Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Correlation of Intra-Tumor 18F-FDG Uptake Heterogeneity Indices with Perfusion CT Derived Parameters in Colorectal Cancer


Application of textural features analysis to 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) images has been used to characterize intra-tumor uptake heterogeneity and has been shown to reflect disease outcome. A current hypothesis is that 18F-FDG uptake heterogeneity may reflect the physiological tracer uptake related to tumor perfusion. The purpose of our study was to investigate the correlations between intra-tumor uptake heterogeneity and vascular parameters derived from dynamic contrast enhanced (DCE) computed tomography (CT) obtained from an integrated 18F-FDG PET/perfusion CT examination.


Thirty patients with proven colorectal cancer prospectively underwent integrated 18F-FDG PET/DCE-CT to assess the metabolic-flow phenotype. Both CT blood flow parametric maps and PET images were analyzed. Correlations between PET heterogeneity and perfusion CT were assessed by Spearman's rank correlation analysis.


Blood flow visualization provided by DCE-CT images was significantly correlated with 18F-FDG PET metabolically active tumor volume as well as with uptake heterogeneity for patients with stage III/IV tumors (|ρ|:0.66 to 0.78; p-value<0.02).


The positive correlation found with tumor blood flow indicates that intra-tumor heterogeneity of 18F-FDG PET accumulation reflects to some extent tracer distribution and consequently indicates that 18F-FDG PET intra-tumor heterogeneity may be associated with physiological processes such as tumor vascularization.


Colorectal cancer is associated with high morbidity, with a 5-year survival of below 50% for rectal cancer [1]. The role of 18F-FDG PET/CT is well established in this cancer model for the detection of recurrent and residual disease as well as in pre-operative staging [2]. In current clinical practice, clinico-pathological staging is used to identify patients who may benefit from neoadjuvant chemo-radiation pre-operatively and adjuvant chemotherapy following surgery [3}. However 18F-FDG PET is becoming an increasingly established imaging modality in colorectal cancer for staging and response assessment [3], [4]. The prognostic and/or predictive value of PET derived parameters with regard to survival or early assessment of response to therapy (during or before treatment), has been the focus of several studies [5][7]. For instance, extraction of parameters with significant predictive value from the baseline 18F-FDG PET scan has been proposed using metabolically active tumor volumes (MATV, the functional volume of the tumor as it can be seen and delineated on an 18F-FDG PET image) [8].

On the other hand, there has recently been increasing interest in the assessment of intra-tumor 18F-FDG heterogeneity, demonstrating an association between such heterogeneity measures on baseline 18F-FDG PET images and overall patient outcome [9]–. Such studies have been performed within the context of locally advanced esophageal cancer [9], lung cancer [10], as well as cervix and head and neck cancer [11]. The characterization of intra-tumor uptake heterogeneity can be categorized into global, regional (tumor sub-volumes) and local (a few voxels) scales. It has been hypothesized that the measured 18F-FDG PET activity distribution heterogeneity may be correlated with several physiological processes including glucose metabolism but also with necrosis, vascularization and angiogenesis [12], [13].Within this context a robust analysis of the intra-tumor 18F-FDG uptake heterogeneity could therefore potentially allow the assessment of such physiological tumor characteristics.

On the other hand, association of 18F-FDG uptake heterogeneity with subjacent physiological processes cannot be performed intuitively or visually, because of the complex and high order nature of the involved parameters obtained through texture analysis. Previous CT studies using dynamic contrast enhanced sequences to measure regional blood flow (BF in ml/100 g/min), blood volume (BV in ml/100 g), mean transit time (MTT in secs) and permeability surface (PS in ml/100 g/min) have suggested that this imaging technique highlights physiological vascular information and may provide an in-vivo marker of tumor angiogenesis [14] or tumor vascularization [15] and can be useful for monitoring neoadjuvant chemotherapy and radiation therapy [16]. BV is defined as the volume of blood within the vasculature of the tumor. BF is defined as BV rate (per min) through the vasculature in a tumor. BV was shown to be a surrogate marker of microvascular density, which is a measure of angiogenesis and has been shown to be an important prognostic factor in many cancer models [17]. MTT is a measurement of the average time necessary for the blood elements to go through the vasculature. Finally, PS is linked to the diffusion of the contrast agent and consequently quantifies the permeability of the blood vessel tumor barrier.

