SEARCH

SEARCH BY CITATION

Keywords:

  • heterogeneity;
  • tumor;
  • Minkowski functionals;
  • image analysis

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES

Purpose

The acquisition of ever increasing volumes of high resolution magnetic resonance imaging (MRI) data has created an urgent need to develop automated and objective image analysis algorithms that can assist in determining tumor margins, diagnosing tumor stage, and detecting treatment response.

Methods

We have shown previously that Minkowski functionals, which are precise morphological and structural descriptors of image heterogeneity, can be used to enhance the detection, in T1-weighted images, of a targeted Gd3+-chelate-based contrast agent for detecting tumor cell death. We have used Minkowski functionals here to characterize heterogeneity in T2-weighted images acquired before and after drug treatment, and obtained without contrast agent administration.

Results

We show that Minkowski functionals can be used to characterize the changes in image heterogeneity that accompany treatment of tumors with a vascular disrupting agent, combretastatin A4-phosphate, and with a cytotoxic drug, etoposide.

Conclusions

Parameterizing changes in the heterogeneity of T2-weighted images can be used to detect early responses of tumors to drug treatment, even when there is no change in tumor size. The approach provides a quantitative and therefore objective assessment of treatment response that could be used with other types of MR image and also with other imaging modalities. Magn Reson Med 71:402–410, 2014. © 2013 Wiley Periodicals, Inc.

Magnetic resonance imaging of tissue morphology has been widely used in oncology to detect the presence of disease and to detect treatment response by measuring decreases in tumor size [1]. Since for some therapies treatment-induced tumor cell death can be correlated with patient survival [2-4], a more general method for detecting treatment response would be to image tumor cell death. This could be used to detect response to those therapies that have little or no effect on tumor size and provide earlier detection of response to those therapies that do eventually induce tumor shrinkage [5].

Even relatively low resolution images, such as those produced by computed tomography (CT) or MRI, can be a sensitive indicator of underlying tissue biology [6] and therefore could potentially be used to interrogate more subtle features of tumor physiology and changes in this physiology in response to treatment. Such an approach, however, requires the development of image metrics that give objective and quantitative assessments of tissue morphology and that capture the underlying biological information in a routine and automated fashion.

A characteristic of tumors is their heterogeneous appearance in MR images, a consequence of their irregular and uncoordinated growth and a chaotic and intermittent blood supply that leads to periods of transient ischemia and hypoxia. The resulting areas of necrosis and hemorrhage can lead to hyper- and hypointensity, respectively, in T2-weighted images [7] and may also be influenced by treatment, where successful treatment can result in a change in the size and distribution of these areas [8].

Various approaches have been adopted for analyzing heterogeneity in MR images of tumors, including k-means clustering [9, 10], texture analysis [11-13], and fractal dimensions [14-16]. However, texture analysis and the k-means clustering algorithm, have some important limitations. Texture analysis requires a large set of image parameters to be calculated with a subset of these parameters then chosen and used to differentiate between tissues of different types. In a recent study, where texture analysis was used to distinguish between benign and malignant soft tissue masses on MR images, only small differences were identified between the two and it was concluded that more data were required to confirm the value of this approach [13]. A number of variations of the k-means clustering algorithm have been used in MRI; however, in most cases, the algorithm requires prior knowledge of the number of expected regions or features and, furthermore, due to the fact that the algorithm has a randomly assigned starting point, the resulting clusters may not always converge to the same point.

