British Journal of Radiology (2005) 78, 338-342
© 2005 British Institute of Radiology
doi: 10.1259/bjr/20260420
Comparison between 1 T MRI and non-MRI based volumetry in inoculated tumours in mice
B Cornelissen, PhD
1
V Kersemans, MSc
1
L Jans, MD
2
L Staelens, PhD
1
R Oltenfreiter, Pharm
1
T Thonissen, MSc
1
E Achten, MD, PhD
2 and
G Slegers, PhD
1
1 Laboratory of Radiopharmacy, Ghent University, Harelbekestraat 72, B-9000 Gent and 2 Department of Radiology and Magnetic Resonance Imaging, Ghent University Hospital, De Pintelaan 185, B-9000 Gent, Belgium
 |
Abstract
|
|---|
Tumour volume is an important therapeutic endpoint for mouse tumour models in the evaluation of new chemotherapeutic drugs and in pre-clinical evaluation of new radioimmunotherapy pharmaceuticals. In this study, two 1 T MRI-based methods both using T1–T2 hybrid weighting, a manual method (determination of the area per slice) and a semi-automated method (using thresholding), are compared with two classical methods, the abovementioned calliper method and volumetry by water displacement after dissection of the tumour. Interoperator and intraoperator differences for both MRI-based methods were good (no differences p<0.05 using a repeated measures analysis of variance (ANOVA) test). Correlation between the different methods was excellent. No significant differences were obtained (p<0.05), except for the semi-automated method, because it automatically excludes necrotic regions from the tumour. Therefore, we conclude that both manual and semi-automated tumour volumetry in subcutaneous tumour bearing athymic mice by low-field MRI are accurate and reliable methods. The semi-automated method is especially useful for larger tumour volumes, since it accounts for necrotic areas within the tumour.
|
Introduction
|
|---|
Tumour volume is an important therapy endpoint in cancers treated with chemotherapy and radiotherapy. Tumour volume measurements are widely used in the staging and evaluation of therapy of various types of tumours in humans, such as brain, prostate, cervical and many other tumours. Regarding MR-based volumetry, manual measurements, semi-automated systems (based on region growing) and automated systems (based on segmentation) have been reported [1–7]. In this study, we report the comparison of MR-based volumetry methods with conventional methods (sliding calliper measurement) in tumour inoculated mice.
In an analogy with the human situation, tumour volume is an important therapeutic endpoint for mouse tumour models in the evaluation of new chemotherapeutic drugs [8] and in pre-clinical evaluation of new radioimmunotherapy pharmaceuticals [9, 10]. The most commonly used method for determination of tumour volume of subcutaneously inoculated tumours in mice is calliper measurement, because of its simplicity. However, due to the approximation of the (ellipsoid) shape of the tumour, this method is prone to error [11]. Volume measurement with MRI-based methods is needed to avoid the propagation of these errors when using the tumour volume for normalization of other parameters, e.g. uptake of a radiopharmaceutical compound. Waterton et al [12] performed a comparison between ultrasound, MR and calliper measurements in mice bearing a subcutaneous tumour. However, they did not report comparisons with a gold standard. In this study, two MRI-based methods, a manual method (determination of the area per slice) and a semi-automated method (using thresholding), are compared with two classical methods, the abovementioned calliper method and volumetry by water displacement after dissection of the tumour (gold standard).
|
Materials and methods
|
|---|
Tumour inoculation
Male Swiss athymic mice (20–25 g) were obtained from Bioservices (Schaijk, The Netherlands). All animals were treated according to the regulations of the Belgian law and animal experiments were approved by the local ethical committee. Athymic mice were inoculated subcutaneously in the right flank with 5.106 cells of HT29, a colon adenocarcinoma, as previously described [8]. Tumours were allowed to grow for 2 to 6 weeks. The in vivo tumour genesis was 100%. If tumour size approximately reached the desired size (from 0.196 cm3 to 2.352 cm3, as determined by water volumetry), the mice were included in the MRI study. Mean tumour size was 1.048±0.687 cm3.
MR image acquisition
The animals were anaesthetised with an intraperitoneal injection of 100 µl of a medetomidin/ketamin – mixture (Domitor, Orion, Espoo, Finland; Ketalar, Parke-Davis, Zaventem, Belgium), 0.1 mg ml–1 and 5 mg ml–1 in water, respectively. After sedation, the animals were placed in a home-built mouse holder in which sterile conditions could be maintained. Two of these holders were fixed into a small surface flex coil.
MRI was performed on a clinical 1 T whole body system (Magnetom SP; Siemens, Erlangen, Germany). The position of the tumour was localized by coronal and sagittal scout images. Imaging was centred on this position.