The main objective of this study is therefore to investigate the correlation between 18F-FDG PET derived parameters, including standardized uptake value (SUV) measurements, MATV and several intra-tumor uptake heterogeneity parameters, previously identified as predictive of response to chemo-radiotherapy [9], and dynamic contrast enhanced CT (perfusion CT) based parameters (BF, BV, MTT and PS). This study represents a first attempt to elucidate on the underlying physiological processes associated with measured regional and local 18F-FDG intra-tumor heterogeneity provided by texture analysis.

Materials and Methods


The study was approved by the University College London (UCL) ethics committee. A written consent also approved by the UCL ethics committee was used in this study for each enrolled patient. Patients with primary colorectal cancer were recruited prospectively from 2007 to 2010 and were scheduled for surgery. In addition to standard staging examinations, eligible adult patients underwent an additional integrated 18F-FDG PET/perfusion CT examination prior to surgery, provided there were no contraindications (uncontrolled diabetes; pregnancy, previous reaction to intravenous contrast agent, renal impairment: serum creatinine>120 µmol/L). The study population consisted of 30 patients (20-male; 10-female; mean age 68±9 y). Staging was performed according to the American Joint Committee on Cancer (AJCC) directives [18]. Seventeen patients had a stage I or II, 13 had a stage III or higher disease. The patients' characteristics are summarized in Table 1.


All patients underwent combined 18F-FDG-PET/perfusion CT on a dedicated integrated PET/64-detector-CT (Discovery VCT, GE Healthcare, Amersham, UK). Fasted patients received an injected dose of (270±80) MBq 18F-FDG and imaging performed on average 66+/-7 min after injection. CT for attenuation correction was from the skull base to upper thigh: 140 kV, 40 mAs, pitch 1.5, 3.75-mm detectors, 5-mm collimation. The PET static emission scan covering the same anatomical area was carried out in 3D mode, consisting of an emission scan of 8 minutes/bed position. Transaxial emission images of 5.47×5.47×3.27 mm (in plane matrix size 128×128) were reconstructed using ordered subsets expectation maximization (OSEM) with two iterations and 28 subsets. The axial field of view was 148.75 mm. No motion correction was applied, since motion can be considered negligible in the body region concerned by the colorectal cancer targeted in this study. This was followed immediately by a perfusion CT examination (120 kV, 60 mAs, axial mode, 2-second interval for first 40 seconds, then 5-second interval; total acquisition: 150 seconds; effective dose: 9 mSv; 10 second delay from injection) which was acquired following intravenous iohexol (50 mL, 350 mg/mL iodine Omnipaque, GE Healthcare; Chalfont St Giles,UK; at 5 mL/s).

Image analysis

PET tumor delineation.

For each patient, primary tumors were identified on 18F-FDG PET images by an experienced nuclear physician. Tumors were then delineated automatically using a previously validated fuzzy locally adaptive Bayesian (FLAB) algorithm [19]. This algorithm allows automatic tumor delineation by computing a probability of belonging to a given “class” (e.g. tumor or background) for each voxel within a 3D region of interest containing the tumor and its surrounding background. This probability is calculated by taking into account the voxel's intensity with respect to the statistical distributions (characterized by their mean and variance) of the voxels in the various regions of the image, as well as its special correlation with neighboring voxels in 3D.

PET extraction of quantitative indices.

Different parameters were extracted from these baseline PET images using the delineated tumor. These parameters included maximum SUV (SUVmax), the metabolic active tumor volume (MATV), the tumor lesion glycolysis (TLG) defined as the product of mean SUV and MATV [20], and 4 local and regional tumor heterogeneity parameters derived from texture analysis [9].