More recent approaches to image analysis have focused on the use of shape-orientated descriptors, such as fractal dimension analysis and Minkowski functionals (MFs), which do not require prior assumptions about the number of regions or features in an image. For example, Rose et al. [17] used fractal dimensions to describe the heterogeneity found in dynamic contrast enhanced (DCE)-MRI parameter maps and showed that the measured heterogeneity could distinguish between low-grade and high-grade gliomas, a distinction that could not be made using distribution-based summary statistics. MFs have been widely employed in cosmology as precise morphological and structural descriptors, and used in the study of the evolution and morphology of galaxies and clusters of galaxies [18, 19]. More recently they have been used as shape functionals in neuromorphometric characterization [20], for classifying normal and pathological pulmonary tissue [21, 22] and as parameters for the analysis of mineral distribution in hip fractures [23]. We demonstrated recently that MFs can be used to parameterize the heterogeneous distribution of a targeted MRI contrast agent for detecting tumor cell death and showed that this increased the sensitivity of cell death detection in a drug-treated tumor [24]. We show here that MFs can be used with T2-weighted images to detect the morphological changes that accompany tumor cell death following drug treatment in the absence of any exogenous contrast agent.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES

Drug Treatment and Tumor Histologic Evaluation

Tumors were grown by subcutaneous injection of 5 × 106 EL-4 murine lymphoma cells into the lower flank of female C57BL/6 mice, and allowed to grow for 10 days. Tumor cell death was induced either by treatment with a cytotoxic drug, etoposide, which also induced tumor shrinkage, or by using a vascular disrupting agent, combretastatin A4-phosphate (CA4P), which produced hemorrhagic necrosis in the absence of any significant change in tumor size. Drug-treated animals received intraperitoneal injections of 67 mg/kg etoposide or 100mg/kg CA4P. Control animals were injected with the solvent vehicle. Procedures were conducted in accordance with project and personal licenses issued under the United Kingdom Animals (Scientific Procedures) Act1986 and were designed with reference to the UK Co-ordinating Committee on Cancer Research Guidelines for the Welfare of Animals in Experimental Neoplasia.

The presence of tumor cell death was confirmed histologically. Tumors were fixed in 10% formalin and embedded in paraffin, and 5 μm sections were stained with hematoxylin and eosin (Fig. 1). The fraction of cells with fragmented nuclei (both apoptotic and necrotic cells) was estimated using ImageJ software (National Institutes of Health) as described in Refs. [8, 25].

image

Figure 1. Representative sections of tumors stained with hematoxylin and eosin from (a) an untreated control tumor, (b) a tumor 24 h post-treatment with etoposide, and (c) 6 h post treatment with CA4P. The etoposide treated tumors show widespread regions of cell death, whereas at 6 h post CA4P treatment these regions are more localized and smaller. Scale bar=300 μm.

Download figure to PowerPoint

MRI

Experiments were performed at 9.4 T using a vertical 89-mm bore Oxford Instruments magnet (Oxford, UK) interfaced to a Varian Inova console (Varian, Palo Alto, CA) and a 45-mm-diameter volume coil (Millipede, Varian). Multi-slice T2-weighted (repetition time (TR)=1.5 s, echo time (TE)=40 ms, four transients per slice, bandwidth=100 kHz, field-of-view (FOV)=35 mm × 35 mm, data matrix 256 × 128, slice thickness 1.5 mm) images were acquired using a spin-echo sequence. The etoposide treatment group (N=8) was imaged pre-treatment, and again 24-h post-treatment, while for the CA4P treatment group imaging was performed pre-treatment and at 6 h (N=8) or 24 h (N=9) post-treatment. A group of untreated control tumor-bearing mice (N=11) were also imaged twice, where the imaging sessions were 24 h apart. A summary of the relevant information about the different treatment categories is given in Table 1.

Table 1. Summary of the Treatment Time Points, Number of Tumors (N), Tumor Size, and Number of Slices Used in the Subsequent Analysis
TreatmentTime (h)NVolume (cm3)Slices
  1. For the CA4P treatment group four of the tumors were imaged at all three time points. A significant difference in volume at baseline between the different treatment groups was observed (P<0.05, one-way ANOVA).

