Multiscale Texture Analysis: From ¹⁸F-FDG PET Images to Histologic Images

Abstract

Characterizing tumor heterogeneity using texture indices derived from PET images has shown promise in predicting treatment response and patient survival in some types of cancer. Yet, the relationship between PET-derived texture indices, precise tracer distribution, and biologic heterogeneity needs to be clarified. We investigated this relationship using PET images, autoradiographic images, and histologic images. Methods: Three mice bearing orthotopically implanted mammary tumors derived from transgenic MMTV-PyMT mice were scanned with 18 F-FDG PET/CT. The tumors were then sliced, and the slices were imaged with autoradiography and stained with hematoxylin and eosin. Six texture indices derived from the PET images, autoradiographic images, and histologic images were compared for their ability to capture heterogeneity on different scales. Results: The PET-derived indices correlated significantly with the autoradiography-derived ones (R 5 0.57-0.85), but the values differed in magnitude. The histology-derived indices correlated poorly with the autoradiography-and PET-derived ones (R 5 0.06-0.54). All indices were slightly to moderately influenced by the difference in voxel size and spatial resolution in the autoradiographic images. The autoradiography-derived indices differed significantly (P , 0.05) between regions with a high density of cells and regions with a low density and between regions with different spatial arrangements of cells. Conclusion: Heterogeneity derived in vivo from PET images accurately reflects the heterogeneity of tracer uptake derived ex vivo from autoradiographic images. Various tumor-cell densities and spatial cell distributions seen on histologic images can be distinguished using texture indices derived from autoradiographic images despite the difference in voxel size and spatial resolution. Yet, tumor texture derived from PET images only coarsely reflects the spatial distribution and density of tumor cells. A current challenge in oncology is to offer each patient a personalized treatment that accounts, at best, for the characteristics of the tumor or tumors. To achieve this goal, the specific histologic or genetic features of the tumor, including intratumoral heterogeneity, have to be determined before therapy is prescribed (1). This determination is usually based only on biopsy analysis, yet one study (2) reported that 63%-69% of all mutations were not detectable within a single tumor sample. Indeed, intratumoral heterogeneity reflects different mutations in cell subpopulations distributed in different regions of the tumor, some of which may be far from the biopsy site. PET provides molecular information about the whole tumor and may therefore be a relevant option to complete the characterization of the tumor. Indeed, PET imaging is noninvasive, can be repeated during the course of therapy, and allows for a comprehensive assessment of each tumor site. Using different tracers, PET can even yield information on the glucose metabolism of cells ( 18 F-FDG), tumor hypoxia ( 18 F-fluoroazomyc...

Understanding Changes in Tumor Texture Indices in PET: A Comparison Between Visual Assessment and Index Values in Simulated and Patient Data

“…The comparison of artificial and real lesions confirms that texture indices from the first family were immune to the nature of the heterogeneity (hypoor hypersignal). This result is also consistent with our previous investigation of the biologic meaning of texture indices (16). In that investigation, we demonstrated that texture indices of the first family measured from autoradiography images were sensitive to the cell pattern seen on a histologic slice, whereas texture indices from the second family were mostly sensitive to cell density.…”

“…These characteristics still certainly reflect some underlying biologic processes. In addition to the visual interpretation of image-derived features addressed here, efforts should thus be pursued to elucidate their microscopic biologic meaning as initiated previously (16,26).…”

Prediction of cervical cancer recurrence using textural features extracted from 18F-FDG PET images acquired with different scanners

“…AUC corresponding to Entropy were slightly higher than those found by Yang et al [24] in both groups (0.70 vs. 0.66), but the difference was not significant. In a previous study, Entropy calculated on PET images was significantly different between regions presenting different patterns of cells identified on pathological slices, suggesting that PET texture analysis captures the cellular heterogeneity of tumor and might provide additional information on tumor aggressiveness [32]. Unlike [24], our results showed a statistically significant correlation between SUV values and response to treatment, as it was previously reported [6, 33].…”

Association between textural and morphological tumor indices on baseline PET‐CT and early metabolic response on interim PET‐CT in bulky malignant lymphomas

“…Recent attempts have also been made to correlate textural features with pathological prognostic factors in breast cancer and to exploit textural analysis for therapeutic response prediction and monitoring in esophageal, cervical, and breast cancers and for planning external radiation beam therapy in head and neck cancers . Beyond clinical considerations, the heterogeneity computed on PET‐CT correlates well with that computed ex‐vivo from autoradiography and the spatial cell patterns observed on pathological slides …”