The parameters under consideration are listed in Table 2. SUVmax provides a global characterization of the uptake radiotracer within the tumors. Textural features analysis on the other hand provides a large number of parameters [9]. These parameters were obtained in two steps. Firstly matrices that describe the relationship between voxels within the tumor delineated volumes were extracted [21], [22] As a second step the derived matrices were used to calculate different texture indices related to the heterogeneity of the intra-tumor activity distribution. In this study, we have considered parameters given by 1) co-occurrence matrices that provide information about heterogeneity at a local scale and 2) intensity size-zone matrices describing homogeneous tumor areas which are able to describe heterogeneity at a regional scale. Only a limited number of texture parameters that have been previously shown to be both reproducible and of potential clinical interest in different cancer models [9], [23], [24], as well as being robust with respect to the overall image spatial resolution and the tumor delineation methodology [25] were considered here. Before the derivation of the texture matrices, the voxels intensity within the delineated volume were resampled into 64 distinct values with the following formula:(Eq.1)where, I is the original voxel intensity, Ω is the set of voxels in the delineated volume.

The normalization value of 64 was chosen according to previously conducted studies [23], [24] that identified this value as optimal for resampling the typical range of intratumor SUV voxel values encountered (between 2 and 25), providing sufficient sampling range to avoid compromising the reproducibility of the associated texture parameter measurements. The co-occurrence matrices (M) were used to describe relationships between contiguous voxels in a specific angular direction by summing all intensity transitions in the delineated volume. From these matrices we have considered the entropy and the homogeneity defined by:(Eq.2)(Eq.3)where, M is a co-occurrence matrix, i,j are the rows and columns index. For each delineation 13 co-occurrences matrices were considered (one for each angular direction) and finally we use as indices the mean of entropy and homogeneity for these 13 directions.

The intensity size-zone matrices (N) were used to describe the relationship between homogeneous areas into the tumor and their intensity. These matrices have 64 lines (the value used for the image discretization) and the number of columns is equal to the number of voxel of the largest homogeneous area found inside the segmented tumor volume. Using these matrices two indices were considered allowing the quantification of intensity variability into the homogeneous areas (IV) and the variability in the size of these homogenous areas (SZV). These indices were calculated using the following formula:(Eq.4)(Eq.5)where, Ω is the number of homogeneous areas within the tumor, L the size of the largest homogeneous area within tumor and N(i,j) the number of areas with intensity i and size j.

Figure 1 illustrates for a patient example the different steps of the implemented heterogeneity analysis process.

Figure 1. Heterogeneity analysis process.

Illustration of heterogeneity analysis process: in upper an example of tumor (PET/CT with transverse, coronal and sagittal view), at the bottom: illustration of the FLAB delineation after a manual ROI selection that allows heterogeneity analysis.

Perfusion CT.

Arterial perfusion fraction was calculated based on the maximum slope and dual-input-compartment model methods using freely available software (Basama Perfusion Ver. [26]. The perfusion CT acquisition was assessed by an experienced radiologist using commercial software based on distributed parameter analysis (Perfusion 3.0; GE Healthcare, Chalfont St Giles, UK). A processing threshold of -50 to 150HU was applied. A smoothed arterial-time enhancement curve was derived by placing a circular 20 mm2 region of interest within the best-visualized artery [27]. From the automatically generated parametric maps, tumor blood flow, blood volume, mean transit time and permeability surface was obtained by defining a tumor region of interest just within the tumor outline on all images where the tumor was visible. The mean value was recorded for each of the four parameters considered.

Statistical analysis

Statistical analyses were performed using MedCalc software (version, MedCalc Software, Belgium). The comparisons between PET and perfusion CT parameters were carried out using Spearman's rank coefficient (ρ), while 95% confidence intervals (CI) for ρ were also obtained. An independent Student t-test was performed to determine the differences between perfusion CT and intra-tumor uptake heterogeneity parameters. The ability to classify patients according to the MATV of their primary tumor with respect to staging was investigated using Mann-Whitney U tests as recommended for small and non-Gaussian distributed samples. P-values under 0.05 were considered significant after the application of Bonferroni correction for multiple comparisons.


