Image Fusion and Patient Specific Dosimetry for Tumor Response and Toxicity Evaluation in Patients with Glioma Undergoing Radionuclide Therapy

B. Wessels¹, R. Jensen², S. Patel³, A. Nelson¹, L. Bouchet⁴, J. Shan⁵ and T. Chew⁵

¹University Hospitals of Cleveland, Case Western Reserve University, Cleveland, Ohio, ²Department of Neurosurgery, University of Utah, Salt Lake City, Utah, ³Department of Neurological Surgery,

Medical University of South Carolina, Charleston, SC,⁴Department of Neurosurgery, University of Florida, Gainesville, FL, ⁵Peregrine Pharmaceuticals, Inc., Tustin, CA

ABSTRACT

A multicenter phase II trial for the treatment of patients with recurrent post-resection glioblastoma is being conducted with a new radionuclide therapy agent (Cotara^TM) labeled with I-131. The active targeting agent in Cotara has been shown to be reactive to DNA/histone H1 which is associated with tumor necrosis. Typically 20 to 100 mCi of the drug is administered intratumorally via continuous pump infusion over a 24 to 48 h time-period. Catheters are placed in previously identified lesions under MR stereotatic localization. The agent distributes in the tumor and its immediate periphery and clears with a mean T_{½ eff} of 45.5 h. The spatial and temporal distribution of the activity depends on a host of physical factors including; the number and placement of the catheters, flow rate, intratumoral pressure, local tumor and normal tissue architecture and total volume infused.

Patient-specific, 3-D dosimetry has been carried out to develop surrogate markers for evaluating tumor control and potential toxicity. The correlation between dosimetry-based measures and standard measures of tumor response and overall patient survival is being determined. In a retrospective analysis, fusion of SPECT and MR images was used to determine whether a better drug distribution around a tumor correlates to an improved absorbed dose pattern and consequently a superior tumor response in the irradiated region on a voxel by voxel basis. Localization of three external fiducial markers was used to co-register MRI/SPECT images. SPECT images were segmented according to standard methods (isoactivity borders set at 45 % of maximum pixel values for a high contrast tumors > 2 cm ). Three-dimensional point kernel calculations were used to compute the absorbed dose based on the measured voxel cumulated activity values. Activity and dose volume histograms were generated for each patient and compared to volume of the Gd-enhanced tumor volume. Separate figures of merit were derived by dividing the activity and dose volume by the MRI volume. Three patients receiving labeled antibody have been analyzed relative to this strategy and resulted in substantial varying figures of merit for activity volume or the prescribed isodose line covering the tumor volume (46 - 88 %). Preliminary results suggest a correlation between underdosed regions and sites of subsequent tumor regrowth. Further optimization of multi-catheter placement and drug delivery may also improve treatment outcome based on individual patient parameters.

INTRODUCTION & BACKGROUND

¹³¹I-ch TNT-1/B is a biotinylated chimeric intact monoclonal antibody recognizing a complex of double-stranded DNA and histone H1. In the phase 1 dose escalation study, doses up to 1.56 mCi/cc GETV were delivered with no significant local or system toxicity. This phase 2 study was designed to confirm the dose selected from the phase 1 study with the use of two interstitial catheters. Also, patients were eligible for a second treatment if they met all original entry criteria and did not have progressive disease at the 8 week evaluation exam.

The primary aim was to determine whether the drug was being delivered to the defined target volume and that target volume was irradiated to a specified absorbed dose level. Consequently, the amount of normal tissue irradiated and the tumor irradiation coverage, as determined using dose volume histogram (DVH), were equally important toxicity endpoints for the analysis. In order to quantitatively assess this information, image fusion sets of combined baseline MR volumes with overlying 14-day post treatment SPECT scans of the infused activity were obtained. Two quantitative metrics were defined to determine tumor coverage and normal tissue at risk of irradiation. These quantities were defined according to concepts outlined in ICRU 50 (1993) and ICRU 62 (1999) for radiation therapy treatment planning:

Metric 1 - Tumor coverage (%) was defined as:

Metric 2 - Normal tissue coverage (FOM) is given by:

FOM represents the "Figure of Merit" for normal tissue irradiation. Typical values of the FOM range from 0.0 to 1.0. Value close to zero means high conformality of the isoactivity distribution to the target (tumor) and little or no irradiation of normal tissues. Value near 0.5 means that a normal tissue volume equals to that of the tumor has been irradiated. Value from 0.75 to 1.00 means an excessive amount of normal tissue has been irradiated (at least 3 times the tumor volume). These metrics were also found to be similar to those proposed by Paddick (2000) in scoring conformity indicies of radiosurgical treatment plans.

