This paper introduces two novel models, RU-Net and R2U-Net, which are based on the U-Net architecture and incorporate Recurrent Convolutional Neural Networks (RCNN) and Recurrent Residual Convolutional Neural Networks (RRCNN). These models aim to improve the performance of medical image segmentation tasks, particularly in the context of retina blood vessel segmentation, skin cancer lesion segmentation, and lung lesion segmentation. The proposed models leverage the strengths of U-Net, Residual Networks, and RCNN to enhance feature representation and training efficiency. Experimental results on three benchmark datasets demonstrate that RU-Net and R2U-Net achieve superior performance compared to existing models, including U-Net and ResU-Net, with the same number of network parameters. The models show better training and validation accuracy, and their effectiveness is further validated through quantitative metrics such as accuracy, sensitivity, specificity, F1-score, Dice coefficient, and Jaccard similarity. The paper also discusses the computational efficiency of the proposed models.This paper introduces two novel models, RU-Net and R2U-Net, which are based on the U-Net architecture and incorporate Recurrent Convolutional Neural Networks (RCNN) and Recurrent Residual Convolutional Neural Networks (RRCNN). These models aim to improve the performance of medical image segmentation tasks, particularly in the context of retina blood vessel segmentation, skin cancer lesion segmentation, and lung lesion segmentation. The proposed models leverage the strengths of U-Net, Residual Networks, and RCNN to enhance feature representation and training efficiency. Experimental results on three benchmark datasets demonstrate that RU-Net and R2U-Net achieve superior performance compared to existing models, including U-Net and ResU-Net, with the same number of network parameters. The models show better training and validation accuracy, and their effectiveness is further validated through quantitative metrics such as accuracy, sensitivity, specificity, F1-score, Dice coefficient, and Jaccard similarity. The paper also discusses the computational efficiency of the proposed models.