According to the Mann-Whitney U test a strong correlation was found between MATV and stage, as MATVs of early stage tumors (I/II) were significantly smaller than for higher stages (III/IV; 15±10 cm3 vs. 31±20 cm3, p = 0.006). Table 3 provides Spearman's rank correlation results and their associated p-value between perfusion CT derived parameters (BF, BV, MTT and PS) and PET derived heterogeneity parameters for stage III/IV patients, since for stage I/II none of the investigated parameters were found to be correlated with perfusion indices.

Table 3. Spearman's rank correlation coefficient, |p|, between PET and CT perfusion parameters and theirs associated p-values for stage III/IV patients.

For stage III and IV patients, IV, homogeneity and MATV were significantly correlated with BF. For example, the regional intra-tumor intensity variability was associated with a Spearman's rank correlation of 0.75 (95% CI: 0.35 to 0.92, p = 0.02), whereas the local homogeneity was associated with a Spearman's rank correlation of 0.78 (95%CI: 0.40 to 0.93, p = 0.01) (figure 2A and 2B). MATV was also correlated with BF albeit with a limited statistical significance (|ρ| = 0.71, 95%CI: 0.26 to 0.90, p = 0.05) (figure 2C). In contrast, for stage III and IV, SUVmax was not found to be significantly correlated with BF.

Figure 2. Correlation between PET heterogeneity and BF.

Scatter diagrams with regression line (solid line) and associated 95% CI (curves above and below regression line represented upper and lower bounds of 95% CI) showing the correlation of BF with A) IV (|ρ| = 0.75, 95%CI: 0.35 to 0.92, p = 0.02), B) local homogeneity (|ρ| = 0.78, 95%CI: 0.40 to 0.93, p = 0.01), C) MATV (|ρ| = 0.71, 95%CI: 0.26 to 0.90, p = 0.05) in stage III/IV tumors.

Finally, none of the parameters under investigation in this study were found to be correlated with PS, BV or MTT, independently on the patient stage.


This study is, to our knowledge, the first of its kind assessing the correlation of recently proposed global, regional and local PET measures of intra-tumor 18F-FDG activity distribution heterogeneity with CT derived tumor perfusion parameters. The role and potential clinical value of tumor heterogeneity characterization is a current topic of increasing interest in 18F-FDG PET tumor imaging. Within this context recent studies have shown that the quantification of intra-tumor intensity voxel arrangements in 18F-FDG PET can provide independent prognostic and predictive factors of disease and associated treatment outcomes [9], [11], [28]. At a local level, these tumor heterogeneity parameters quantify the differences of intensity between contiguous voxels, while at the regional level they quantify relationships between homogenous areas of different intensities, associated with tumor regions of increased or reduced metabolism.

One of the remaining issues and recurrent questions is the relationship of the measured tumor 18F-FDG uptake heterogeneity with underlying biology. 18F-FDG uptake heterogeneity can be the result of many physiological processes, such as vascularization, perfusion, tumor aggressiveness, or hypoxia [29], [30]. As already mentioned, our study is the first to evaluate the correlation between PET 18F-FDG uptake heterogeneity patterns and CT perfusion parameters which have previously shown to reflect tumor vascularization and aggressiveness [31]. Several previous studies have in the past investigated the metabolism-perfusion correlation, comparing standard 18F-FDG PET image derived parameters providing a global tumor assessment (SUVmean, SUVmax) and various perfusion CT parameters. These investigations have led to varying results depending on tumor type as well as tumor stage [32]. For example, Janssen et al. found that highly perfused tumors had a higher FDG uptake than relatively low perfused tumors in rectal cancer [33]. Groves et al. showed that there was a significant positive correlation between tumor SUV and tumor perfusion normalized to cardiac output for breast cancer patients [34]. Hirasawa et al. found an inverse significant correlation between tumor perfusion and glucose uptake in human head and neck tumors [35]. On the other hand, Bisdas et al. have shown a significant correlation between SUVmax and BF as well as between SUVmean and PS for head and neck squamous cell carcinomas [36]. Miles et al found a non-consistent relationship between BF and SUVs in non-small cell lung cancer, depending on tumor size and stage, with a coupling of flow and metabolism in small tumors [37]. Finally, Goh et al. found a statistically significant correlation between tumor BF and tumor metabolism for high tumor stages (III and IV) but not for stage I or II in colorectal cancer [38].

