The paper introduces MultiResUNet, an enhanced version of the U-Net architecture for multimodal biomedical image segmentation. The authors found that while U-Net performs well, it has limitations in handling images with varying scales and challenging segmentation tasks. To address these issues, they proposed modifications to the U-Net architecture, including the introduction of MultiRes blocks and Res paths. MultiRes blocks allow for multi-resolutional analysis by incorporating different convolutional filter sizes, while Res paths enhance feature consistency between encoder and decoder layers through additional processing and residual connections. These modifications lead to improved performance, particularly on challenging images. The proposed architecture was tested on five different medical image datasets, achieving relative improvements of 10.15%, 5.07%, 2.63%, 1.41%, and 0.62% over the classical U-Net. The results show that MultiResUNet outperforms U-Net in terms of segmentation accuracy and robustness, especially in handling images with irregular boundaries, perturbations, and outliers. The 3D version of MultiResUNet also performs better than the 3D U-Net. The study concludes that MultiResUNet is a significant advancement in the field of biomedical image segmentation, offering improved performance and reliability compared to the traditional U-Net architecture.The paper introduces MultiResUNet, an enhanced version of the U-Net architecture for multimodal biomedical image segmentation. The authors found that while U-Net performs well, it has limitations in handling images with varying scales and challenging segmentation tasks. To address these issues, they proposed modifications to the U-Net architecture, including the introduction of MultiRes blocks and Res paths. MultiRes blocks allow for multi-resolutional analysis by incorporating different convolutional filter sizes, while Res paths enhance feature consistency between encoder and decoder layers through additional processing and residual connections. These modifications lead to improved performance, particularly on challenging images. The proposed architecture was tested on five different medical image datasets, achieving relative improvements of 10.15%, 5.07%, 2.63%, 1.41%, and 0.62% over the classical U-Net. The results show that MultiResUNet outperforms U-Net in terms of segmentation accuracy and robustness, especially in handling images with irregular boundaries, perturbations, and outliers. The 3D version of MultiResUNet also performs better than the 3D U-Net. The study concludes that MultiResUNet is a significant advancement in the field of biomedical image segmentation, offering improved performance and reliability compared to the traditional U-Net architecture.