Prior to the imaging of the series of mice described below, a comparison was made between three-dimensional (3D) dual echo steady state (DESS), T1 and T2 weighted images (axial 3D DESS, repetition time/echo time (TR/TE)=25.7 ms/9.0 ms, 128 x 256 matrix, slice thickness 1.0 mm, field of view (FOV) 124 mm x 124 mm x 42 mm, taking 21 min 05 s; axial T1 weighted spin echo, a TR/TE=500 ms/15.0 ms, 128 x 256 matrix, slice thickness 3.0 mm, taking 4 min 19 s; and axial T2 weighted spin echo, a TR/TE=3204 ms/119.0 ms, 90 x 256 matrix, slice thickness 3.0 mm, taking 10 min 18 s. We selected the 3D DESS sequence for further analyses. The DESS sequence (dual-echo in the steady state) acquires two gradient-echo signals in one repetition time, a fast imaging with steady state precession (FISP) and a reversed FISP (PSIF) signal. During image reconstruction, the strong T2 weighted PSIF signal is added to the FISP signal. This yields images with improved signal-to-noise (two raw data sets added) and has stronger T2 contrast all within a measurement time comparable with a FISP sequence [13]. The 3D DESS sequence not only yielded better signal-to-noise ratios but also significantly better delineation of the tumour, as compared with the T1 and T2 weighted sequences. Moreover, 3D DESS yielded better tumour to muscle and tumour to fatty tissue ratios. 3D techniques offer superior spatial resolution, and they do not suffer from interslice saturation and they do not yield interslice gaps (contiguous slices were acquired throughout the tumour without intersectional space). Additionally, they are much faster comparing the same image quality obtained by 2D sequences. For all these reasons, we decided in favour of the 3D DESS sequence.
Manual volume measurements
Manual volumetry was performed on a MagicView 1000 workstation (Siemens AG, Erlangen, Germany). The tumour was delineated in each slice. Summing the obtained areas and multiplication with the slice thickness resulted in the tumour volume. Three operators (BC, VK and RO) independently performed the manual volumetry three times on different days. The rationale to use non-radiologists to perform the repetitive measurements of the manual method was that tumour volume measurements are typically performed by MR-technologists, and not by the local radiologists. Interoperator and intraoperator reliability was assessed from these data.
Semi-automated region growing measurements
The semi-automatic measurement is based on region growing [3]. A seed point was manually selected in the most intensely coloured part of the tumour. All non-diagonal adjacent voxels that have signal intensities above a certain threshold were included in the volume. For every newly included voxel, this algorithm was repeated until no voxels were added. An advantage of this method is that non-enhanced (necrotic) areas are automatically detected and not included (Figure 1
). The measurements were performed twice on two different days by BC and TT, who have experience in radiology, but are not radiologists, for the abovementioned reasons. Interoperator and intraoperator reliability was assessed from these data.
View larger version (32K):
[in this window]
[in a new window]
|
Figure 1. Region growing tumour volumetry on 1 T MR images. (a) Axial slice through the tumour. (b) The same slice with a section of the "grown" tumour volume. As shown, non-enhanced parts of the tumour are not included in the volume.
|
|
Calliper measurements
Volumetry by calliper measurement is performed by measuring (a) the short and (b) the long axis of the tumour by calliper. The volume of the tumour is calculated from the estimation of an ellipsoid: V=0.5 x a2 x b [11].
Water displacement measurements
To measure the volume of the tumour by water displacement volumetry, the tumour was resected. Subcutaneously inoculated tumours, which do not have adjacent tissue infiltration, are easy to resect by an experienced technologist. Additionally, the lining of the tumour is very well defined by 3D DESS MRI. The tissues measured by MRI-based methods and water displacement are therefore the same. After dissection of the tumour, water displacement volumetry is performed immediately to prevent shrinking of the tumour due to dehydration. A scheme of the home-built system is shown in Figure 2. The system consists of three connected recipients, A, B (reading error 0.005 ml) and C (reading error 0.005 ml), filled with water. The water levels in B and C are marked on the recipients. The tumour is placed in recipient A, which is large, and thus not suited for exact volume readings. By placing the tumour in recipient A, the liquid level in pipette B and C also rises. B is used as calibration for the initial water level. Pipette C is now moved downwards until the water level in B has reached its original level. The difference in C corresponds with the volume of the tumour.
