| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Hepatobiliary Imaging |
1 Department of Radiology, Division of Abdominal Imaging and Intervention,
Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115.
2 Department of Radiology, University of Massachusetts, 5 Lake Ave., Worcester,
MA 01604.
Received September 16, 2003; accepted after revision March 22, 2004.
Address correspondence to S. G. Silverman
(sgsilverman{at}partners.org).
OBJECTIVE. We report our initial investigation of the use of a 3D method for assessing percutaneous tumor ablations. We hypothesized that these 3D techniques could be used to assess the technical success of ablations and that 3D metrics would be predictive of treatment response.
CONCLUSION. Three-dimensional assessment of percutaneous tumor ablations provides a quantitative evaluation of the technical success of the procedure. Three-dimensional computer-based techniques can both quantify coverage of a tumor and create a virtual ablation margin for percutaneous procedures, akin to a surgical margin. Although results are preliminary, 3D metrics were useful in predicting treatment response.
Percutaneous imaging-guided thermal ablation techniques, principally using radiofrequency, have recently become available, offer treatment of liver metastases with low morbidity rates, and are therefore well-suited for patients who are not surgical candidates [1]. Cryotherapy is a thermal ablation technique that has been used extensively in open surgical settings [2] and more recently was applied percutaneously to treat liver tumors [3, 4]. For cryotherapy and, in fact, for all percutaneous thermal ablation techniques, a method is needed to assess how well a tumor has been covered by the treatment either during or soon after the procedure. The term "technical success" has been described recently and is used to address whether the tumor was treated according to protocol and covered completely [5]. Complete coverage would be considered a technically successful procedure. Currently, the most common method of assessing whether an ablation procedure was technically successful is based on visual inspection of 2D contrast-enhanced CT scans or MR images at the end of the procedure or within 48 hr. If no residual tumor is detected, patients are then followed up with imaging typically at 3- to 6-month intervals. Patients are retreated if and when residual tumor is identified.
Three-dimensional measurements, in contrast, allow the determination of tissue volumes and are likely to represent tumor burden and coverage more fully than measurements of maximum diameters obtained from 2D data. A technique that provides the periprocedural 3D assessment of the tumor volume, the additional targeted ab lation margin volume, and the entire volume of the treatment effect would be of benefit in determining whether an ablation was technically successful. We describe herein such a method and its application to patients with liver metastases treated with MRI-guided cryotherapy. Furthermore, we determined whether 3D assessment metrics were predictive of treatment response.
Materials and Methods
Patients
The protocol for the selection of patients to undergo MRI-guided
percutaneous cryotherapy of liver tumors was approved by the institutional
review board of our hospital. This clinical trial was limited to patients with
liver tumors less than or equal to 5 cm in diameter who either refused surgery
or whose malignancy was considered inoperable for technical or medical
reasons. Image data from nine patient procedures were analyzed for this 3D
assessment. Preliminary results of feasibility and safety were reported in
seven patients [3]. The study
comprised five women and four men, 50–81 years old. All patients gave
informed consent to enter the clinical trial for treatment; additional
institutional review board approval was obtained to review image data for this
retrospective image analysis. Nine biopsy-proven liver metastases (mean
maximal diameter, 3.1 cm; range, 1.5–5.0 cm) were analyzed. These
included metastases from adenocarcinoma of the colon (n = 5),
esophagus (n = 1), and unknown origin (n = 1);
gastrointestinal stromal sarcoma arising from the stomach (n = 1);
and non–small cell lung cancer (n = 1).
Imaging and Procedures
Percutaneous MRI-guided cryotherapy of liver tumors was performed using a
0.5-T open-configuration MRI system (Signa SP/i, GE Healthcare) and an argon
gas–based MRI-compatible cryotherapy delivery system (CryoHit, Galil
Medical). Cryotherapy involved the placement of 2.2- to 2.4-mm diameter
needlelike cryoprobes into each liver tumor using MRI guidance, followed by
freezing of the probes to create a zone of frozen tissue surrounding the
tumor. Postprocedural MR images were acquired in all patients using a 1.5-T
system (EchoSpeed, GE Healthcare) 24–48 hr after cryotherapy. MRI
consisted of transverse T1-weighted spin-echo imaging (TR range/TE range,
300–600/4.2–14; section thickness, 4–5 mm; field of view,
34–40 cm), transverse T2-weighted fast spin-echo imaging
(2,200–5,100/100–106; echo-train length, 12–19; section
thickness, 4 mm; field of view, 30–40 cm), and transverse fast
multiplanar spoiled gradient-echo imaging (TR range/TE, 285–310/1.6;
flip angle, 75°; section thickness, 5–6 mm; interslice gap, 1 mm;
field of view, 34–40 cm; with fat suppression) performed before and
after (at 30, 60, 90 sec and 5 min) the IV injection of 20 mL of gadopentetate
dimeglumine (Magnevist, Berlex Laboratories).