Control0111.1±0.76–17
Control24111.3±0.96–15
Etoposide081.8±0.611–17
Etoposide2481.3±0.59–13
CA4P0130.9±0.45–13
CA4P680.7±0.44–12
CA4P2491.0±0.47–11

Minkowski Functional Analysis

Tumors were segmented manually from transverse T2-weighted images, with a contiguous non-square region extracted, using the image analysis tools in MATLAB (MathWorks, Natick, MA). Representative images of tumors before and after treatment with etoposide and CA4P are shown in Figure 2. Where large fat deposits were present within the extracted tumor region of interest they were removed due to their significantly higher signal intensity. Depending on the size of each tumor, between 4 and 18 slices were segmented.

image

Figure 2. Representative T2-weighted images of EL4 tumors from a control tumor imaged at baseline (a) and 24 h later (b); before (c) and 24 h after treatment with etoposide (d), and before (e) and 6 h after treatment with CA4P (f).

Download figure to PowerPoint

The MF analysis of shape structure in a scalar field is specifically structured as to be invariant with respect to the effects of normalization of the data. In comparing the MFs of two scalar fields, i.e. the intensity distributions in two separate images, each image must be rescaled to span the same range in order to eliminate amplitude differences from the analysis. Therefore, tumor image intensities were linearly remapped onto the uniform interval 0 to 1, independently for each slice of an individual tumor. This was performed by firstly mapping the median of the bottom 1% of image intensities to 0, together with all other pixels having a lower intensity than this median value. Next, the median of the top 1% of image intensities was mapped to 1, together with all other pixels having a higher intensity value than this median value. This was performed to avoid the [0,1] range being influenced by a few anomalous dark or bright pixels. The number of such anomalous pixels was typically less than 1%.

The normalized tumor images were converted into binary datasets by thresholding the images as a function of gray scale. Ten threshold steps were chosen to sample the gray level variation in the image, giving 11 thresholded images per slice (Fig. 3). Increasing the number of threshold steps to 20 did not improve the discrimination between treated tumors and untreated controls (data not shown). In each thresholded image the visible pixels were considered in the computation of the MFs.

image

Figure 3. Thresholded images of the representative tumor images shown in Figure 2 at 11 gray scale threshold steps. a: Control tumor imaged at baseline and 24 h later (b); before (c) and 24 h after treatment with etoposide (d), and before (e) and 6 h after treatment with CA4P (f). In each case the threshold value indicates the normalized signal intensity cutoff in the image.

Download figure to PowerPoint

MFs were calculated using software developed by us and which we have used previously [24]. The software calculates the three 2D MFs; area (A), perimeter (U), and genus (χ), as a function of the image threshold (i) using the following formulae [26],

  • display math(1)
  • display math(2)
  • display math(3)

where pi, ei, and vi are the numbers of pixels, edges, and vertices at threshold i, respectively, with the common edges and vertices of connected pixels counted only once. In Figure 4a for example p=8, e=22 and v=15 which gives A=8, U=12, and χ=1. The area and perimeter were normalized to remove any dependence on the total number of pixels in an image, p0, using nAi=Ai/p0 and nUi=Ui/p01/2, where nAi and nUi are the normalized area and perimeter at threshold i, respectively; the genus is already independent of the number of pixels.

Genus is often known as the connectivity number; in a 2D image this can be considered as the number of regions of connected white pixels minus the number of completely enclosed regions of black pixels (Fig. 4).

image

Figure 4. The 2D MFs: Area (A), Perimeter (U), and Genus (χ), where the genus is the number of connected components minus the number of holes. Considering the white pixels to be signal and the black pixels to be the background we have in (a) A=8, U=12, and χ=1 (one block of connected white pixels and no holes). (b) A=8, U=16, and χ=2 (two blocks of connected white pixels and no holes). (c) A=8, U=16, and χ=0 (one block of connected white pixels and one hole).

