# Quantitative evaluation of apparent diffusion coefficient in a large multi-unit institution

## TL;DR

This study evaluated the accuracy of ADC measurements across 23 MRI scanners and found higher errors at lower ADC values and in sagittal orientation.

## Contribution

The study provides a quantitative evaluation of ADC accuracy across a large multi-vendor MRI fleet using a NIST-traceable phantom.

## Key findings

- Lower ADC values and sagittal orientation showed increased error and lower signal-to-noise ratio.
- Bland-Altman analysis found no significant differences between 1.5T and 3T scanners.
- ADC measurements among three phantoms were not significantly different.

## Abstract

•An established diffusion phantom was scanned across 23 magnetic resonance imaging scanners, two manufacturers, and two field strengths, and consistently showed increased error and lower signal–noise-ratio at lower apparent diffusion coefficient (ADC) values.•ADC measurements scanned in sagittal orientation consistently showed higher coefficient-of-variation across the reference ADC values.•No significant differences were seen in Bland-Altman analysis across field strengths.

An established diffusion phantom was scanned across 23 magnetic resonance imaging scanners, two manufacturers, and two field strengths, and consistently showed increased error and lower signal–noise-ratio at lower apparent diffusion coefficient (ADC) values.

ADC measurements scanned in sagittal orientation consistently showed higher coefficient-of-variation across the reference ADC values.

No significant differences were seen in Bland-Altman analysis across field strengths.

Diffusion weighted imaging (DWI) and their calculated apparent diffusion coefficient (ADC) values in magnetic resonance imaging (MRI) has grown with applications across multiple anatomical sites and settings across diagnostic and oncologic therapy settings, necessitating characterizing its quantitative performance across multiple scanners. Therefore, the purpose of this study was to determine the quantitative accuracy of DWI and their calculated ADC values across a fleet of MRI systems using a NIST-traceable diffusion phantom.

Three NIST/QIBA DWI phantoms were imaged with protocols provided with the phantom on a multi-vendor fleet of 23 clinical MRI scanners. User-directed regions-of-interest on each vial provided ADC measurements among a wide range of NIST-traceable ADC values. Across all the MRI systems and data, the coefficient-of-variation was calculated, and Bland-Altman analysis was performed. The measurements between phantoms were analyzed to assess their agreement to determine if multiple phantoms can be used for quality assurance across a large institution.

Lower ADC values and the sagittal orientation exhibited the largest error range and coefficient-of-variation (CoV) and was directly related to SNR. From a one-way analysis of the variance (ANOVA), the mean and standard deviation of the percent errors from each phantom were not significantly different from one another. No strong differences in Bland-Altman plots were seen between 1.5T and 3T scanners, however confidence intervals were different among vendors. The ADC values among the three phantoms were not significantly different from one another.

A comprehensive analysis of institution-wide MRI scanners was conducted, motivating further work to utilize these results to implement a comprehensive quality assurance program and advanced scheduling system to better match similarly performing MRI scanners.