Normal Tissue Volume covered by activity is defined as that volume which is circumscribed (segmented) by pixel values of 40 % of the maximum gray scale value for the highlighted tumor as determined in the SPECT scan. This definition is predicated on the phantom work performed by Mortelmans et al. (1986) and Erdi et al. (1996) for SPECT imaging. These investigators have shown that for uniform activity distribution in objects greater than 2 cm in diameter and contrast ratios in excess of 5 (object/background activity), the edge of these objects corresponded to a threshold setting of 40 - 45 % of the maximum counts/pixel for those objects. These assumptions are satisfied by the tumors under examination here except for the assumption of uniformity of activity throughout the target tissue. Hence, the volumes of activity determined here are limited in accuracy by the relative non-uniformity of the imaged tumor and should be used as relative values for comparison purposes.

METHODS

Image fusion program

MR and SPECT images for these 6 patients receiving Cotara^TM treatment were received through the core laboratory’s FTP site from participating institutions. MR-SPECT image fusion was performed using THE MIM^TM imaging software. The MIM^TM software is an FDA 510 k approved product made available through co-author AN. In order to maintain original scan volume accuracy, translation and rotation are the only two operations used during this procedure. Image magnification is set directly from the incoming DICOM file and was checked with a scale bar as provided on the image. A simulated true color overlay displays the merged image volumes through color summation of the MR and SPECT image volumes. In addition, both tumor region contour and isoactivity distribution are displayed on the fused image. Controls for image alignment, contrast control and slice selection are available based on cursor location on the MIM^TM image platform. External fiducial markers (Figure 1) are also available for image alignment control. For setting the positions of the markers, Markers 1 and 3 are set 2 cm anterior to the auditory canal on a theoretical reference line extending from the outer canthus of either eye to the auditory canal. Marker 2 is located at the bridge of the nose.

A similar method was used to overlay the isodose curve with the fused MR/SPECT image. A commercially available software package (Micrografx Picture Publisher 8) was used to display partially transparent images of all three imaging modalities. The isodose curve was rescaled with respect to the MR/SPECT image. The transparency was adjusted to display isodose curve, isoacitvity distribution and Gd-enhanced tumor region.

Volume determination for SPECT and MR imaging modalities

In this study, isoactivity volume, tumor volume and overlap volume of each patient were obtained by using Volume Estimator with a complete series of fused images mentioned above.

Volume Estimator (VE) is a volume-calculating program which may be used to identify regions of interest (ROI) in 2-D images and accumulate all these regions to obtain a total volume. In order to sum the volume correctly, the area of each image was multiplied by the corresponding slice thickness.

Tumor coverage (%) and normal tissue coverage (FOM) were assessed from isoactivity volume, tumor volume and overlap volume of the whole ROI according to the definitions above.

Absorbed dose computation

Calculations of average dose to the tumor (Table 1) are based on the mean biodistribution results from the Phase I Cotara^TM clinical trial (Patel, 1999). These include an average tumor uptake maximum of 34.2 % and T 1/2 eff of 45.1 hr. Volume determinations for the isoactivity distribution from SPECT were performed using the previously described Volume Estimator program. A mean absorbed dose to the tumor was calculated according to the MIRD formulation (Snyder 1975, 1978) for NP (non-penetrating) radiation only using the equilibrium dose constant for I-131 of 0.409 g*rad/uCi*hr with total absorption of the beta radiation localized to the tumor and immediate surrounding tissues.

To take into account non-uniform activity distribution in the tumor, a 3D dose calculation algorithm was used. This algorithm was based on point and voxel kernels (voxel S values) as outlined by Sgouros (1993) and Bolch (1999). The total activity in the tumor was determined from the administered activity and the pharmacokinetic parameters provided by the phase I dosimetry measurements. This total cumulated activity was then apportioned to the tumor using the spatial distribution defined by the SPECT image intensity values. This was accomplished by dividing each voxel value by the sum of all voxel values within the tumor and then multiplying by the total cumulated activity for the tumor. The image intensity or gray scale in each voxel was assumed to be linearly correlated with cumulated activity. This assumption was further substantiated by the radial crossplot measurements taken for each of the six patients and shown in Figure 4. The calculated cumulated activity per voxel was then convoluted with the point and voxel kernels to generate a 3D-dose volume at the resolution of the SPECT scan volume. 2D isodose displays are shown as overlays to the fused SPECT/MR central slice images (Figures 5 a-d).

Fiducial Marker Alignments:

A total of 3 radioactive markers were placed on the skin to assist in the 3-D alignment process for the MR and SPECT images using the MIM^TM program. Marker 1 & 3 were typically placed 2 cm anterior to the auditory canals on a line from the outer canthus of either eye to the auditory canals. Marker 2 was placed at the bridge of nose. The image intensity has been maximized so that these 100 mCi fiducial sources markers are visible

Figure 1 - External Fiducial Markers

RESULTS

Image Fusion:

Image fusion between the MR and infused activity SPECT scans for the six patients receiving Cotara ^TM are shown in Figure 2.

