Segmenting nuclei cells is a serious challenge in the field of medical science, as nuclear morphology is an essential component of most cancer grading schemes. It can be quite difficult to identify nuclei in biomedical images, as well as to precisely identify their borders and/or separate overlapping nuclei. This study attempted to create a deep learning algorithm to efficiently segment such biomedical images and isolate individual nuclei from cell images. The U-Net model has been used in this study because this architecture has demonstrated remarkable effectiveness in picture segmentation tasks in the medical domain. In order to determine the ideal parameters for nuclei segmentation, the modified U-Net model is optimized during training using a combination of loss functions, such as dice loss and binary cross-entropy loss. The segmentation performance of the improved U-Net model is assessed using unseen nucleus images after it has been trained. Here, experimental results show that the suggested method performs well in terms of accuracy and efficiency, achieving good performance in nuclei cell segmentation. The suggested approach has a lot of opportunities for use in cytology, drug development, cell biology research, and histopathology analysis, which could help progress biomedical science and healthcare. Lastly, our study addresses the increasing need for accurate and automated biological image analysis across a range of biomedical domains by advancing nucleus cell segmentation approaches.