Improved segmentation of ultrasound brain tissue incorporating expert evaluation.

The quantitative analysis of medical ultrasound images for the purpose of diagnosis is a difficult task due to the speckle noise present in the images. Nowadays medical doctors depend strongly on the visual interpretation of the images which is subjective to some account. Trying to reduce this noise should assist the experts in a better understanding of some pathologies. We focus on a brain disease called periventricular leukomalacia, also called white matter damage, which occurs frequently on premature neonates. For the moment the affected brain tissue is segmented semi-automatically using two different techniques that take the speckle noise into little account. Here we propose a framework which includes an efficient preprocessing step and relying on expert-based evaluation we develop an integrated segmentation method, which yields a more accurate and better reproducible segmentation.