Statistics
In order to assess intraoperator differences and interoperator differences, a repeated measures analysis of variance (ANOVA) was performed on the data set of manual volume measurements after confirming that the data were normally distributed by means of the Shapiro-Wilks test. Furthermore, the data from the manual volume measurements were compared with those obtained from water displacement measurements. This was done by performing a paired samples t-test, comparing the mean of the three observations from the manual volume measurements and the values obtained from the water displacement measurements. Semi-automated and calliper measurements were also compared with the water displacement values by means of a paired samples t-test.
|
Results
|
|---|
Shapiro-Wilks analysis showed normally distributed results for all manual volume measurements. Repeated measurement ANOVA results for interoperator and intraoperator differences for the manual volumetry on MR images revealed no significant (p<0.05) interoperator differences between the measurements performed by any of the three operators on different days. Furthermore, no significant differences were found between the values obtained by the three operators.
A matrix regression analysis of all four methods showed good correlation between all four methods (r2>0.95, p<0.0001). Bland-Altman plots of all measurements vs gold standard are shown in Figure 3.
View larger version (7K):
[in this window]
[in a new window]
|
Figure 3. Bland-Altman plots for all methods compared with water volumetry. (a) Water vs calliper. (b) Water vs manual (region of interest (ROI)-drawing). (c) Water vs semi-automated (region growing).
|
|
Comparison between the calliper method and the water displacement method yielded no significant difference (p=0.7064). Correlation was good (r2=0.9583). Bland-Altman analysis showed a mean overestimation of the tumour volume of 0.056 cm3 (2.81%) compared with the gold standard. The Bland-Altman plot also shows a larger deviation for larger tumour volumes.
The manual method took approximately 5 min per mouse per operator. Linear regression analysis showed good correlation between operators and between repeated measurements of the same operator (r2>0.95; p<0.003). Comparison of the manual volume and water displacement methods also showed no significant difference between the gold standard and the values obtained by any of the three operators (p=0.1945 (RO); p=0.1281 (BC); p=0.1221 (VK)). Correlation was good (r2>0.9738). Bland-Altman analysis showed a mean underestimation of the tumour volume of 0.102 cm3 (5.13%) compared with the gold standard.
The region growing method took approximately 1 min per mouse per operator. However, comparison between the region growing method and the water displacement method did yield a significant difference (p=0.0155). Also here, correlation was good (r2=0.9907). Bland-Altman analysis showed a mean underestimation of the tumour volume of 0.076 cm3 (3.77%) compared with the gold standard. Interoperator and intraoperator reliability for the region growing method was good: Shapiro-Wilks coefficients for all repeated measurement sets were >0.95. No significant (p<0.05) interoperator differences were found between the measurements performed by any of the two operators on different days. Furthermore, no significant differences were found between the values obtained by the two operators.
|
Discussion
|
|---|
In this paper, we evaluated the use of four different techniques for measurement of the volume of subcutaneous tumours in mice. Calliper measurement, and two MRI-based methods (a manual method and a region growing method) are compared with water displacement.
Calliper measurement is mostly used when assessing tumour volume in therapy evaluation studies on large numbers of mice. However, the accuracy of the method is doubtful, especially when larger tumour volumes are included. This is also shown by our results. Therefore, we wanted to examine the use of classical non-dedicated 1 T MRI for tumour volume measurements of subcutaneous tumours.
The large deviations from the gold standard as measured with the calliper method imply that calliper measurement is not suited for pre-clinical studies using large tumours. One possible way to overcome this issue is to use many mice and thus levelling out the mistakes made by calliper measurement. By using MRI-based methods the number of animals can be reduced drastically.
Despite the fact that dedicated small animal strong field MRI systems are available, we used a simple clinical 1 T machine, which is routinely used in clinical practice. The main advantage of this approach is the more abundant availability of clinical 1 T systems, and their reduced cost compared with dedicated systems. However, a disadvantage is the poorer image quality compared with dedicated systems. Nevertheless, our results show that 1 T systems are suitable for application in volumetry of inoculated mouse tumours. Owing to its large cost, the use of MRI-based volumetry is not justified in the case of small tumours. However, when using sliding calliper measurements, huge errors are encountered when trying to measure large tumours (i.e. larger than 1.5 cm3). Moreover, when using tumour models which present large necrotic areas (or a large number of smaller necrotic areas), sliding calliper measurements cannot account for these necrotic parts, thus causing errors in the measurement of the tumour tissue. MRI-based methods can account for these necrotic areas (see below). A third, less important reason why MRI-based volume measurements should be considered, is its superior spatial resolution. Especially in subcutaneously inoculated tumours, the volume of the tumour can be measured very accurately and reproducibly.
