Image segmentation, whose objectives are set by well-defined tasks and which is generally significant for automatic applications, deals with isolating image pixels of interest. Data; which might be redundant, useless or even task-complicating due to amount and rawness, must be reduced, compact representations must be obtained and useful features must be extracted. In this work, a heuristic approach for detecting skull contours in monochrome ultrasound images displaying transcerebellar fetal skulls is described. At the start of the process, the user is expected to mark few points manually along the skull boundary on the input image. Due to the bright-pixels composition of skull contours and discontinuities between edge segments arising from imaging modality, the heuristic method utilizes the concepts of average shape model and intensity-based computation of average positions. The results on sample images, when compared to ground-truth segmentations, indicate a 96,5% similarity on average. When the visually-satisfactory outputs are used as inputs in diagnosis systems for spina bifida detection, either better or comparable results are obtained in terms of F-measure and GMRP. The proposed method is supposed to be a facilitative factor in certain automatic diagnosis systems.