Patient Utah-KG MRI

Figure 2a - Patient Utah-KG MRI

Figure 2b – Patient Utah-KG Fusion

Patient MUSC-KMG MRI

Figure 2c – Patient MUSC-KMG MRI

Figure 2d – Patient MUSC-KMG Fusion

Patient MUSC-JGL MRI

Figure 2e – Patient MUSC-JGL MRI

Figure 2f - Patient MUSC-JGL Fusion

Patient MUSC-MEP MRI

Figure 2g - Patient MUSC-MEP MRI

Figure 2h – Patient MUSC-MEP Fusion

Patient Utah-JHA MRI

Figure 2I – Patient Utah-JHA MRI

Figure 2J – Patient Utah-JHA Fusion

Patient Utah-DLG MRI

Figure 2K – Patient Utah-DLG MRI

Figure 2L- Patient Utah-DLG Fusion

Collage showing MR, SPECT alone and fusion pair images in the axial, sagittal and coronal views

Figure 2M – Patient MUSC-KMG

Figure 2N – Patient MUSC-KMG Isodose Overlay

Figure 3

Volume Determination: Isoactivity distribution volume, tumor volume and overlap region between these two areas on the central slice of patient Utah-DG are illustrated in Figure 3 when using the volume estimator ROI computation tool. The volume estimator was also used as a summing tool for all serial images extending through the tumor and isoactivity distribution by multiplying with the appropriate slice thickness to obtain a volume for all ROI’s.

Figure 3a – Red ROI denotes isoactivity distribution area for patient Utah-DG on the central slice

Figure 3b – Red ROI denotes tumor area for patient Utah-DG on the central slice

Figure 3c – Red ROI denotes overlap area between the tumor area and the isoactivity distribution area for patient Utah-DG on the central slice

Figure 4:

Radial Cross Plots: These graphs show the fall-off of intensity (gray scale) versus the distance out from the center of the infused activity distribution for 2 patients. These data show a first order approximation of linear decrease of activity vs. radial distance near the edge of the activity profile.

Figure 4a - Radial Cross Plot for patient MUSC-MEP

Figure 4c - Radial Cross Plot for patient Utah-JHA

Figure 5

Absorbed Dose Overlay: Using 3-D methods of calculating dose for internal emitter distribution according to methods by Bolch (1999) and Sgouros (1993), semi – transparent isodose overlays were fused onto the activity distribution – MRI fusion to estimate the absorbed dose distribution to both the tumor and the surrounding normal tissue.

Figure 5a – Patient MUSC-JGL Isodose Overlay

Figure 5b – Patient MUSC-MEP Isodose Overlay

Figure 5c – Patient Utah-DLG Isodose Overlay

Figure 5d – Patient Utah-JHA Isodose Overlay

Table 1 - Summary of quantitative assessment for all 6 patients

Quantitative Assessment of Tumor and Normal Tissue Coverage

with Cotara Infusion

DISCUSSION & CONCLUSIONS

We have used SPECT images of localized Cotara^TM to glioma registered with underlying MR anatomy to assess several surrogate treatment outcome parameters. These parameters include 1) percent of Gd enhanced tumor volume covered by the activity distribution, 2) a figure of merit which assesses the amount of normal tissue at risk of irradiation and 3) the amount and distribution of absorbed dose in the region of the tumor. Guided by standard radiation therapy treatment planning criteria (ICRU 50), we have developed a set of unique outcome parameters directly applicable to radionuclide therapy.

For these six patients, the data indicates that some basic conclusions or recommendations for optimum treatment can be made. First, the localization of the drug in a volume larger than was anticipated using the baseline treatment MR can substantially decrease the mean absorbed dose to the target volume. For these patients, the resultant dose localized in the expanded volume decreased by more than a factor of four compared to the MR volume dependent dosing metric. Secondly, various degrees of physical geographic miss of the activity distribution overlaying the tumor volume were apparent and quantified in this study. Guided by external radiation therapy experience, the more dose on the target volume with the minimum of normal tissue irradiated is the most desirable method of delivering effective therapy. For this study, the highest survival post treatment (Pat. MUSC- JGL, Utah - KG, Utah -DLG, MUSC - KMG) is consistent with higher percent of tumor covered by the activity. The relationship between absorbed dose dependence and response for this data set is not immediately discernable. Thirdly, some dosing of the region outside the GTV (Gross Tumor Volume) may not be deleterious to the clinical outcome of the study considering the regrowth mechanism for Stage 4 gliomas. Properly, a CTV (Clinical Treatment Volume) should be entered into the dosing algorithm. This may include treatment margin several cm beyond the baseline Gd enhanced MR volume as well as a correction estimate of mean geographic misalignment of the isoactivity distribution with the tumor volume. In addition, for this small cohort of patients, the volume of drug distribution appears to be more dependent on the volume of infusion rather than the total time of activity administration (24 versus 48 hours).

Society for Neuro-Oncology, November 2001