Download figure to PowerPoint

Multiple slices were obtained for each tumor, and the resulting MF values for an individual slice are not independent of those from other slices of the same tumor. To account for this the MFs from each slice were combined into a single set of MFs for each tumor. It is possible to do this by either calculating mean MF values at each threshold for each tumor or by applying a weighting by slice area prior to calculating the mean. The weighted mean value approach was chosen as variations in the size and shape of the tumor in different slices are accounted for. The weighted mean values were calculated using the formulae,

  • display math(4)

where wnA is the weighted normalized area, wnU is the weighted normalized perimeter, is the normalized genus, i is the threshold and j is the jth slice of the tumor.

Support Vector Machine Classification

The ability of MFs to detect changes in image heterogeneity post-treatment was tested by building a support vector machine (SVM) classifier [27]. MF features with significant differences (P<0.05) in MF values between the particular treatment and the untreated controls in a two-tailed unpaired t-test were chosen for the classification (Fig. 5-7). The validity of applying a t-test to the data was first confirmed using a D'Agostino and Pearson omnibus normality test (Graphpad Prism, San Diego, CA). The values for each selected feature were scaled to lie in the interval [0,1] prior to classification to avoid bias in the classification as the range of values for the perimeter and genus features were different. SVM classification was performed using the LIBSVM program [28] with a radial basis function kernel. This kernel has two parameters, C the soft margin parameter, and γ the kernel parameter. These parameters were optimized using leave-one-out cross validation for the classifier generated for each treatment.

Classification performance using the optimal SVM parameters was assessed using a receiver operating characteristic (ROC) curve. Vectors classed (by the classifier) as positive when their true classification is positive are termed true positives (TP). Equally, vectors classified as negative by the classifier when they are truly negative are termed true negatives (TN). Finally, vectors classed as positive, which are truly negative, are false positives (FP) and vectors classed as negative, which are truly positive, are false negatives (FN). The “true positive rate” (TPR) or sensitivity is defined as:

  • display math(5)

The “false positive rate” (FPR) is defined as:

  • display math(6)

It is also common to define the specificity as one minus the FPR.

The area under the ROC curve (AUC), which is a graphical plot of TPR versus FPR, provides a measure of how well the data vectors have been classified by this method. Classifiers that can exactly divide the data into true positives and negatives have an ROC area of unity, while classifiers that randomly mix the two classes have an expected ROC area of ½.

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES

Histology

Histological analysis of tumors excised from drug-treated animals was used to determine the extent of cell death following drug treatment. Representative hematoxylin and eosin stained sections from control tumors and tumors treated with etoposide and CA4P are shown in Figure 1. At 24 h following treatment with the cytotoxic drug, etoposide, the area of the tumor sections containing dead cells increased from a control value of (4.8±0.3)% to (33±1)% (N=3) at 24 h [8]. CA4P treatment caused widespread tumor cell death which rose from (6±1)% (N=7) in control tumors to (16±4)% (N=3) at 6 h after treatment, and (36±5)% (N=7) at 24 h post-treatment [25]. The distribution of the regions of apoptosis and necrosis was qualitatively different with the two drugs; etoposide produced widespread regions of apoptosis/necrosis 24-h post treatment (Fig. 1b), as compared to smaller diffuse regions of necrosis in the CA4P-treated tumors (Fig. 1c).

Tumor Signal Intensity

The absolute signal intensity for each tumor was normalized to the signal intensity from surrounding muscle. No significant differences in tumor to muscle signal intensity ratio were seen between pre- and post-treatment images for the control, etoposide, or CA4P treatment groups (P>0.05, two-tailed paired t-test) (Fig. 2).

Tumor Volume

The change in segmented tumor volume between pre- and post-treatment images was assessed for each treatment group. For the untreated control group there was a general but not significant increase in tumor volume between images (P>0.05). For the etoposide treatment group, there was a (29±7)% reduction (P<0.001) in tumor volume post-treatment. For the CA4P treatment group, no significant differences in tumor volume after treatment were observed at either time point (P>0.05). There was a significant difference between the initial tumor volumes between the three groups (P<0.05, one-way ANOVA), with the etoposide treatment group having significantly larger tumors in the baseline scan (Table 1). However, subsequent MF analysis was based on the differences in the MF parameters for images acquired from the same tumor before and after treatment (or no treatment in the case of the control group) and therefore each tumor acted as its own control.

