Liver segmentation plays a crucial role in the diagnosis and treatment of liver diseases. The proposed method for liver lesion segmentation utilizes the DeepLesion dataset which contains many bio-medical CT scan images with a variety of liver pathologies. The method is based on a popular driven image segmentation deep learning model called U-Net segmentation model. The research includes several regularization techniques such as batch normalization, spatial dropout, and Gaussian noise, to prevent overfitting and also to improve generalization performance has used Adam optimizer. The model is trained by combining binary accuracy with Dice coefficient loss and evaluated the performance of our model using standard evaluation metrics model correctly predicted the lesion with true positive rate of 94%. The proposed automated liver segmentation method has the capability to increase the efficiency and accuracy of diagnosing and treating liver disease. It can be incorporated into existing clinical workflows to aid radiologists in the interpretation of CT scans.