# Normalization of the percentage of signal recovery derived from dynamic susceptibility contrast perfusion MRI in brain tumors

## TL;DR

This paper proposes a normalization method for signal recovery metrics in brain tumor MRI scans to improve consistency across patients and studies.

## Contribution

The novel contribution is the use of logarithmic normalization with normal-appearing white matter to reduce variability in PSR values.

## Key findings

- Logarithmic normalization using NAWM reduced inter-patient variability in PSR values.
- nPSRln outperformed no normalization and linear normalization in maintaining consistency under varied acquisition parameters.
- The method improves comparability of PSR values across different studies.

## Abstract

The universalizability of the metric percentage of signal recovery (PSR) derived from dynamic susceptibility contrast (DSC) perfusion MRI is limited by its dependency of acquisition parameters. In this technical assessment, we tested different reference tissues for PSR normalization and found the normal-appearing white matter (NAWM) to have the least inter-patient variability when using a fixed PSR-optimized protocol. A logarithmic normalization using NAWM improved the consistency of PSR values when a cohort of brain tumor patients was analyzed by synthetically changing acquisition parameters (while keeping the protocol uniform within the cohort). Additionally, the NAWM logarithmic normalization was better than no normalization and linear normalization at maintaining the consistency of both values and ranks within the cohort while a synthetic random variation of the acquisition parameter was applied (i.e., with a heterogeneous protocol within the cohort). Future PSR studies may consider reporting logarithmic normalized PSR (nPSRln) values to potentially improve the comparability across studies.