UCSF

Recent interest in exploring the clinical relevance of cerebral microbleeds (CMBs) has motivated the search for a fast and accurate method to detect them. Visual inspection of CMBs on MR images is a lengthy, arduous task that is highly prone to human error because of their small size and wide distribution throughout the brain. Several computer-aided CMB detection algorithms have recently been proposed in the literature, but their diagnostic accuracy, computation time, and robustness are still in need of improvement. In this study, we developed and tested a semi-automated method for identifying CMBs on minimum intensity projected susceptibility-weighted MR images that are routinely used in clinical practice to visually identify CMBs. The algorithm utilized the 2D fast radial symmetry transform to initially detect putative CMBs. Falsely identified CMBs were then eliminated by examining geometric features measured after performing 3D region growing on the potential CMB candidates. This algorithm was evaluated in 15 patients with brain tumors who exhibited CMBs on susceptibility-weighted images due to prior external beam radiation therapy. Our method achieved heightened sensitivity and acceptable amount of false positives compared to prior methods without compromising computation speed. Its superior performance and simple, accelerated processing make it easily adaptable for detecting CMBs in the clinic and expandable to a wide array of neurological disorders.