A Population-Based Gaussian Mixture Model Incorporating ¹⁸F-FDG PET and Diffusion-Weighted MRI Quantifies Tumor Tissue Classes

Abstract

The aim of our study was to create a novel Gaussian mixture modeling (GMM) pipeline to model the complementary information derived from 18 F-FDG PET and diffusion-weighted MRI (DW-MRI) to separate the tumor microenvironment into relevant tissue compartments and follow the development of these compartments longitudinally. Methods: Serial 18 F-FDG PET and apparent diffusion coefficient (ADC) maps derived from DW-MR images of NCI-H460 xenograft tumors were coregistered, and a population-based GMM was implemented on the complementary imaging data. The tumor microenvironment was segmented into 3 distinct regions and correlated with histology. ANCOVA was applied to gauge how well the total tumor volume was a predictor for the ADC and 18 F-FDG, or if ADC was a good predictor of 18 F-FDG for average values in the whole tumor or average necrotic and viable tissues. Results: The coregistered PET/MR images were in excellent agreement with histology, both visually and quantitatively, and allowed for validation of the last-time-point measurements. Strong correlations were found for the necrotic (r 5 0.88) and viable fractions (r 5 0.87) between histology and clustering. The GMM provided probabilities for each compartment with uncertainties expressed as a mixture of tissues in which the resolution of scans was inadequate to accurately separate tissues. The ANCOVA suggested that both ADC and 18 F-FDG in the whole tumor (P 5 0.0009, P 5 0.02) as well as necrotic (P 5 0.008, P 5 0.02) and viable (P 5 0.003, P 5 0.01) tissues were a positive, linear function of total tumor volume. ADC proved to be a positive predictor of 18 F-FDG in the whole tumor (P 5 0.001) and necrotic (P 5 0.02) and viable (P 5 0.0001) tissues. Conclusion: The complementary information of 18 F-FDG and ADC longitudinal measurements in xenograft tumors allows for segmentation into distinct tissues when using the novel GMM pipeline. Leveraging the power of multiparametric PET/MRI in this manner has the potential to take the assessment of disease outcome beyond RECIST and could provide an important impact to the field of precision medicine.

Probabilistic classification of tumour habitats in soft tissue sarcoma

“…Our hypercellular tumour habitat probably corresponds to the viable tumour reported in literature, extremely cell dense and presumably well oxygenated with low ADC and T 2 . Necrosis featuring long T 2 and high ADC resembles that reported in the literature, generally acellular and similar to bulk water . The collagenous stroma habitat is new and warrants further investigation.…”

“…Connective tissue, viable tumour, and necrosis were distinguished based on FDG‐PET and ADC in lung cancer xenografts . Our hypercellular tumour habitat probably corresponds to the viable tumour reported in literature, extremely cell dense and presumably well oxygenated with low ADC and T 2 . Necrosis featuring long T 2 and high ADC resembles that reported in the literature, generally acellular and similar to bulk water .…”

“…Four compartments were identified in prior studies of colon cancer xenografts using k ‐means clustering algorithms, including two types of viable tumour and two types of necrosis . Connective tissue, viable tumour, and necrosis were distinguished based on FDG‐PET and ADC in lung cancer xenografts . Our hypercellular tumour habitat probably corresponds to the viable tumour reported in literature, extremely cell dense and presumably well oxygenated with low ADC and T 2 .…”

“…Recent preclinical studies have reported that the multispectral analysis of ADC, T 2 , and proton density parameter maps with k ‐means clustering algorithms can aid the differentiation between viable and necrotic tumour, and the identification of multiple compartments within necrotic tissue . A similar approach using with k ‐means clustering and Gaussian mixture modelling of ADC and positron emission tomography with fluorodeoxyglucose (FDG‐PET) differentiated two types of viable tissue and necrosis tissue in xenograft tumours . While these are promising methods to study intra‐tumour heterogeneity, k ‐means clustering partitions voxels into mutually exclusive clusters.…”

“…[18][19][20] A similar approach using with k-means clustering and Gaussian mixture modelling of ADC and positron emission tomography with fluorodeoxyglucose (FDG-PET) differentiated two types of viable tissue and necrosis tissue in xenograft tumours. 21 While these are promising methods to study intra-tumour heterogeneity, k-means clustering partitions voxels into mutually exclusive clusters. This imposes that a tumour voxel can only belong to a single habitat, which might not reflect the complex microenvironment in a voxel, especially at the border of two habitats.…”