Differentiating Gastric Cancer and Gastric Lymphoma Using Texture Analysis (TA) of Positron Emission Tomography (PET)
Introduction
Gastric cancer and gastric lymphoma are two distinct malignancies with overlapping clinical and radiological features but differing treatment strategies and outcomes. Accurate differentiation between these conditions is crucial for effective management. While endoscopic biopsy remains a definitive diagnostic method, it is invasive and limited in assessing lesions beyond the submucosal layer. Non-invasive imaging modalities, such as computed tomography (CT) and magnetic resonance imaging (MRI), have been widely used but face challenges in reliably distinguishing between these tumors. Positron emission tomography (PET) with 18F-fluorodeoxyglucose (FDG) offers a metabolic imaging approach that reflects glucose metabolism, providing valuable insights into tumor characteristics. However, the variability in FDG uptake within both gastric cancer and gastric lymphoma limits the diagnostic accuracy of standard PET parameters like the standard uptake value (SUV).
Texture analysis (TA) is a computational technique that quantifies intra-tumor heterogeneity by analyzing the spatial distribution of pixel or voxel intensities in medical images. TA has shown promise in various oncological applications, including diagnosis, staging, and treatment response assessment. In this study, we explored the role of PET TA in differentiating gastric cancer from gastric lymphoma, as well as low-grade from high-grade gastric lymphoma, to improve diagnostic accuracy and clinical decision-making.
Methods
Patient Selection
This retrospective study included 79 patients (45 with gastric cancer and 34 with gastric lymphoma) diagnosed between January 2013 and February 2018. Inclusion criteria required a confirmed diagnosis through gastroscopic biopsy or surgical pathology and a pre-treatment 18F-FDG PET/CT scan. Patients with a history of cancer, prior treatment, incomplete data, or no observable FDG uptake in lesions were excluded.
Image Acquisition
Patients fasted for at least six hours before receiving an intravenous injection of 18F-FDG. PET/CT scans were performed using a 16-slice hybrid scanner, with CT images used for attenuation correction. PET images were reconstructed with a voxel size of 4 mm³.
Texture Analysis
Regions of interest (ROIs) were manually drawn on PET images to encompass the entire tumor volume, avoiding the gastric lumen and adjacent metastatic lesions. Texture features were extracted using in-house software, including first-order histogram features (mean, standard deviation, skewness, kurtosis, entropy, and percentiles) and second-order grey-level co-occurrence matrix (GLCM) features (entropyGLCM and inertiaGLCM). Intra- and inter-observer agreement was assessed to ensure reliability.
Statistical Analysis
Differences in texture features and SUVs between groups were analyzed using the Mann-Whitney test. Receiver operating characteristic (ROC) analysis evaluated the diagnostic efficacy of these parameters. Intra-class correlation coefficients (ICCs) assessed the reproducibility of texture feature measurements.
Results
Patient Characteristics
The study cohort included 47 males and 32 females, with a median age of 60 years. Gastric cancer cases comprised various histological subtypes, while gastric lymphoma cases included mucosa-associated lymphoid tissue (MALT) lymphoma, diffuse large B-cell lymphoma, and others.
Differences Between Gastric Cancer and Gastric Lymphoma
Significant differences were observed in SUVmax, SUVmean, and several texture features between gastric cancer and gastric lymphoma. InertiaGLCM, which quantifies local variation in FDG uptake, was significantly lower in gastric cancer than in gastric lymphoma, with the highest area under the curve (AUC) of 0.714 for differentiation.
Differences Between Low-Grade and High-Grade Gastric Lymphoma
Low-grade gastric lymphoma exhibited significantly lower SUVmax, SUVmean, and entropyHIST compared to high-grade lymphoma. SUVmax and SUVmean were the most effective parameters for distinguishing these groups, with AUCs of 0.818 and 0.842, respectively.
Differences Between Low-Grade Gastric Lymphoma and Gastric Cancer
EntropyGLCM was significantly lower in low-grade gastric lymphoma than in gastric cancer, with an AUC of 0.770 for differentiation. This suggests that entropyGLCM can effectively distinguish low-grade lymphoma from gastric cancer, even when SUVs are similar.
Correlations Between SUVs and Histogram Features
Histogram features, including mean, standard deviation, and percentiles, showed strong positive correlations with SUVmax and SUVmean, indicating their dependence on FDG uptake levels.
ROC Analysis
InertiaGLCM was the most discriminating feature for differentiating gastric lymphoma from gastric cancer, while entropyGLCM was effective in distinguishing low-grade lymphoma from gastric cancer.
Discussion
This study highlights the potential of PET TA in improving the differential diagnosis of gastric neoplasms. InertiaGLCM, which reflects local variation in FDG uptake, was particularly effective in distinguishing gastric lymphoma from gastric cancer. EntropyGLCM, which quantifies tumor heterogeneity, was valuable in differentiating low-grade lymphoma from gastric cancer. These findings suggest that texture features can provide additional diagnostic information beyond traditional SUVs, especially in cases with overlapping FDG uptake levels.
The lower inertiaGLCM in gastric cancer may reflect more uniform FDG uptake patterns compared to the heterogeneous uptake in gastric lymphoma. Similarly, the lower entropyGLCM in low-grade lymphoma suggests a more homogeneous cellular composition, consistent with the aligned monomorphic lymphoid cells characteristic of these tumors.
Limitations
This study has several limitations, including its retrospective design and relatively small sample size. The heterogeneity of pathological subtypes within gastric lymphoma and gastric cancer may also influence the results. Future studies with larger cohorts and prospective designs are needed to validate these findings and explore the potential of PET TA in clinical practice.
Conclusion
PET TA, particularly inertiaGLCM and entropyGLCM, offers a promising non-invasive approach to differentiating gastric cancer from gastric lymphoma and low-grade from high-grade lymphoma. By providing insights into tumor heterogeneity and local FDG uptake patterns, texture analysis can enhance diagnostic accuracy and support personalized treatment strategies.
doi.org/10.1097/CM9.0000000000001206
Was this helpful?
0 / 0