The high sensitivity of susceptibility weighted imaging (SWI) to iron has spurred applications in safety monitoring for microbleeds and investigation of cerebral iron homeostasis and its disruption in neurodegenerative diseases¹. However, the limit of detection, sensitivity and specificity of SWI in such applications has not been fully explored. The aim of this study was to examine the sensitivity and specificity of the various SWI modalities to detect cerebral microbleeds (CMB) and changes in iron content associated with bleeding, leaky vasculature, vascular thrombosis or mineralization with histopathological confirmation.

Procedures: The study included 5 cynomologus macaques. A tool monoclonal antibody was administered intravenously once weekly for 8 weeks. MRI was collected on a Philips Achieva 3T scanner at baseline, 2, 4, 6 and 8 weeks in all animals with the exception of one animal receiving an additional scan at 16 weeks. 4 animals were euthanized after 8 weeks and 1 after 16 weeks. Sequences included 2D T2w (TR=3000ms, TE=80ms, flip angle=90, slice thickness = 0.8mm), 2D T2*w (TR=650ms, TE=20ms, flip angle=20, slice thickness =1.5mm) and 3D SWI (TR=30ms, TE=20ms, flip angle=15, reconstructed slice thickness = 0.8,1,2 and 4mm) scans of the brain with an in-plane resolution of 0.5x0.5 mm². For histological confirmation, the perfused, fixed brains were serial sectioned into 40 µm thick axial sections. Every 18^th and 19^th section (approximately 680 µm spacing) were stained with H&E to detect erythrocytes and Perl’s Prussian blue stain to detect granules of iron in blood cells (hemosiderin) respectively.

Analysis: For radiologic-neuropathologic assessment of the CMBs and thrombotic lesions, lesions on MRI characterized by hypointensity on SWI, T2, T2* and hyperintensity on SWI phase, SWIM (Susceptibility Weighted Imaging and Mapping)² and lesions on histology characterized by intense Prussian blue staining clusters were identified for each animal by careful visual inspection of the datasets independently. To quantify the change in iron content and the spatial extent over time, the susceptibility values derived from SWIM were estimated within regions of interest.

One out of the five animals developed CMBs and potential thrombotic lesions evidenced by increased susceptibility contrast that were detectible on MRI by week 2 and confirmed by histology. Once lesions appeared, they remained on all follow up scans, however new lesions appeared during the 8 week monitoring period. The MRI immediately prior to euthanasia was used to count lesions and confirm with those detected on histology. 22 lesions were detected on the 0.8mm thick SWI image and confirmed on histology. 20, 16 and 6 lesions were detected on the SWI images with 1.0mm, 2.0mm and 4.0mm slice thickness respectively demonstrating reduced sensitivity with increased slice thickness. The SWIM and T2w images presented 17 lesions. Majority of the lesions were within basal ganglia structures. In relation to the histological findings, the SWI, T2w, and SWI magnitude images demonstrated zero false positives whereas SWIM demonstrated one false positive (a false interpretation of a vein). On SWIM parametric image, the caudate nucleus exhibited an increasing trend in susceptibility and its spatial extent over time consistent with increased iron build up.CMBs and vascular lesions of varying dimensions were detected on MRI which was in concordance with independent histological findings. The detection sensitivity was the highest for the SWI image followed by SWIM and T2w; longitudinal assessment improved the confidence in radiological detection. Increasing slice thickness had an adverse impact on the detection sensitivity to CMBs. Increases in punctate and diffuse iron content in basal ganglia structures were also detected and quantified using SWIM. The increase in iron content in the caudate over time is likely due to the higher incident lesions in this structure compared to the other brain regions in this animal. SWIM images have been shown to offer improved sensitivity and specificity over SWI in prior reports in humans. However, in the current study SWIM was slightly less sensitive in detecting relatively small CMBs than SWI due to insufficient phase dipole effect at the current spatial resolution (slice thickness) and signal to noise ratio.