Segmentation
Determining tumor and cryonecrosis volumes.—Postprocedural
MR images were transferred to a UNIX workstation (Sparc, Sun Microsystems) and
reviewed to delineate the liver, tumor, and the zone of cryonecrosis.
Cryonecrosis was identified by a lack of enhancement on contrast-enhanced
T1-weighted MR images. In two patients, hyperintensity on T2-weighted images
was also used to help define the regions of interest. Ablated tumors may be
visible soon after cryotherapy
[3]. The boundaries of the
tumor could be distinguished on postprocedural MRI (Figs.
1A,
1B, and
1C).
|
|
|
Using software with an open license that provides navigation, segmentation, and measurements within multiple-slice data sets [6], we segmented 2D images by identifying the pixels within each MR image and manually contouring the regions of interest (liver, tumor, and cryonecrosis). All segmentations were conducted by a trained medical student and approved by a staff radiologist who performed the ablation. The same software was used to generate 3D models from the segmented regions (liver, tumor, and cryonecrosis) (Figs. 1A, 1B, and 1C). A voxel-based measurement was used to calculate volumes.
Determining target volume.—The "target volume" was defined as the tumor plus an ablation margin. The target volume was mathematically computed using a dilation function applied to the tumor [7]. The ablation margin consisted of a cuff of normal liver extending 1.0 cm in all directions perpendicular to the tumor surface (Figs. 2A, 2B, 2C, and 2D). A voxel-based volume measurement was used to calculate volumes.
|
|
|
|
Assessment of Technical Success
Three metrics of technical success were calculated for each patient:
percentage of tumor coverage, percentage of target volume coverage, and Dice
similarity coefficient (DSC)
[8]. Percentage of tumor
coverage is equal to (a / b) x 100, where a is the
number of voxels common to both tumor and cryonecrosis volumes and b
is the number of voxels belonging only to the tumor volume. Percentage of
target volume coverage is equal to (a / b) x 100, where
a is the number of voxels common to both target and cryonecrosis
volumes and b is the number of voxels belonging only to the target
volume. The DSC is a measure of agreement between two data sets
[8,
9]. Its use for the comparison
of overlap between voxel-based data sets, including the description of
segmentation intersection, has been previously described
[10]. It is calculated as
 |
|
Three-Dimensional Assessment Metrics as Predictors of Treatment Response
Treatment response was determined for each patient on the basis of
follow-up MR images (range, 2–27 months; mean, 12.7 months). These
examinations were compared with preprocedural MR images by two staff abdominal
radiologists blinded to the results of the 3D metrics. The nine treated liver
metastases were divided into two groups: group 1 included those tumors that
exhibited a complete response to treatment (defined as no evidence of tumor at
follow-up) and group 2 included those that had a partial response (defined as
tumors that were not completely ablated but were smaller, stable, or showed
growth in only portions at follow-up). With these data, mean values of
percentage of tumor coverage, percentage of target volume coverage, and DSC
were tested as predictors of treatment response. For all three predictors, a
standard one-tailed Student's t test was used to test their mean
values by dichotomized outcome (complete response vs partial response).
Results
The maximum tumor diameter, tumor volume, target volume, cryonecrosis volume, percentage of tumor coverage, percentage of target volume coverage, DSC, and treatment response for each of the nine patients are presented in Table 1. The mean liver tumor volume was 22.9 cm2 (range, 3.5–56.9 cm2), and the mean target volume was 72.1 cm2 (range, 28.3–129.6 cm2). Percentage of tumor coverage values ranged from 29.8% to 100%, with a mean of 78.8%. Percentage of target volume coverage values ranged from 22.7% to 99.8%, with a mean of 53.7%. DSC values ranged from 0.27 to 0.75, with a mean of 0.52.
|
The mean percentage of tumor coverage for the complete response group (n = 3) and the partial response group (n = 6) was 100% and 68.2%, respectively. These means were significantly different (p = 0.04). The mean percentages of target volume coverage for the complete response group and the partial response group were 86.8% and 37.2%, respectively. These means were also significantly different (p = 0.003). The mean DSC values for the complete response and partial response groups were 0.5 and 0.525, respectively, with no significant difference (p = 0.29).
