Validation of diffusion measurements obtained on a 0.35 T MR in Malawi: Important insights for radiologists in low income settings with low field MRI

Abstract

Purpose

To investigate the reliability of diffusion weighted image (DWI) measurements obtained on a 0.35 T MR scanner in Malawi for malaria research.

Materials and methods

The same healthy volunteers (n = 6) were scanned on a 0.35 T MR scanner in Malawi and a 3 T scanner in the US. Three subjects had two repeated DWI scans at 0.35 T. Due to scanner constraints, only three diffusion gradient directions for DWI on 0.35 T could be obtained. An apparent diffusion coefficient (ADC) map was reconstructed from the 0.35 T and the result was compared to standard DWI acquisition on the 3 T scanner. The mean ADC from 15 different regions and the voxel-wise coefficient of variation (CV) were calculated to investigate the intra-scanner and inter-scanner variability. Reproducibility was calculated using intra-class correlation coefficient (ICC).

Results

The 0.35 T intra-scanner ADC repeatability was high for all three subjects with repeated scans (ICC > 0.7). The intra-scanner correlation between repeated scans was also high (r > 0.67, p < 0.01). Comparing the ADC findings from the 0.35 T and 3 T MRs, the high inter-scanner correlation suggested that the 0.35 T ADC results were valid (ICC > 0.7, r > 0.5, p < 0.01). Voxel-wise CV revealed a few regions with larger variation (CV > 20%), which were primarily located in peripheral regions and the boundary of lateral ventricles, and likely due to partial volume effects in low field scans.

Conclusion

These findings support the validity of DWI obtained from low field MR scanners used in many low income countries.

Introduction

Efforts on the control and prevention of malaria have not yet achieved a sustained impact in the epicenter of the disease, sub-Saharan Africa. Cerebral malaria (CM) is a severe complication of malaria that is clinically characterized by coma. The fatality rate is 15–25%, the annual incidence is 1–12 cases per 1000 children in malaria-endemic regions [1], [2], and 85% of the morbidity and mortality of CM occurs in young African children [3]. Even now, almost half a million lives are lost to malaria each year [4] and ~ 300,000 pediatric CM survivors sustain brain injuries resulting in long term neurologic sequelae [5]. In Malawi alone, malaria accounts for ~ 18% of all hospital deaths [6].

In order to better elucidate the pathophysiology of cerebral malaria, an MRI facility was established at a Hospital in Malawi in 2008. Due to poor access to dependable cryogen delivery, a 0.35 T MRI scanner with a resistive magnet was installed. From 2009 to 2012, children with pediatric cerebral malaria admitted to the Hospital underwent serial brain MRIs. Diffusion-weighted image (DWI) abnormalities were commonly identified [7], but the low field scanner does not generate quantitative parameters that are commonly used in modern diffusion imaging such as apparent diffusion coefficient (ADC) maps. Since either high or low ADC levels can lead to positive findings on DWI, corresponding to different pathological states (vasogenic and cytotoxic edema, respectively), the Malawi-based cerebral malaria MRI work underscored the need to validate DWI findings obtained on such low field scanners with limited capabilities. It should be noted that such scanners are the mainstay of MRI imaging in many resource limited settings since they are less expensive to purchase and operate and do not require reliable access to cryogens. Additionally, common neurologic conditions in resource-limited settings including stroke are also imaged utilizing these low field scanners accentuating the need for understanding the validity of DWI findings from low field MR to assure diagnostic accuracy. To this end, we sought to evaluate the validity of DWI findings at low field using conventional diffusion studies typically provided at higher fields.

We chose to use a modern 3 T MR scanner as a reliable reference in our study, since several groups have previously shown the ADC measurements are highly reproducible between different sites and vendors using high field MR [8], [9]. An open low-field MR scanner has been assessed for detecting cerebral ischemia using DWIs compared to a 1.5 T MRI scanner [10], however, it merely demonstrated the ability of detecting high intensity lesions on DWI using the low-field MR. In our study, we focused on quantitatively assessing the voxel-by-voxel ADC values using the DWIs acquired on a low-field MR scanner, compared to a modern 3 T MR scanner.