Meeting Banner
Abstract #3205

Semi-Automated Segmentation of Microhemorrhages Revealed by SWI

Randall R. Benson1, Ramtilak Gattu2, Balaji Myrtheunjayan3, Zhifeng Kou4, Ewart M. Haacke4

Until now neuropsychological and cognitive deficits in treated PKU patients could not be correlated to their known brain lesions found in conventional MRI. In this work, in addition to the findings that the brain lesions are not corresponding to the therapeutic compliance of the patients, microstructural changes in normal appearing brain tissue in treated PKU patients are disclosed by using quantitative proton/T2-mapping and DTI, which may indicate a global neurotoxic effect of the elevated phenylalanine levels, and contributes to the understanding of the pathomechanisms in PKU patients.

Conventional MRI is insensitive to milder traumatic brain injury. Susceptibility-weighted imaging (SWI) has demonstrated superior sensitivity to microhemorrhages in TBI. Manual lesion counting is labor intensive and operator dependent. Automated methods offer many advantages. The current study summarizes our recent efforts to develop an automated method of lesion segmentation and quantification. The method relies on intensity based probability mapping along with masking the major sources of false positive artifact. Results from 16 TBI patients demonstrate the feasibility, accuracy and clinical correlation of the method.