SAMUS was an interactive foundational model for medical image segmentation, demonstrating outstanding performance on the echocardiography dataset, such as CAMUS. Its point-prompt feature had a significant impact on segmentation accuracy, and to comprehensively evaluate the effectiveness of point-based prompts in echocardiographic segmentation, we designed three types of point prompt: random point, middle point, and fixed point, and systematically compared their effects on segmentation performance. The experiments showed that a more precise point enhanced the ability of the model to find the right object with better overall performance. It was also shown that the use of point prompts introduced a new type of error as the model segmented another object when the points were around the borders with low contrast, causing catastrophic failures. Training individual models to segment each label of interest removed the likelihood of the confusion and the use of point prompts showed promising improvements through indicating the position of the border. In conclusion, point prompts could be used to identify the objects as well as guiding the fitting of the border, and should be carefully designed to further improve the segmentation accuracy.