Limits...
Lesion identification using unified segmentation-normalisation models and fuzzy clustering.

Seghier ML, Ramlackhansingh A, Crinion J, Leff AP, Price CJ - Neuroimage (2008)

Bottom Line: Second, we adopt a fuzzy clustering procedure to identify outlier voxels in normalised gray and white matter segments.These two advances suppress misclassification of voxels and restrict lesion identification to gray/white matter lesions respectively.From a clinical perspective, our method should help to compute the total volume of lesion or to trace precisely lesion boundaries that might be pertinent for surgical or diagnostic purposes.

View Article: PubMed Central - PubMed

Affiliation: Wellcome Trust Centre for Neuroimaging, Institute of Neurology, UCL, London UK. m.seghier@fil.ion.ucl.ac.uk

ABSTRACT
In this paper, we propose a new automated procedure for lesion identification from single images based on the detection of outlier voxels. We demonstrate the utility of this procedure using artificial and real lesions. The scheme rests on two innovations: First, we augment the generative model used for combined segmentation and normalization of images, with an empirical prior for an atypical tissue class, which can be optimised iteratively. Second, we adopt a fuzzy clustering procedure to identify outlier voxels in normalised gray and white matter segments. These two advances suppress misclassification of voxels and restrict lesion identification to gray/white matter lesions respectively. Our analyses show a high sensitivity for detecting and delineating brain lesions with different sizes, locations, and textures. Our approach has important implications for the generation of lesion overlap maps of a given population and the assessment of lesion-deficit mappings. From a clinical perspective, our method should help to compute the total volume of lesion or to trace precisely lesion boundaries that might be pertinent for surgical or diagnostic purposes.

Show MeSH

Related in: MedlinePlus

(A) For simulated cases 06 (circles) and 10 (squares), illustration of the Dice index (left) at different U thresholds and ROC curves (right) that plot the true positive rate (sensitivity) on the false positive rate (one minus specificity) for different U thresholds. (B) Axial slices illustrating the lesion boundaries of all simulated cases.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2724121&req=5

fig6: (A) For simulated cases 06 (circles) and 10 (squares), illustration of the Dice index (left) at different U thresholds and ROC curves (right) that plot the true positive rate (sensitivity) on the false positive rate (one minus specificity) for different U thresholds. (B) Axial slices illustrating the lesion boundaries of all simulated cases.

Mentions: At the voxel level, we illustrate the sensitivity of the method on simulated cases 06 and 10 (those presented in Fig. 1C and D). Fig. 6A shows the Dice similarity index at different U thresholds. At low U values (e.g. U < 0.05), the Dice index was small due to high false positive rates. The Dice index reached high values (> 0.7) at intermediate U values suggesting a remarkable correspondence between the identified lesions and the known simulated lesions. The method is also highly sensitive and specific as illustrated by the ROC curves (i.e. curves near to the top-left corner). At the global level, the boundaries of all simulated lesions are shown in Fig. 6B. All lesions were identified successfully, including the extensive oedema in simulated case 01, both the aneurysm and infarct in simulated case 03, dysplasia in simulated case 04, large tissue loss and tissue damage in simulated case 06 and atrophy in simulated case 08. Critically, although a low U threshold was used (U = 0.1) in Fig. 6B, false positives (i.e., intact tissue identified as damaged) were very limited (e.g. brainstem of simulated case 05 and WM tracks in simulated case 07).


Lesion identification using unified segmentation-normalisation models and fuzzy clustering.

Seghier ML, Ramlackhansingh A, Crinion J, Leff AP, Price CJ - Neuroimage (2008)

(A) For simulated cases 06 (circles) and 10 (squares), illustration of the Dice index (left) at different U thresholds and ROC curves (right) that plot the true positive rate (sensitivity) on the false positive rate (one minus specificity) for different U thresholds. (B) Axial slices illustrating the lesion boundaries of all simulated cases.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2724121&req=5

fig6: (A) For simulated cases 06 (circles) and 10 (squares), illustration of the Dice index (left) at different U thresholds and ROC curves (right) that plot the true positive rate (sensitivity) on the false positive rate (one minus specificity) for different U thresholds. (B) Axial slices illustrating the lesion boundaries of all simulated cases.
Mentions: At the voxel level, we illustrate the sensitivity of the method on simulated cases 06 and 10 (those presented in Fig. 1C and D). Fig. 6A shows the Dice similarity index at different U thresholds. At low U values (e.g. U < 0.05), the Dice index was small due to high false positive rates. The Dice index reached high values (> 0.7) at intermediate U values suggesting a remarkable correspondence between the identified lesions and the known simulated lesions. The method is also highly sensitive and specific as illustrated by the ROC curves (i.e. curves near to the top-left corner). At the global level, the boundaries of all simulated lesions are shown in Fig. 6B. All lesions were identified successfully, including the extensive oedema in simulated case 01, both the aneurysm and infarct in simulated case 03, dysplasia in simulated case 04, large tissue loss and tissue damage in simulated case 06 and atrophy in simulated case 08. Critically, although a low U threshold was used (U = 0.1) in Fig. 6B, false positives (i.e., intact tissue identified as damaged) were very limited (e.g. brainstem of simulated case 05 and WM tracks in simulated case 07).

Bottom Line: Second, we adopt a fuzzy clustering procedure to identify outlier voxels in normalised gray and white matter segments.These two advances suppress misclassification of voxels and restrict lesion identification to gray/white matter lesions respectively.From a clinical perspective, our method should help to compute the total volume of lesion or to trace precisely lesion boundaries that might be pertinent for surgical or diagnostic purposes.

View Article: PubMed Central - PubMed

Affiliation: Wellcome Trust Centre for Neuroimaging, Institute of Neurology, UCL, London UK. m.seghier@fil.ion.ucl.ac.uk

ABSTRACT
In this paper, we propose a new automated procedure for lesion identification from single images based on the detection of outlier voxels. We demonstrate the utility of this procedure using artificial and real lesions. The scheme rests on two innovations: First, we augment the generative model used for combined segmentation and normalization of images, with an empirical prior for an atypical tissue class, which can be optimised iteratively. Second, we adopt a fuzzy clustering procedure to identify outlier voxels in normalised gray and white matter segments. These two advances suppress misclassification of voxels and restrict lesion identification to gray/white matter lesions respectively. Our analyses show a high sensitivity for detecting and delineating brain lesions with different sizes, locations, and textures. Our approach has important implications for the generation of lesion overlap maps of a given population and the assessment of lesion-deficit mappings. From a clinical perspective, our method should help to compute the total volume of lesion or to trace precisely lesion boundaries that might be pertinent for surgical or diagnostic purposes.

Show MeSH
Related in: MedlinePlus