|
Automatic characterization of renal stones composition from CT data |
|
The objective was to develop an automated tool,
usable in routine clinical practice, allowing determining the composition of
a renal stone from the knowledge of the voxels intensities
in CT exams. A set of 118 renal stones which composition was
determined from infrared spectrometry have been scanned and classified. The voxels were automatically extracted in the image data and
a dissimilarity index has been used to characterize the local homogeneity of
the stone. The classification has been established from the homogeneous
regions of the stones. Stone composition is
correctly determined in 52% of the cases. Sensitivity and specificity
according to the components are the following: uric acid (65%-92.5%), struvite
(19%-93.5%), cystine (78%-96.5%), carbapatite (33.5%-89%), dihydrated
calcium oxalate (57%-86.5%), monohydrated calcium oxalate (66.5%-89%), brushite (75%-95.6%). The CT scan Low Dose acquisition
does not negatively influence significantly the results (p<0.05). Such an
automatic classification enables envisioning the optimal destruction mode of
the stone. |
|
|
|
Reference : Chevreau G, Troccaz J, Conort P, Renard-Penna R, Mallet A, Daudon M, Mozer P. Estimation of urinary stone composition by automated processing of CT images. Urol Res. Oct 2009 37(5):241-245. (pdf ) |
|
Collaborations : Pitié Salpétrière Hospital
(Dr |
