Texture-based tissue characterization: A novel predictor for brain injury?

Premature infants are prone to white matter damage (WMD), which is associated with cerebral palsy (CP) and cognitive impairment. Ultrasound (US) is the preferred imaging modality to detect WMD. To improve on existing diagnostic rates, quantitative measures incorporating new information are needed. We are investigating US texture measures as new indicators of white matter health. We are developing techniques which enhance image texture differences that may reflect pathological changes. Earlier experiments using data from 30 patients indicate that tissue types may be segmented based on texture measures. Present experiments using images from 18 patients (12 with normal outcome, 6 who developed CP) reveal that the measures from CP patients form separate populations from healthy patients. Texture measures were obtained without compensating for operator-dependent machine settings and without suppressing speckle. Digitized analogue ultrasound films are used as the input data since modern digital ultrasound machines have bandwidth restrictions. The effect of image resolution reduction on the texture-based methods is part of an on-going investigation.