Abstracts

RATIONALE: High-resolution MRI of the brain has made it possible to identify focal cortical dysplasia (FCD) in an increasing number of patients. However, in many of them, FCD lesions are characterized by minor structural abnormalities that often go unrecognized or are too subtle to be detected by standard radiological analysis. We previously developed a voxel-based image processing method including first-order texture and morphological analysis using features that were chosen to model in vivo the MRI characteristics of FCD (1). However, the visual assessment of these maps can be a subjective process.
The purpose of this study was to implement an objective method of detection of FCD lesions by developing a fully automated classifier. We hypothesized that FCD lesions could be detected automatically by combining the models described above with second-order texture analysis, which has the ability to quantify spatial patterns of gray-level intensities within an MRI that may reflect tissue organization.
METHODS: We selected 18 patients with histologically proven FCD and 20 neurologically normal controls. FCD lesions were visually identified and manually segmented on T1-weighted 3D MRI prior to the automatic classification for later use in validation. Two sets of feature maps were calculated over the 3D T1-weighted MRI. The first set of maps was based on the characteristics of FCD as seen on T1-weighted MRI: gray matter thickness to model cortical thickening; gradient magnitude to model blurring of the GM-WM junction; relative intensity to model hyperintense signal within the FCD lesion. The next set of feature maps was generated by a series of second-order texture operators. Both sets of feature maps were used to train a Bayesian classifier to identify FCD lesions.
RESULTS: FCD lesions were correctly detected by the classifier in 16 of 18 patients, as indicated by agreement with the location of the manually segmented lesion on the T1-weighted 3D MRI. The classifier did not detect any lesions in the control group.
CONCLUSIONS: We demonstrate the ability of a novel and objective computer-aided method to automatically detect focal cortical dysplasia on high-resolution T1-weighted MRI. In addition to the use of discernable MRI characteristics of focal cortical dysplasia , our method is based upon the analysis of spatial patterns of gray level intensities, which are more difficult to appreciate visually. Therefore, this technique has the potential for detecting abnormalities in patients with [dsquote]non-lesional[dsquote] partial epilepsy.
References
1. Bernasconi A, Antel S, Collins DL, Bernasconi N, Olivier A, Dubeau F, Pike GB, Andermann F, Arnold DL. Texture analysis and morphological processing of MRI assist detection of focal cortical dysplasia in extra-temporal partial epilepsy. Annals of Neurology 49:770-775, 2001.
[Supported by: Canadian Institutes of Health Research and Savoy Foundation]