BV and PS have been previously shown to correlate with angiogenesis in colorectal tumors [14]. In our study BV, which highlights the volume of circulating blood within the tumor vasculature, and provides a measure of ‘functional vessel density', was not correlated with SUVmax. This result might be due to the limited number of patients included in this study. If a correlation was confirmed in a larger cohort for patients with stage I and II, this result might be explained by the fact that tracer delivery is optimal in highly vascularized small tumors leading to higher 18F-FDG radiotracer uptake. Also, previous studies have shown that in stage III/IV tumors, other phenomena such as hypoxia and the increased presence of necrotic regions can influence this relationship, which can explain the observed lack of significant correlation for high stages [39].

For stage I and II patients there were no statistically significant correlations between any perfusion parameters and heterogeneity parameters. The absence of any correlation with heterogeneity parameters may be due to the small size of tumors in stage I/II patients (MATV of 14±10 cm3) which in combination with the limited PET spatial resolution reduces the potential of local 18F-FDG tracer heterogeneity characterization.

On the other hand, when considering stage III/IV patients (MATV of 31±20 cm3), several statistically significant correlations were found between intra-tumor uptake heterogeneity features and perfusion CT parameters. These correlations concerned both local and regional intra-tumor 18F-FDG tracer heterogeneity parameters as well as tumor MATVs. Amongst the CT perfusion parameters considered blood flow exhibited the highest correlation with the different scale heterogeneity parameters, which suggests that 18F-FDG PET local and regional intra-tumor heterogeneity measurements are associated with tumor vascularization, similarly to BF.

Regarding MTT (BV/BF ratio), no statistically significant correlations were found with local or regional intra-tumor uptake heterogeneity parameters. Again the limited number of patients relative to the number of considered parameters might limit the statistical power of the study. Without correcting for multiple testing, IV, SZV and entropy were indeed correlated with MTT and BV. However if these correlations were highlighted by including more patients, they would probably remain comparatively smaller than the correlation coefficients found for BF. This can be explained by the fact that BV was not significantly correlated with the considered intra-tumor uptake heterogeneity features, leading to a weaker, according to the Spearman's rank, correlation for the ratio of BV and BF.

The choice of carrying out the analysis separately for the two subgroups was motivated by two main reasons. The first is that other phenomenon, such as hypoxia or necrotic areas, are more likely to occur in higher stage tumors. In addition, on smaller stages tumors the heterogeneity parameters are mostly correlated with MATV without providing complementary information. Dichotomization of the patients according to MATV or T stage did not modify the results obtained by dichotomizing according to AJCC stage.

One of the limitations of our study is the small number of included patients, and therefore our results need to be confirmed on a larger tumor population considering also different cancer types. A second limitation concerns the accuracy and robustness of assessing tumor perfusion using DCE-CT that is less accurate than measurements assessed with oxygen-15 water PET. Within this context it has been demonstrated that acquisition settings can have an impact on modeling methodology reproducibility and as a result on the derived perfusion parameter values [40], [41]. Although histopathological data such as VEGF, GLUT, CD105 and HiF1α could have been of interest for the present investigation, it was available for only a few patients and we could therefore not include these biomarkers in the undertaken analysis.

In this study the correlation between perfusion CT parameters and PET was investigated only using PET images with 18F-FDG that reflect tumor metabolism. Obviously as a perspective it will be interesting to investigate other radiotracers with higher specificity in terms of physiological processes targeted such as deoxy-3′-[18F]-fluorothymidine (cell proliferation) or [18F]-fluoromisonidazole (hypoxia). Since tumor heterogeneity characterization aims at quantifying such underlying physiological processes, the use of these alternative radiotracers might provide higher correlations than those obtained with FDG, whose accumulation is mediated by a combination of all these physiological processes and others (hypoxia, proliferation, vascularization, angiogenesis). Any such future study should clearly concentrate on tumor stages III and IV, since textural analysis for tumor heterogeneity characterization may be compromised for small tumors defined by a limited number of voxels in the PET images.