2D MFs

Simple visual inspection of the thresholded images of tumor slices from the control, etoposide, and CA4P-treated groups showed changes in image heterogeneity (Fig. 3), which were not readily discernible in the raw T2-weighted images (Fig. 2). The image heterogeneity was quantified using 2D MF values, and the data for control tumors at baseline and 24 h are shown in Figure 5. Similar plots were obtained for both the etoposide and CA4P treatment groups (data not shown). Changes in MF values in the 24 h between scans for the control group, as well as following etoposide and CA4P treatment were observed; however, the drug treatments yielded different patterns of changes. These changes were quantified by calculating differences in the mean weighted 2D MF values (Eq. (4)) pre- and post-treatment. Figure 6 shows the change in mean weighted MF values for the etoposide-treated animals compared to the untreated controls. Significant differences between the control and etoposide treatment groups were seen for perimeter at thresholds 0.7–1.0 and for genus at thresholds 0.4–0.6 and 0.9–1.0 (P<0.05) (Fig. 6). Changes in the mean weighted MF values for the CA4P-treated animals at both treatment time points compared to the controls at 24 h are shown in Figure 7. For CA4P at 6 h post-treatment, significant differences between treated and control tumors were seen for area at threshold 0.8, perimeter at thresholds 0.8–1.0, and genus at thresholds 0.2, 0.7, and 0.9–1.0 (P<0.05) (Fig. 7). In the 24 h CA4P treatment group, significant differences were observed for perimeter at threshold 1.0, and genus at thresholds 0.7 and 1.0 (Fig. 7).

image

Figure 5. Weighted mean (Eq. (4)) 2D MF parameters (a) area, (b) perimeter, and (c) genus, as a function of gray scale threshold for control tumors imaged at baseline and again 24 h later. Values shown are mean±SD of the 11 control tumors, and an offset along the x-axis has been applied to the second scan data.

Download figure to PowerPoint

image

Figure 6. Changes in the weighted mean 2D MF parameters (a) area, (b) perimeter, and (c) genus, as a function of gray scale threshold. Data shown are for control tumors, and tumors 24 h after treatment with etoposide. Each line (control and etoposide-treated) represents the mean change in MF value between images collected before and after drug- or vehicle-treatment (post-treatment minus pre-treatment (± SEM)). An offset has been applied to the treated data in the x-axis. Thresholds with significant differences (P<0.05) between treated and control tumors are indicated.

Download figure to PowerPoint

image

Figure 7. Changes in the weighted mean 2D MF parameters (a) area, (b) perimeter, and (c) genus, as a function of gray scale threshold. Data shown are for control tumors, and tumors 6 and 24 h after treatment with CA4P. Each line (control and CA4P-treated) represents the mean change in MF value between images collected before and after drug- or vehicle-treatment (post-treatment minus pre-treatment (±SEM)). An offset along the x-axis has been applied to the treated data. Thresholds with significant differences (P<0.05) between treated and control tumors are indicated for 6 h CA4P treatment by *, and 24 h CA4P treatment by †.

Download figure to PowerPoint

SVM Analysis

The MF features described above that showed significant differences (P<0.05) between treated tumors and non-treated controls were chosen for use in an SVM-based classifier. The number of features, and the optimized parameters, C and γ, used for each of the classifiers are given in Table 2. The SVM classifiers constructed for each treatment performed well, achieving an accuracy of 75% or higher (Table 2). For etoposide treatment the sensitivity and specificity were good, as were these parameters for the CA4P treatment group at the 6 h time point. 2D MFs at 24 h post CA4P treatment did not perform as well, with the classifier only giving a sensitivity of 66.7%. The results of the ROC analysis are also given in Table 2.

