[The value of tumor volumetry as opposed to bidimensional determination of tumor size during follow-up of hepatic metastases from colorectal carcinoma]

Size determination and response assessment of liver metastases with computed tomography--comparison of RECIST and volumetric algorithms

OBJECTIVE

To compare different three-dimensional volumetric algorithms (3D-algorithms) and RECIST for size measurement and response assessment in liver metastases from colorectal and pancreatic cancer.

METHODS

The volumes of a total of 102 liver metastases in 45 patients (pancreatic cancer, n=22; colon cancer, n=23) were estimated using three volumetric methods (seeded region growing method, slice-based segmentation, threshold-based segmentation) and the RECIST 1.1 method with volume calculation based on the largest axial diameter. Each measurement was performed three times by one observer. All four methods were applied to follow-up on 55 liver metastases in 29 patients undergoing systemic treatment (median follow-up, 3.5 months; range, 1-10 months). Analysis of variance (ANOVA) with post hoc tests was performed to analyze intraobserver variability and intermethod differences.

RESULTS

ANOVA showed significant higher volumes calculated according to the RECIST guideline compared to the other measurement methods (p<0.001) with relative differences ranging from 0.4% to 41.1%. Intraobserver variability was significantly higher (p<0.001) for RECIST and threshold based segmentation (3.6-32.8%) compared with slice segmentation (0.4-13.7%) and seeded region growing method (0.6-10.8%). In the follow-up study, the 3D-algorithms and the assessment following RECIST 1.1 showed a discordant classification of treatment response in 10-21% of the patients.

CONCLUSIONS

This study supports the use of volumetric measurement methods due to significant higher intraobserver reproducibility compared to RECIST. Substantial discrepancies in tumor response classification between RECIST and volumetric methods depending on applied thresholds confirm the requirement of a consensus concerning volumetric criteria for response assessment.

Flat Panel Detector-Based Volumetric Computed Tomography (fpVCT)

OBJECTIVES

Evaluation of a silicon-based flat panel volumetric computed tomography (fpVCT) and multislice CT in terms of volumetry of phantoms with different algorithms. Furthermore, to compare the different volumetric analysis methods themselves.

MATERIALS AND METHODS

Four phantoms of different materials have been scanned with fpVCT (GE prototype with circular gantry with 2 aSi/CsI flat panel detector) and a 64-slice spiral CT (MSCT: LightSpeed VCT). Three spherical phantoms of different materials and 1 phantom with an irregular shape were evaluated. True volumes were calculated in dependence from the diameter or by water displacement method. Imaging parameters (80 kVp, 100 mA) and the position of the phantoms were identical in both techniques. After reconstruction of the images different algorithms have been used 4 times for each phantom. These analysis methods have been performed: Region growing, threshold method, planimetry, 3-dimensional volumetry measurement by using the equation of an ellipsoid (ellipse) and an advanced lung analysis modus [single advanced lung analysis (ALA)]. The mean values and the standard deviations have been evaluated and compared with the true volumes.

RESULTS

In all phantoms fpVCT showed better results with lower deviations from the true values than in MSCT, especially for small volumes of the phantoms. However, the results of the ALA single method demonstrated no significant difference between the fpVCT and MSCT. The comparison of the different analysis methods revealed that 3-dimensional measurement with the ellipse method was the worst method for volume estimation, especially for the irregularly formed phantom.

CONCLUSION

fpVCT was superior to MSCT in the volumetry of small objects. The ellipse method has been shown to be the worst for volumetry with the highest relative deviations from the true volume value. The single ALA method shows the lowest standard deviation thereby revealing a reproducible volumetric method for small nodules. However, further future developments of volumetric analysis methods are necessary to use them accurately in daily routine. Due to the truly isotropic volume data set with high spatial resolution fpVCT is a powerful tool for the volumetry of small nodules.

Usability of semiautomatic segmentation algorithms for tumor volume determination

RATIONALE AND OBJECTIVES

Tumor volume is an important parameter for clinical decision making. At present, semiautomatic image segmentation is not a standard for tumor volumetry. The aim of this work was to investigate the usability of semiautomatic algorithms for tumor volume determination.

METHODS

Semiautomatic region- and volume-growing, isocontour, snakes, hierarchical, and histogram-based segmentation algorithms were tested for accuracy, contour variability, and time performance. The test were performed on a newly developed organic phantom for the simulation of a human liver and liver metastases. The real tumor volumes were measured by water displacement. These measured volumes were used as the gold standard for determining the accuracy of the algorithms.

RESULTS

Variability of the segmented volumes ranging from 3.9 +/- 3.2% (isocontour algorithm) to 11.5 +/- 13.9% (hierarchical segmentation) was observed. The segmentation time per slice varied between 32 (volume-growing) and 72 seconds (snakes) on an IBM/RS6000 workstation.

CONCLUSIONS

Only the region-growing and isocontour algorithms have the potential to be used for tumor volumetry. However, further improvements of these algorithms are necessary before they can be placed into clinical use.

Factors influencing the accuracy of volume measurements in spiral CT: a phantom study

PURPOSE

The purpose of this study was to investigate the influence of different parameters (object size, shape, orientation, contrast, observer, and window setting) on the accuracy of volume measurements in spiral CT.

METHOD

The phantom study consisted of two parts. First, well-circumscribed ellipsoid objects were scanned with conventional and spiral CT. Volumes were determined by two observers using manual contour delineation and summation-of-areas. The influence of object size, CT technique, and observer was assessed. Second, irregularly shaped gingerroots were scanned and the effect of observer, contrast, and orientation was investigated. Also assessed in both parts of the study was the effect of window setting.

RESULTS

Spiral CT offered higher accuracy and reproducibility than conventional CT did. The accuracy of volume measurements was mainly determined by the window setting used for image display. Using an "optimized" window center value halfway between object density and background density resulted in overestimation of true object volumes if the number of slices through the object (number of samples) was low. Other factors (contrast, observer-related errors in contour delineation) were less important.

CONCLUSIONS

Volume measurements of small objects, obtained with spiral CT and using manual contour delineation, may be grossly erroneous. Major causes of inaccuracy are an inappropriate window center selection and undersampling.