [B] 1991;18:27-42

5. Lamki LM, Patt YZ, Murray LJ, et al. (1990) Superiority of 111In- monoclonal antibody to CT and other conventional imaging studies in the diagnosis of occult recurrence of colorectal cancer (abstr). Radiology 1990;177(suppl P):258

6. Jaszczak RJ, Whitehead FR, Lim CB, Coleman RE. (1982) Lesion detection with single-photon emission computed tomography (SPECT) compared with conventional imaging. J Nucl Med 1982;23:97-102

7. Holcomb HH, Loats HL, Tamminga C, et al. (1988a) Registration of MRI-PET images, procedure and validation. J Nucl Med 1988;29:853

8. Delaloye B, Bischof-Delalaye A, Grob JB, et. al. SPECT with 123 I-Labeled F(ab’)2 Fragments from monoclonal Anti-CEA Antibodies in Colon Carcinoma (Abstr.) J Nucl Med 1984;25:17

9. Kramer EL, Noz ME, Sanger JJ, et al. CT-SPECT Fusion to Correlate Radiolabeled Monoclonal Antibody Uptake with Abdominal CT Findings. Radiology 1989;172:861-885.