An equilibrium model for characterization and classification of dominant intraprostatic lesions and normal tissues in patients with prostate cancer.
Bagher-Ebadian H, Janic B, Liu C, Pantelic M, Hearshen D, Chetty IJ, Elshaikh M, Movsas B, and Wen N. An equilibrium model for characterization and classification of dominant intraprostatic lesions and normal tissues in patients with prostate cancer. Med Phys 2017; 44(6):3096.
Purpose: This study introduced an Equilibrium-Model (EM) as an innovative approach for MR-based classification of Dominant-Intraprostatic- Lesions (DILs) from normal tissues (NT) in patients with prostate cancer (PCa). Methods: Twenty-one patients with evidence of PCa with no priortreatment underwent MRI study. An ultrasound-guided needle-biopsy was performed to confirm the diagnosis. T2-weighted-images (T2WI) and Diffusion- Weighted-images were acquired from the pelvis of patients using a 3TMR scanner. The Apparent-Diffusion-Coefficient (ADC) map was constructed from Diffusion-Weighted-images. Using the diagnostic report of each patient, a set of ROIs delineating DILs and NTs were drawn on their image modalities. It was hypothesized that arrangement and decoration of voxel-intensities on the outer-layer of tumor can serve as a prescribed-Dirichlet boundary-constrain to derive an EM of its inner-core information under steady-state condition. An iterative method was used to solve the Laplace equation for estimating the EM intensities for inner-core of DILs/NTs measured from T2WI and ADC-map. An EM-distance (EMD) was calculated from normalized difference between the averages of MRI and EM intensities. The EMD was used to classify DILs from NTs and the results were evaluated with one-way ANOVA. Results: The EMD measured from T2WI showed strong classification performance (19/21-90%, p < 0.0002, FC = 4.08, FClass = 17.24) for DILs (EMD = 0.0576 ± 0.0733) and NTs (EMD = -0.0187 ± 0.0415). The EMD measured from ADC also showed a high classification performance (19/21-90%, p < 0.0007, FC = 4.08, FClass = 13.56) for DILs (EMD = 0.0228 ± 0.0284) and NTs (EMD = -0.0055 ± 0.0208). Also, the low Correlations (r = 0.37, p < 0.09 for T2WI and r = 0.51, p < 0.015 for ADC-map) between the DILs-EMDs and NTs-EMDs confirmed their robustness. Conclusion: This study demonstrated the feasibility of using an EM to construct a matric for classification of DILs and TNs from multi-modal MR information in PCa patients. The performance of the classifier may be improved if the two EMDs measured from T2WI and ADC-map are combined and DILs-EMDs are adjusted with the inter-variation of the NTs-EMDs.