Nikk Conneman, Abstract Co-Author: Nothing to Disclose

Maarten Lequin MD, PhD, Abstract Co-Author: Nothing to Disclose

Ewout Vansteenkiste, Presenter: Nothing to Disclose

Paul Govaert, Abstract Co-Author: Nothing to Disclose

In analyzing white matter damage (WMD) we hitherto depended on the sonographer to compare the intensity of flaring with the “brightness” of choroid plexus and healthy subcortical white matter. Visual grading of the severity of WMD with ultrasound is not objective. The aim of this study was to find an objective tool for measuring the extent of WMD on ultrasound.

Starting from expert-based subjective evaluation we developed an integrated semi-automatic segmentation method, that yields a reproducible estimation of the extent of injury. The technique combines both textural information of the investigated tissue as well as mathematical morphology in order to delineate the boundaries of the affected area. Added right and left flaring extent in pixel number was compared at different segmentation thresholds. Thresholds (0.7,0.8,0.9,1.0,1.1) were determined by the relative grey value of the region of interest to plexus. Blinded to classification 40 images were segmented by a clinician. The images were coronal sections through the atrium, taken with an Acuson Sequoia 8.5 MHz scanhead, at identical gain and depth. All US frames were of preterms < 32 w gestation who had first week MRI or where lesions became cystic. The patients were graded as follows: [0 normal MRI] n=18 [1 haemorrhagic WMD on MRI] n=14 [2 extensive cystic PVL] n=8.

Descriptive statistics of the flaring extent in pixels were compared between cohorts of different MR grade. Comparing the normal group with the pathological one (1 and 2) the two-tailed P value was < 0.0001, extremely significant for the threshold 0.8 (Mann-Whitney U). Normal flaring extent at this threshold was 3383 pixels (sd 1390). Pathological flares exceeded + 2sd of normal in 17 of 22 cases. The sensitivity of a flaring area exceeding the mean + 2 sd of normal for predicting WMD grade 1 or 2 was 77 %, the specificity 94 %. The positive predictive value of the test was 94 %.

This semi-automatic segmentation method of flaring is able to distinguish between normal and WMD. If applied within the ultrasound machine it could be a robust bedside method for quantification of WMD.

Conneman, N, Lequin, M, Vansteenkiste, E, Govaert, P, Sonographic Segmentation of Preterm White Matter Damage.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4413487.html