Table 2. SVM Classification Parameters and Results for Detection of Treatment Response Using 2D MFs with 11 Threshold Steps Using Differences in Tumor MFs (treated–control)
TreatmentFeaturesCγAccuracyAUCTPRFPR
  1. The features column gives the number of MF features used in the classifier. C and γ are the optimized parameters for the SVM classifier, and the accuracy is the result of classification using the SVM trained with these values with leave one out cross validation. Classification performance of the SVM classifier for each treatment group was also assessed using an ROC curve generated from the selected features. The area under this curve (AUC) and the calculated true positive rate (TPR, sensitivity) and false positive rate (FPR, 1—specificity) are indicated.

Etoposide90.5115/190.867/83/11
CA4P 6 h80.6116/190.827/82/11
CA4P 24 h31.0215/200.646/92/11

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES

Tumor responses to treatment in the clinic are conventionally assessed from CT or MR imaging measurements of decreases in tumor size [1]. We have shown previously that cell death in EL4 lymphomas induced by etoposide treatment could be imaged using 2D MFs at 24 h after treatment using a targeted MR imaging agent that binds to the phosphatidylserine exposed by dying cells [24]. We have also shown that the cell death induced by CA4P treatment could be detected at 6 h post-treatment using 13C-MRS measurements of the increased rates of conversion of hyperpolarized [1,4-13C2]fumarate to malate and at 24 h using diffusion-weighted imaging [29]. Since the cellular necrosis resulting from etoposide treatment was expected to lead to regions of increased intensity in T2-weighted images [30] whereas the hemorrhagic necrosis resulting from CA4P treatment was expected to lead to regions of hypo- and hyper-intensity [7]. We have examined here the possibility of detecting treatment response by quantitatively assessing, using 2D MFs, the changes in heterogeneity of T2-weighted images following drug treatment.

The ability of MFs to detect treatment response is based on changes in the MF values with treatment, particularly changes in perimeter and genus values. At low threshold values the “holes” in the thresholded images (Fig. 2), produced by eliminating dark pixels, result in negative genus values (Fig. 5c), which is also seen as a corresponding increase in the perimeter MF (Fig. 5b). As the threshold is increased the genus is sensitive to the number of isolated regions of bright pixels in the original image and the genus values become positive. At higher thresholds the perimeter tends to decrease as the number of remaining pixels decreases.

Changes in the 2D MF values were observed in control tumors in the 24 h between scans (Fig. 5), in particular the area MF at the high thresholds was increased in the 24 h scan, indicating that more bright pixels are present. This may represent changes in tumor morphology, particularly an increase in isolated necrotic areas associated with continued tumor growth. The volume of the segmented tumor was increased in the control group between scans, although this increase was not statistically significant. To account for these changes in the control untreated tumors we subsequently used changes in MF values (treated minus control) for each tumor in the construction of the SVM classifiers. In this way each tumor acted as its own control.

In the case of etoposide treatment of EL4 tumors, there was a significant reduction in the segmented tumor volume 24 h post-treatment, which could be explained by the increase in cell death observed in histological sections. In addition, there were changes in the 2D MF values post-treatment relative to the pre-treatment scans, specifically the perimeter and genus functionals were decreased in the post-treatment scans, indicating changes in the image heterogeneity post-treatment (Fig. 6). These changes in MFs were similar to those seen previously with contrast enhanced T1-weighted images of EL4 tumors post etoposide treatment [24]. The decrease in both genus and perimeter is possibly due the presence of fewer but larger hyperintense regions due to necrosis in the post-treatment images.