Finally, as a future perspective, it may be also interesting to investigate the correlation of functional and anatomical heterogeneity provided by respectively 18F-FDG PET and DCE-CT images. According to our knowledge there is no study that had investigated this kind of correlation, although anatomical DCE-CT image and FDG PET heterogeneity parameters have been both previously shown to provide valuable indicators of patient survival in non-small cell lung cancer irrespective of treatment regime [42].


Statistically significant correlations were determined between BF perfusion parameter acquired on DCE-CT images and the measurement of uptake heterogeneity patterns on 18F-FDG PET images for stage III/IV patients. Our results confirm that intra-tumor 18F-FDG PET local and regional heterogeneity measures are linked to tumor vascularization as measured by DCE-CT. Intra-tumor 18F-FDG uptake heterogeneity patterns in the PET images may provide complementary information to perfusion parameters provided by DCE-CT.

Author Contributions

Conceived and designed the experiments: FT. Performed the experiments: FT. Analyzed the data: FT MH CCLR. Wrote the paper: FT AMG VG MH CCLR DV. Responsible for data collection: AMG VG. Statistical analysis: PI. Overall data analysis supervision: DV.


  1. 1. Tepper JE, O'Connell M, Niedzwiecki D, Hollis DR, Benson AB 3rd, et al. (2002) Adjuvant therapy in rectal cancer: analysis of stage, sex, and local control final report of intergroup 0114. J Clin Oncol 20: 1744–1750.
  2. 2. Hebertson RA, Scarsbrook AF, Lee ST, Tebbutt N, Scott AM (2009) Established, emerging and future roles of PET/CT in the management of colorectal cancer. Clinical radiology 64: 225–237.
  3. 3. Liebig C, Ayala G, Wilks J, Verstovsek G, Liu H, et al. (2009) Perineural invasion is independent predictor of outcome in colorectal cancer. J Clin Oncol 27: 5131–5137.
  4. 4. Chowdhury FU, Shah N, Scarsbrook AF, Bradley KM (2010) [18F]FD PET/CT imaging of colorectal cancer: a pictorial review. Postgrad Med J 86: 174–182.
  5. 5. Vikram R, Iyer RB (2008) PET/CT imaging in the diagnosis, staging, and follow-up of colorectal cancer. Cancer Imaging 8: S46–S51.
  6. 6. Dimitrakopoulou-Strauss A, Strauss LG, Rudi J (2003) PET-FDG as predictor of therapy response in patients with colorectal carcinoma. Q J Nucl Med 47: 8–13.
  7. 7. Hendlisz A, Golfinopuloulos V, Garcia C, Covas A, Emonts P, et al. (2011) Serial FDG-PET/CT for early outcome prediction in patients with metastatic colorectal cancer undergoing chemotherapy. Ann Oncol 23: 1687–1693.
  8. 8. de Geus-Oei LF, Vriens D, van Laarhoven HW, van der Graff WT, Oyen WJ (2009) Monitoring and predicting response to therapy with 18F-FDG PET in colorectal cancer: a systematic review. J Nucl Med 50S1: 43S–54S.
  9. 9. Tixier F, Cheze Le Rest C, Hatt M, Albarghach N, Pradier O, et al. (2011) Intratumor heterogeneity characterized by textural features on baseline 18F-FDG PET images predict response to concomitant radiochemotherapy in esophageal cancer. J Nucl Med 52: 369–378.
  10. 10. Cook GJ, Yip C, Siddique M, Goh V, Chicklore S, et al. (2013) Are pretreatment 18F-FDG PET tumor textural features in non-small cell lung cancer associated with response and survival after chemoradiotherapy? J Nucl Med 54: 19–26.
