H-Index, a Metric for Tumor Hypoxia Discrimination Derived from the IVIM Diffusion MRI Model

Motivation: There are large variations in perfusion-related quantifications using the IVIM model in disease diagnosis and therapeutic response evaluation. Goal(s): This study aims to improve the reliability of applying IVIM model in hypoxic level classification by considering the model parameter collinearity. Approach: We introduced a Bayesian inference method to estimate IVIM parameter probability distribution, followed by a linear discrimination analysis. This analysis produces a robust metric for distinguishing hypoxic and non-hypoxic tumor tissues. Results: A reliable metrics, as a linear combination of two IVIM parameters (Dt and Fp), accurately reflects tumor tissue hypoxia levels using established outcomes as training dataset and a reference. Impact: We addressed the challenge of uncertainty in IVIM parameter fitting arising from the strong collinearity inherent in the IVIM biexponential model. Additionally, we introduced a robust metric, H-index, for distinguishing between hypoxic and non-hypoxic tumors, referencing previous histopathologically confirmed data.