There was no significant change in tumor volume either 6 or 24 h post CA4P treatment, however changes in MF values were observed. The pattern of changes in MF values was different for CA4P treated tumors compared to etoposide treatment. The change in both perimeter and genus functionals at low thresholds in the 6 h post-treatment data (Fig. 7) suggests that there was an increase in the number of regions of dark pixels, possibly due to the presence of regions of hemorrhage. At higher thresholds for the 6 and 24 h treatment data there was a reduction in both perimeter and genus functionals, which may indicate an increase in the size of the regions of necrosis similar to that seen with etoposide treatment.

The potential of 2D MFs for detecting treatment response from T2-weighted images was evaluated using ROC analysis, where the curves were generated after SVM classification based on the MF parameters for control tumors and for tumors 6 and 24 h after drug-treatment. The SVM classifier for etoposide performed well, with an AUC of 86% and high specificity and sensitivity (Table 2). While these tumors also showed a significant reduction in tumor volume post-treatment, these results demonstrate that changes in image heterogeneity can be used to detect treatment response in the absence of exogenous contrast. For CA4P at 6 h following treatment the SVM classifier gave an AUC of 82%, and again scored well on specificity and sensitivity. Unlike etoposide treatment, however, there was no observable change in tumor volume; therefore 2D MFs may be applicable clinically to detect treatment response before any changes in tumor volume occur. The 24-h post-treatment group performed poorly in the ROC analysis with an AUC of 64%, a low sensitivity of 67%, but a good specificity. Increasing the number of features by using the criteria of P<0.1 as a feature selector did not improve the classification performance (data not shown). Since CA4P-treated tumors show a loss of tumor perfusion using DCE-MRI at 6 h post-treatment but restored perfusion by 24 h post-treatment [25], the levels of perfusion may be more important than changes in the levels of necrosis in determining whether the 2D MF analysis of image heterogeneity detects treatment response in this case.

Manual segmentation of the tumors, as performed here, does introduce an element of operator-dependent subjectivity into the analysis. The use of automatic segmentation algorithms would be preferable. While the present study has focused on the application of 2D MFs in the analysis of treatment response, 3-dimensional equivalents of the 2D MFs used in this study are available [26]. 3D MFs may be expected to perform as well or better than 2D MFs as in principle they capture more information about the heterogeneity of the tumor as a whole. However, an important consideration is the large difference in resolution between the 2D slices (0.14 × 0.28 mm) when compared to the thickness of the imaging slice, which was 1.5 mm in the dataset used here. This may reduce the capability of the 3D MFs to capture additional heterogeneity when compared to that reported on by the 2D MFs.

Demonstrating that specific aspects of tumor morphology are indeed responsible for the observed features in the T2-weighted MR images requires careful registration of the MR and histological images; a process that is made difficult by the very different slice thicknesses involved; 1.5 mm for the MR slice versus 5 µm for the histological slice. It may be possible to use MF analysis to assign features in the MR image to specific features in the corresponding histological section by demonstrating that they have the same MFs. For example, showing that lightly stained regions in an H&E stained section, which represent regions of necrosis, have the same MFs as hyperintense regions in the corresponding T2-weighted image. However, this analysis was not attempted here.

In conclusion, we have demonstrated that we can detect early responses of tumors to drug treatment by parameterizing, using MFs, changes in heterogeneity in T2-weighted images after drug treatment. This was possible in the absence of observable changes in tumor size in the case of treatment with CA4P. Such an approach provides a quantitative and therefore objective assessment of treatment response. In principle, this method could be used to enhance response detection in any type of MR image, e.g. diffusion-weighted, contrast agent-enhanced and, indeed, in images acquired using other imaging modalities, such as positron emission tomography (PET) and single-photon emission computed tomography (SPECT).

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES

This work was supported by Cancer Research UK programme grant to KMB and a Biotechnology and Biological Sciences Research Council (BBSRC) to KMB and MPH. TB was supported by grants from the Medical Research Council UK, the Royal College of Radiologists and the Addenbrookes Charitable Trust.

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENTS
  7. REFERENCES