Proceedings 2004 IEEE International Symposium on Biomedical Imaging: From Nano to Macro - ISBI’04 vol:2 pages:1127-1130
2004 IEEE International Symposium on Biomedical Imaging: From Nano to Macro - ISBI’04 location:Arlington, VA, USA date:15-18 April 2004
We present a framework for automatic detection of Focal Cortical Dysplastic (FCD) lesions from MR images of human brain. Our method extends, and improves the lesion detection specificity of a previously published voxel-based technique  using cortical thickness and signal gradient as discriminating features of FCD lesions. In absence of any prior anatomical hypothesis regarding the spatial location of the lesion, the method examines each intra-cerebral voxel individually and simultaneously, and constructs a statistical parametric map indicating evidence against a null hypothesis of no effect in the patient versus a normal control group. Upon interrogation of the statistical map with an optimally selected threshold, the voxels demonstrating the improbability of the null hypothesis are reported as lesions. The method correctly detects 55 out of the 10 cases with a very high significance. The cases we did not detect were in deep gray matter regions, where the variance in the feature maps was high. decreasing the significance of the effect.
Srivastava S., Maes F., Vandermeulen D., Van Paesschen W., Dupont P., Suetens P., ''Feature-based statistical analysis of structural MR data for automatic detection of focal cortical dysplastic (FCD) lesions'', Proceedings 2004 IEEE International Symposium on Biomedical Imaging: From Nano to Macro - ISBI’04, vol. 2, pp. 1127-1130, April 15-18, 2004, Arlington, VA, USA.