  11. 11. El Naqa I, Grigsby PW, Apte A, Kidd E, Donnelly E, et al. (2009) Exploring feature-based approaches in PET images for predicting cancer treatment outcomes. Pattern Recognit 42: 1162–1171.
  12. 12. Basu S, Kwee TC, Gatenby R, Saboury B, Torigian DA, et al. (2011) Evolving role of molecular imaging with PET in detecting and characterizing heterogeneity of cancer tissue at the primary and metastic sites, a plausible explanation for failed attempts to cure malignant disorders. Eur J Nucl Med Mol Imaging 38: 987–991.
  13. 13. Pugachev A, Ruan S, Carlin S, Larson SM, Campa J, et al. (2005) Dependence of FDG uptake on tumor microenvironment. Int J Radiat Oncol Biol Phys 62: 545–553.
  14. 14. Goh V, Halligan S, Daley F, Wellsted DM, Guenther T, et al. (2008) Colorectal tumor vascularity: quantitative assessment with multidetector CT – Do tumor perfusion measurements reflect angiogenesis? Radiology 249: 510–517.
  15. 15. Lazanyi KS, Abramyuk A, Wolf G, Tokalov S, Zöphel K, et al. (2010) Usefulness of dynamic contrast enhanced computed tomography in patients with non-small-cell lung cancer scheduled for radiation therapy. Lung Cancer 70: 280–285.
  16. 16. Bellomi M, Petralia G, Sonzogni A, Zampino MG, Rocca A (2007) CT Perfusion for the monitoring of neoadjuvant chemotherapy and radiation therapy in rectal carcinoma. Radiology 244: 486–493.
  17. 17. Weidner N (1995) Intratumor microvessel density as a prognostic factor in cancer. Am J Pathol 147: 9–19.
  18. 18. Henson JS, Huttler DE, O'Sullivan B, Sobin LH, Yarbro JW (1997) AJCC cancer staging manual. 5th ed. Philadephia: Lippincott-Raven.
  19. 19. Hatt M, Cheze le Rest C, Turzo A, Roux C, Visvikis D (2009) A fuzzy locally adaptive Bayesian segmentation approach for volume determination in PET. IEEE Trans Med Imaging 28: 881–893.
  20. 20. Larson SM, Erdi Y, Akhurst T, Mazumdar M, Macapinlac HA, et al. (1999) Tumor treatment response based on visual and quantitative changes in global tumor glycolysis using PET-FDG imaging. The visual response score and the change in total lesion glycolysis. Clin Positron Imaging 2: 159–171.
  21. 21. Haralick RM, Shanmugam K, Dinstein I (1973) Textural features for image classification. IEEE Trans Syst Man Cybern 3: 610–621.
  22. 22. Thibault G, Fertil B, Navarro C, Pereira S, Cau P, et al.. (2009) Texture indexes and gray level size zone matrix: application to cell nuclei. Pattern Recognition Inf Process 140–145.
  23. 23. Tixier F, Hatt M, Cheze-le-Rest C, Le Pogam A, Corcos L, et al. (2012) Reproducibility of tumor uptake heterogeneity characterization through textural feature analysis in 18F-FDG PET imaging. J Nucl Med 53: 693–700.
  24. 24. Willaime JMY, Turkheimer FE, Kenny LM, Aboagye EO (2013) Quantification of intra-tumour cell proliferation heterogeneity using imaging descriptors of 18F fluorothymidine-positron emission tomography. Phys Med Biol 58: 187–203.
  25. 25. Hatt M, Tixier F, Cheze-Le-Rest C, Pradier O, Visvikis D (2013) Robustness of intratumour 18F-FDG PET uptake heterogeneity quantification for therapy response prediction in oesophageal carcinoma. Eur J Nucl Med Mol Imaging 40: 1662–1671.
  26. 26. Tsushima Y, Funabasama S, Sanada S, Aoki J, Endo K (2009) Development of perfusion CT software for personal computers. Acad Radiol 9: 922–926.
  27. 27. Kanda T, Yoshikawa T, Ohno Y, Kanata N, Takenaka D, et al. (2012) CT hepatic perfusion measurement: Comparison of three analytic methods. Eur J Radiol 81: 2075–2079.