The fact that semi-automated volume measurements yielded significantly different values compared with the gold standard is explained by the exclusion of necrotic tissue areas by the semi-automated method. Necrosis is primarily seen in larger tumours. Indeed, when excluding tumours bigger than 1.5 cm3 as measured by water volumetry, the Wilcoxon signed-rank test does not show any significant difference between both groups (assymptomatic significance=0.063). The paired samples t-test could not been used anymore because the data for region growing volumetry were not normally distributed anymore (Shapiro-Wilks coefficient=0.87). Interoperator and intraoperator reliability for the region growing method was however good. Therefore, it is recommended to use the semi-automated method for the measurement of tumour volumes in mice, because it is both reliable and accurate for smaller tumour volumes (<1.5 cm3). In the case of larger tumours, the method is reliable and it accounts for necrotic areas within the tumour.
|
Conclusion
|
|---|
Both manual and semi-automated tumour volumetry in subcutaneous tumour bearing athymic mice by low-field MRI are accurate and reliable methods. The semi-automated method is especially useful for larger tumour volumes, since it accounts for necrotic areas within the tumour.
Received for publication June 4, 2004.
Revision received November 4, 2004.
Accepted for publication December 13, 2004.
|
References
|
|---|
- Coakley FV, Kurhanewicz J, Lu Y, Jones KD, Swanson MG, Chang SD, et al. Prostate cancer tumor volume: measurement with endorectal MR and MR spectroscopic imaging. Radiology 2002;223:91–7.[Abstract/Free Full Text]
- Gibbs P, Buckley DL, Blackband SJ, Horsman A. Tumour volume determination from MR images by morphological segmentation. Phys Med Biol 1996;41:2437–46.[CrossRef][Medline]
- Joe BN, Fukui MB, Meltzer CC, Huang Q-S, Day RS, Greer PJ, et al. Brain tumor volume measurement: comparison of manual and semiautomated methods. Radiology 1999;212:811–6.[Abstract/Free Full Text]
- Kaus MR, Warfield SK, Nabavi A, Black PM, Jolsz FA, Kikinis R. Automated segmentation of MR images of brain tumors. Radiology 2001;218:586–91.[Abstract/Free Full Text]
- Lundin P, Engström BE, Karlsson FA, Burman P. Long-term octreotide therapy in growth hormone-secreting pituitary adenomas: evaluation with serial MR. Am J Neuroradiol 1997;18:765–72.[Abstract/Free Full Text]
- Mayr NA, Taoka T, Yuh WT, Denning LM, Zhen WK, Paulino AC, et al. Method and timing of tumor volume measurement for outcome prediction in cervical cancer using magnetic resonance imaging. Int J Radiat Oncol Biol Phys 2002;52:14–22.[CrossRef][Medline]
- Stamey TA, McNeal JE, Yemoto CM, Sigal BM, Johnstone IM. Biological determinants of cancer progression in men with prostate cancer. JAMA 1999;281:1395–400.[Abstract/Free Full Text]
- End DW, Smets G, Todd AV, Applegate TL, Fuery CJ, Angibaud P, et al. Characterisation of the antitumour effects of the selective farnesyl protein transferase inhibitor R115777 in vivo and in vitro. Cancer Res 2001;6:131–7.
- Orlova A, Bruskin A, Sjöström A, Lundqvist H, Gedda L, Tolmachev V. Cellular processing of 125I- and 111In-labelled epidermal growth factor (EGF) bound to cultured A431 tumor cells. Nucl Med Biol 2000;27:827–35.[CrossRef][Medline]
- Sundberg AL, Almqvist Y, Orlova A, Blomquist E, Jensen HJ, Gedda L, et al. Combined effect of gefitinib (‘Iressa’, ZD1839) and targeted radiotherapy with 211At-EGF. Eur J Nucl Med Mol Imaging 2003;30:1348–56.[CrossRef][Medline]
- Van den Berg AP. Enkele kernpunten bij onderzoekskunst en bewerking van resultaten. Personal communication.
- Waterton JC, Alott CP, Pickford R, Hamilton EA, Valcaccia B. Assessment of mouse tumour xenograft volumes in vivo by ultrasound imaging, MRI and calliper measurement. Proceedings of the 19th LH Gray Conference: Quantitative Imaging in Oncology 146–149; Eds Faulkner K, Carey B, Crellin A, Harisson RM. BIR, London 1997.
- Hardy PA, Recht MP, Piraino D, Thomasson D. Optimization of a dual echo in the steady state (DESS) free-precession sequence for imaging cartilage. J Magn Reson Imaging 1996;6:329–35.[Medline]
Top
Abstract
Introduction
