About 18-35% of the preterm infants suffer from punctuate white matter lesion (PWML). Accurately assessing the volume and localisation of these lesions at the early postnatal phase can help paediatricians adapting the therapeutic strategy and potentially reduce severe sequelae. MRI is the gold standard neuroimaging tool to assess minimal to severe WM lesions, but it is only rarely performed for cost and accessibility reasons. Cranial ultra-sonography (cUS) is a routinely used tool, however, the visual detection of PWM lesions is challenging and time consuming due to speckle noise and low contrast image. In this paper we perform semantic detection and segmentation of PWML on 3D cranial ultrasonogra-phy. We introduce a novel deep architecture, called Priority U-Net, based on the 2D U-Net backbone combined with the self balancing focal loss and a soft attention model focusing on the PWML localisation. The proposed attention mask is a 3D probabilistic map derived from spatial prior knowledge of PWML localisation computed from our dataset. We compare the performance of the priority U-Net with the U-Net baseline based on a dataset including 21 exams of preterm neonates (131 PWMLs). We also evaluate the impact of the self-balancing focal loss (SBFL) on the performance. Compared to the U-Net, the priority U-Net with SBFL increases the recall and the precision in the detection task from 0.4404 to 0.5370 and from 0.3217 to 0.5043, respectively. The Dice metric is also increased from 0.3040 to 0.3839 in the segmentation task.