  28. 28. van Velden FHP, Cheebsumon P, Yaqub M, Smit EF, Hoekstra OS, et al. (2011) Evaluation of cumulative SUV-volume histogram method for parameterizing heterogenous intratumoural FDG uptake in non-small cell lung cancer PET studies. Eur J Nucl Med Mol Imaging 38: 1636–1647.
  29. 29. Rajendran JG, Schwartz DL, O'Sullivan J, Peterson LM, Ng P, et al. (2006) Tumor hypoxia imaging with 18F fluoromisonidazole positron emission tomography in head and neck cancer. Clin Cancer Res 12: 5435–5441.
  30. 30. Kunkel M, Reichert TE, Benz P, Lehr HA, Jeong JH, et al. (2003) Overexpression of Glut-1 and increased with oral squamous cell carcinoma. Cancer 97: 1015–1024.
  31. 31. García-Figueiras R, Goh VJ, Padhani AR, Baleato-González S, et al. (2013) CT perfusion in oncologic imaging: a useful tool? AJR Am J Roentgenol 200: 8–19.
  32. 32. Rajpopat M, Miles KA (2012) Imaging vascular-metabolic relationships in tumors. Diagnostic imaging Europe 1: 11–14.
  33. 33. Janssen MHM, Aerts HJWL, Buijsen J, Lambin P, Lammering G, et al. (2012) Repeated positron emission tomography-computed tomography and perfusion-computed tomography imaging in rectal cancer: fluorodeoxyglucose uptake corresponds with tumor perfusion. Int. J. Radiation Oncology Biol. Phys 82: 849–855.
  34. 34. Groves AM, Wishart GC, Shastry M, Moyle P, Iddles S, et al. (2011) Metabolic-flow relationships in primary breast cancer: feasibility of combined PET/dynamic contrast-enhanced CT. Eur J Nucl Med Mol Imaging 36: 416–421.
  35. 35. Hirasawa S, Tsushima Y, Takei H, Hirasawa H, Taketomi-Takahashi A, et al. (2007) Inverse correlation between tumor perfusion and glucose uptake in human head and neck tumors. Acad Radiol 14: 312–318.
  36. 36. Bisdas S, Spicer K, Rumboldt Z (2008) Whole-tumor perfusion CT parameters and glucose metabolism measurements in head and neck squamous cell carcinomas: a pilot study using combined positron-emission tomography/CT imaging. AJNR AM J Neuroradiol 29: 1376–1381.
  37. 37. Miles KA, Griffiths MR, Keith CT (2006) Blood flow-metabolic relationships are dependent on tumour size in non-small cell lung cancer: a study using quantitative contrast-enhanced computer tomography and positron emission tomography. Eur J Nucl Med Mol Imaging 33: 22–28.
  38. 38. Goh V, Engledow A, Rodriguez-Justo M, Shastry M, Peck J, et al. (2012) The flow-metabolic phenotype of primary colorectal cancer: assessment by integrated 18F-FDG PET/Perfusion CT with histologic correlation. J Nucl Med 53: 687–692.
  39. 39. Kan Z, Phongkitkarun S, Koyabashi S, Tang Y, Ellis LM, et al. (2005) Functional CT for quantifying tumor perfusion in antiangiogenic therapy in a rat model. Radiology 237: 151–158.
  40. 40. Goh V, Bartram C, Halligan S (2009) Effect of intravenous contrast agent volume on colorectal cancer vascular parameters as measured by perfusion computed tomography. Clin Radiol 64: 368–372.
  41. 41. Goh V, Liaw J, Bartram CI, Halligan S (2008) Effect of temporal interval between scan acquisitions on quantitative vascular parameters in colorectal cancer: implications for helical volumetric perfusion CT techniques. AJR Am J Roentgenol 191: W288–292.
  42. 42. Win T, Miles KA, Janes SM, Ganeshan B, Shastry M, et al. (2013) Tumor heterogeneity and permeability as measured on the CT component of PET/CT predict survival in patients with non-small cell lung cancer. Clin Cancer Res 19: 3591–3599.