Of six patients in whom a partial response was seen, three (patients 3, 6, and 7) had tumors that decreased in size at follow-up. The other three (patients 1, 2, and 4) had tumors that were stable or showed some tumor growth at follow-up (Fig. 4). Of the three patients who achieved a complete response, two (patients 8 and 9) achieved complete (100%) tumor coverage and near-complete (93.2% and 99.8%) target volume coverage (Figs. 5 and 6); the other (patient 5) achieved complete tumor coverage but only moderate (67.3%) coverage of the target volume.
|
|
|
Discussion
Our study, using data derived from liver tumors treated with percutaneous MRI-guided cryotherapy, shows how 3D segmentation of postprocedural MR images permits the generation of 3D models that quantify tumor and treatment volumes. This technique provides precise determination of the tumor volume, the percentage of tumor covered, and the volume of healthy tissue ablated by the treatment. In addition, it permits the designation of an expanded target region incorporating a digitally created ablation margin. Three-dimensional modeling therefore allows the calculation of additional assessment metrics that is not possible with the current method of visually inspecting 2D images.
A technically successful liver tumor ablation should incorporate both complete coverage of the tumor with an adequate margin and maximal preservation of healthy surrounding tissue. Each of the 3D assessment metrics used in this study (percentage of tumor coverage, percentage of target volume coverage, and DSC) describes different aspects of these characteristics of technical success. Percentage of tumor coverage is an intuitive and important component of assessment of the technical success of ablation procedures; however, it does not address the issues of ablation of excessive normal tissue and coverage of a 1-cm ablation margin. Percentage of target volume coverage complements percentage of tumor coverage by allowing the ablation margin to be included in the analysis. The DSC provides additional information about the agreement of target and ablation volumes. Whereas percentage of tumor coverage varies only with the intersection of the tumor and ablation volumes, the DSC reflects both the extent of intersection and the extent of nonintersection of these volumes.
Whereas both percentage of tumor coverage and percentage of target coverage were each statistically significant predictors of treatment response, percentage of target volume coverage achieved a greater degree of statistical significance (p = 0.003 vs 0.04). These data suggest the importance of assessing coverage of the ablation margin in addition to the tumor. This concept resonates with surgical principles: liver resections are targeted not only to accomplish resection of the tumor mass itself but also to create an appropriate margin of normal-appearing tissue, or surgical margin—a practice that has been shown to aid in prevention of residual disease [11, 12]. Although the data are preliminary, the determination of percentage of tumor coverage and percentage of target coverage 24–48 hr after an ablation procedure could be used to select patients who should be retreated immediately, rather than waiting for recurrences to become evident at follow-up.
As our results show, the DSC would not necessarily predict treatment response because a low DSC could describe both a poorly covered tumor (Fig. 4) and a well-covered tumor with large normal-tissue kill (Fig. 6). The latter scenario might lead to complications either as a result of ablating an adjacent important structure or because of excessive normal-tissue damage. Correlation of the DSC with procedural safety was not possible in this study because of the lack of complications. However, it is presented here as an important component of the evaluation of technical success of percutaneous ablation procedures. High values of the DSC always represent appropriate coverage with an appropriately sized zone of cryonecrosis and are therefore desirable. A DSC greater than 0.7 is generally considered to represent excellent agreement of volumes [10]. Achieving a high DSC is less important if the tumor is not located adjacent to critical structures. For example, for tumors located deep within the liver (Fig. 6) or surrounded by a similar safe zone of noncritical structures, overcoverage (high percentage of coverage and low DSC) may be satisfactory. However, for tumors in proximity to critical structures (e.g., gallbladder), a high DSC is an important goal.
Our study had limitations. The sample size was small because we chose to simply show a novel technique of assessing ablations. As this technique matures, larger prospective studies will be conducted to validate its utility. Also, we applied these methods to patients undergoing MRI-guided cryotherapy. Although the concepts are applicable to other ablative agents and imaging-guidance techniques, further study is needed. With further maturation, this technique may provide valuable information that could be beneficial for the practice of tumor ablation and, in the future, applied intraprocedurally.
Acknowledgments
We thank Felix Dahm for technical assistance and Donna L. Vega for secretarial assistance.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  I. Bricault, R. Kikinis, P. R. Morrison, E. vanSonnenberg, K. Tuncali, and S. G. Silverman Liver Metastases: 3D Shape-based Analysis of CT Scans for Detection of Local Recurrence after Radiofrequency Ablation Radiology, October 1, 2006; 241(1): 243 - 250. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
