Evaluation of diffusion models in breast cancer.
View/ Open
Date
2015-08-01Author
Panek, R
Borri, M
Orton, M
O'Flynn, E
Morgan, V
Giles, SL
deSouza, N
Leach, MO
Schmidt, MA
Type
Journal Article
Metadata
Show full item recordAbstract
PURPOSE: The purpose of this study is to investigate whether the microvascular pseudodiffusion effects resulting with non-monoexponential behavior are present in breast cancer, taking into account tumor spatial heterogeneity. Additionally, methodological factors affecting the signal in low and high diffusion-sensitizing gradient ranges were explored in phantom studies. METHODS: The effect of eddy currents and accuracy of b-value determination using a multiple b-value diffusion-weighted MR imaging sequence were investigated in test objects. Diffusion model selection and noise were then investigated in volunteers (n = 5) and breast tumor patients (n = 21) using the Bayesian information criterion. RESULTS: 54.3% of lesion voxels were best fitted by a monoexponential, 26.2% by a stretched-exponential, and 19.5% by a biexponential intravoxel incoherent motion (IVIM) model. High correlation (0.92) was observed between diffusion coefficients calculated using mono- and stretched-exponential models and moderate (0.59) between monoexponential and IVIM (medians: 0.96/0.84/0.72 × 10(-3) mm(2)/s, respectively). Distortion due to eddy currents depended on the direction of the diffusion gradient and displacement varied between 1 and 6 mm for high b-value images. Shift in the apparent diffusion coefficient due to intrinsic field gradients was compensated for by averaging diffusion data obtained from opposite directions. CONCLUSIONS: Pseudodiffusion and intravoxel heterogeneity effects were not observed in approximately half of breast cancer and normal tissue voxels. This result indicates that stretched and IVIM models should be utilized in regional analysis rather than global tumor assessment. Cross terms between diffusion-sensitization gradients and other imaging or susceptibility-related gradients are relevant in clinical protocols, supporting the use of geometric averaging of diffusion-weighted images acquired with diffusion-sensitization gradients in opposite directions.
Collections
Subject
Breast
Humans
Breast Neoplasms
Diffusion Magnetic Resonance Imaging
Bayes Theorem
Phantoms, Imaging
Diffusion
Models, Theoretical
Image Processing, Computer-Assisted
Adult
Aged
Aged, 80 and over
Middle Aged
Female
Research team
Magnetic Resonance
Language
eng
License start date
2015-08
Citation
Medical physics, 2015, 42 (8), pp. 4833 - 4839
Publisher
WILEY