Michael Berks

Purpose: To apply voxel-wise tracer kinetic model selection, characterise the spatial distribution of best-fitting models across the brain, and evaluate whether model selection improves sensitivity for differentiating normal-appearing tissue from pathological tissue compared to the Patlak model. Methods: Extended Tofts, Patlak, and intravascular models were fit to DCE-MRI data from stroke survivors and controls, as well as simulated data. The best-fitting model was chosen for each voxel using the Akaike Information Criterion, and model selection K ^trans (estimates from the best-fitting model for each voxel) compared to Patlak model K ^trans . Results: In simulated data, the Extended Tofts model was best-fitting at K ^trans >10 ⁻³ min ^-1 , where the Patlak model systematically underestimated K ^trans . Patlak was optimal at K ^trans between 10 ⁻⁴ -10 ⁻³ min ^-1 , where Extended Tofts estimates had greater variability. The intravascular model was selected for K ^trans ∼10 ⁻⁴ min ^-1 . The Patlak model was chosen in most control voxels. In chronic stroke, the Extended Tofts model was preferred in most cortical and white matter hyperintensity voxels, while the Patlak model was selected in most deep grey matter and normal-appearing white matter voxels. Model selection K ^trans estimates were significantly greater than Patlak estimates in the cortex and white matter hyperintensities, with greater inter-patient variability, likely reflecting biological variability in blood-brain barrier leakage resulting from stroke. Conclusion: Voxel-wise model selection may provide more accurate estimates of a wider range of K ^trans values than any single model, revealing greater differences between normal and pathological tissue and offering a more sensitive and physiologically appropriate framework for DCE-MRI analysis of blood-brain barrier dysfunction.

Supplementary Figure S2. Illustration of the data that is used to derive the ΔR1, HFMRI and HVMRI parameters.

Supplementary Figure S8. Treatment-induced changes in OE-MRI biomarkers relative to baseline (BL).

Supplementary Figure S4. Motion correction effect on ΔR1 time-series.

Supplementary Table S4. Summary of cohort lesion parameter median values at baseline and W2 and W4.

Evaluation of concordance of hypoxia modification in patients with two or more lesions, by assessing serial values of HVMRI at pretreatment and W2 and W4.

Supplementary Figure S7. Treatment-induced changes in OE-MRI biomarker ΔR1.

HVMRI changes at W2. A, Waterfall plot showing the percentage change in HVMRI from BL to W2 in all lesions, categorized as primary tumors or nodal metastases. B, Analysis of the 17 lesions imaged at both W2 and W4 compares the reduction from BL at W2 (light blue) and W4 (dark blue). Dashed lines are the asymmetrical RC LOA for HVMRI (i.e., −45.7% and +84.1%).

Waterfall plots showing change in HVMRI, NVMRI, and WTV at W2. Four patterns seen were (A) lesions with significant hypoxia modification and reduction in WTV. Diamonds indicate tumors in which the hypoxic fraction remained unchanged; B, tumors with significant hypoxia modification that did not have change in WTV; (C) tumors that did not have significant individual changes in hypoxia, but significant reduction in WTV was driven by reduction in NVMRI; (D) tumors with no reduction in HVMRI, NVMRI, or WTV.

Supplementary Figure S3. Patient recruitment and imaging sample size.