ABSTRACT

In vivo molecular imaging techniques is increasingly used in the management of oncological patients, allowing different aspects of oncological pathologies to be assessed (e.g. metabolism, hypoxia) non invasively. The possibility to extract indexes of disease from in vivo biomedical images and to associate them with their biological drivers opens new prospective on the role of in vivo molecular imaging and expedites the translation of novel biomarkers from the bench to the clinical environment. In this work we investigate the relationship between ¹⁸F-FDG uptake measured by BodyWeight Standardized Uptake Value (SUV_BW) as index of cell glucose metabolism, and histological indices for gastric and gastro-oesophageal cancer. For this purpose, Partial Volume Effect Correction (PVC) has been properly compensated prior to the measurement of the PET index (PVC-SUV_BW). The correlation of ¹⁸F-FDG PVC-SUV_BW with histology data was evaluated by bivariate and multivariate statistical analysis. Although obtained in a limited number of patients, our results suggest that correlations can be found when PVC is applied to SUV_BW and that ¹⁸F-FDG PET can provide information on biological characteristics of gastric and gastro-oesophageal cancer lesions.

1. Introduction

In vivo disease biomarkers are increasingly demanded by clinicians in order to characterize a disease, to make prognosis and to predict response to treatment non invasively.

Body-Weight Standardized Uptake Value (SUV_BW), measured by Fluorodeoxyglucose and Positron Emission Tomography (¹⁸F-FDG PET), has been extensively used as semi-quantitative index accounting for altered glucose metabolism of an oncological lesion. In order to obtain an accurate measurement of ¹⁸F-FDG SUV_BW, a Correction for Partial Volume Effect (PVC) has been proved mandatory, since this effect causes severe underestimation of SUV_BW (up to 80% - 90% for small lesions) [1,2].

Aim of this work was to evaluate the metabolic impact of PVC on SUV_BW as potential in vivo prognostic biomarker of gastric and gastro-oesophageal cancer, reflecting ex vivo histo-pathological characteristics.

2. Materials and Methods

Fourty-nine patients (31 men, 18 women; mean age 63 ± 13 years; age range: 33 - 83 years) with biopsy-proven gastric and gastro-oesophageal cancer underwent a basal ¹⁸F-FDG PET-CT study. ¹⁸F-FDG PET/CT sensitivity was assessed. Patient weight, and injected/residual dose were measured in order to calculate PVC-SUV_BW on primitive gastric and gastro-oesophageal lesions detected on ¹⁸F-FDG PET images.

Patients fasted for twelve hours before the exams and were intravenous injected with ¹⁸F-FDG (1 mCi/10 kg). The PET-CT protocol began 60 minutes after the injection. All PET-CT studies were performed according to the oncological clinical protocol implemented on the discovery STE scanner, including a SCOUT scan at 40 mA, a CT scan at 140 keV and 150 mA (10 s) and 3D PET scans (2.5 min/scan) for adjacent bed positions. PET images were reconstructed by a 3D ordered subset expectation maximization algorithm (OSEM, 28 subsets, 2 iterations, 5.14 mm Gaussian post-smoothing) with corrections for random, scatter and attenuation incorporated into the iterative process.

An Operator Independent technique using an automatic threshold was used to define Regions of Interest (ROIs) on PET images [2] and quantitative analysis was performed by calculating mean SUV_BW for each primitive gastric and gastro-oesophageal lesions. The RC-based correction methods developed in [2] was used to correct in order to account for Partial Volume Effect.

Both SUV_BW and PVC-SUV_BW were obtained by TouchSUV software [3,4].

Correlation tests (Mann-Whitney and Kruskal Wallis tests for univariate analysis and hierarchical clustering combined with a pre-processing k-means analysis for multivariate analysis) were performed in order to evaluate the relationships between ¹⁸F-FDG PVC-SUV_BW and biopsy-evaluated histotype (signet ring cell carcinoma (SR), squamous cell carcinoma (S) and other adenocarcinoma (ADK) subtype) and grade (G1, G2, G3) (according to WHO and Lauren classifications).

3. Results

¹⁸F-FDG PET/CT was able to detect gastric and gastrooesophageal cancer with a sensitivity of 82%: ¹⁸F-FDG PET/CT images of 9 (18%) biopsy-proven gastric cancers were classified as negative, showing no ¹⁸F-FDG uptake in the primitive lesions. Negative PET images were not quantified and were excluded by correlation analysis.

Mean primitive lesion diameter (sphere-equivalent diameter) was 2.15 ± 1.17 cm, ranging from 0.99 cm to 6.25 cm. Lesion size confirmed the need of PVC for accurate PET quantification for more than 75% of lesions [2].

Signet ring cell carcinomas showed a lower ¹⁸F-FDG PVC-SUV_BW compared to squamous cell carcinomas and to other adenocarcinoma subtypes (PVC-SUV_BW: 5.57 ± 3.22 g/cc vs 9.90 ± 1.91 g/cc vs 9.32 ± 4.26 g/cc; Mann-Whitney test, p < 0.05). No correlation was found when PVC was not applied to SUV_BW.

No correlations were found between grade and ¹⁸FFDG PVC-SUV_BW or ¹⁸F-FDG SUV_BW (Kruskal Wallis test, p > 0.05).

Figure 1 shows the results for univariate analysis.

The pre-processing k-means cluster analysis allowed to stratify patients in three different groups on the basis of PVC-SUV_BW (PVC-SUV_BW ≤ 6.25 g/cc; 6.25g/cc < PVC-SUV_BW < 11.60 g/cc; PVC-SUV_BW ≥ 11.60 g/cc). Using these groups, the hierarchical cluster analysis performed on PVC-SUV_BW, histotype and grade showed that poorly differentiated (G3) signet ring cell carcinomas were significantly associated with ¹⁸F-FDG PVCSUV_BW ≤ 6.25 g/cc (p < 0.05), while moderately differentiated (G2, G1)  squamous cell carcinoma were significantly associated with ¹⁸F-FDG PVC-SUV_BW > 11.6 g/cc; (p < 0.05).

Figure 2 shows the results of the hierarchical cluster analysis.