MRI has defined “cerebral microbleeds” as a common phenomena of the aging brain, becoming increasingly frequent beginning at age 60. By age 80, nearly 40% of the population is likely to show evidence of microbleeds (MBs). However, in pathological examination only a very small brain region can be sampled, and not able to provide a comprehensive examination to evaluate the full extent of MBs. It is believed that the prevalence of cerebral microbleeds could easily be higher when more sensitive imaging techniques are used. In this project we developed an MR imaging method to detect possible microbleeds on postmortem brains of patients with and without dementia, hoping to provide information to guide neuropathological examination to sample the suspicious MBs areas, and improve the chance of identifying true microbleeds to better understand its role in normal aging and development/progression of dementia. The postmortem brain contains air-bubbles, and GRE sequence or SWI are very sensitive to these air-bubbles which will generate major artifacts and greatly affect MBs detection. Typical MBs are < 200 µm, and even with blooming effect, the size of MBs is around 0.5 mm, only one pixel or a few pixels in high resolution MR images. Therefore, it is very challenging to detect small MBs in postmortem brain. The goal of this project is to optimize MR imaging procedures for detecting MBs on postmortem brain for guiding tissue pathological examinations; and further to develop streamlined automatic MBs detection software.The air bubbles on the surface and gyri/sulci of the fixed brain specimens will lead to severe artifacts and present as small signal voids on MR images mimicking MBs. To overcome this problem we have developed a special experimental set-up using a vacuum chamber with an ultrasound sonicator probe to remove the bubbles. After the de-bubbling the specimen was imaged. And then the specimen was flipped, and went through the same de-bubbling procedure and imaging again. Only signal voids that were present on both side-A and side-B images were identified as possible microbleeds. The MRI imaging was performed on a Philips 3T scanner using a gradient echo sequence with TR= 47 ms, TE=6.1 ms; flip angle=20°; matrix size 800x800; in plane resolution, 0.25x 0.25 mm; slice thickness= 0.7 mm. Figure 1 shows picture of one brain specimen and the corresponding specimen MRI image. Figure 2 shows the images acquired from side-A and side-B that do not show the same signal voids, thus can be used to identify and exclude bubble artifacts. Two cases (one true positive one true negative) that were sampled and examined thoroughly in the pathological study were used for training purposes to help us develop microbleeds detection criteria on MRI. They were: a) dark signal void on T2*-weighted images; b) round or oval shape; c) blooming on T2*-weighted MRI; d) near or in gray matter not in deep white matter; f) distinct from other potential mimics such as iron/calcium deposits, or remaining blood in vessels. Figure 3 shows the in-vivo MRI and postmortem brain specimen images from one patient with early onset familiar Alzheimer’s disease due to genetic disposition, who was confirmed to have severe amyloid angiopathy in pathological examination. Based on the criteria we could identify 112 MBs lesions on 2 brain specimens for this patient. Figure 4 shows a microbleed-like artifact and Figure 5 shows a case with MBs detected.The specimen imaging was performed in 70 cases with brains coming from an oldest-old 90+ aging study, Alzheimer’s patient cohort, and Down’s syndrome patient cohort. In 18 cases (5 normal aging mean age 92, 10 dementia mean age 85, and 3 Down’s mean age 54), the microbleeds were detected by using Prussian Blue staining and a special image analysis software to measure the staining area. The number of detected microbleeds on specimen MRI was analyzed. The Total Prussian Blue Area was: 2021±936 for Normal aging, 2036±675 for dementia, and much lower 1515±640 for Down’s. The mean number of MBs on specimen MRI was: 1.4±1.2 for normal aging, 0.6±0.2 for dementia, and 0.3±0.3 for Down’s. Excluding Down’s patients and analyzing the association of MBs with age, the prevalence of MBs is 2 of 5 (40%) in younger subjects < 90 years old, and 7 of 9 (78%) in older subjects ≥ 90 years old, P=0.08 for one-sided test. This study and the analysis is on-going, and our results so far seemed to confirm that aging, not dementia, was a major factor associated with MBs.