# DB-SAM: Delving into High Quality Universal Medical Image Segmentation Source: [DB-SAM: Delving into High Quality Universal Medical Image Segmentation]() Collection index: [llms.txt]() Data updated: 2026-01-02T07:21:19.128Z Authors: Chao Qin, Jiale Cao, Huazhu Fu, Fahad Shahbaz Khan, Rao Muhammad Anwer Published: 2024-10-05T14:36:43.000Z ## Abstract Recently, the Segment Anything Model (SAM) has demonstrated promising segmentation capabilities in a variety of downstream segmentation tasks. However in the context of universal medical image segmentation there exists a notable performance discrepancy when directly applying SAM due to the domain gap between natural and 2D/3D medical data. In this work, we propose a dual-branch adapted SAM framework, named DB-SAM, that strives to effectively bridge this domain gap. Our dual-branch adapted SAM contains two branches in parallel: a ViT branch and a convolution branch. The ViT branch incorporates a learnable channel attention block after each frozen attention block, which captures domain-specific local features. On the other hand, the convolution branch employs a light-weight convolutional block to extract domain-specific shallow features from the input medical image. To perform cross-branch feature fusion, we design a bilateral cross-attention block and a ViT convolution fusion block, which dynamically combine diverse information of two branches for mask decoder. Extensive experiments on large-scale medical image dataset with various 3D and 2D medical segmentation tasks reveal the merits of our proposed contributions. On 21 3D medical image segmentation tasks, our proposed DB-SAM achieves an absolute gain of 8.8%, compared to a recent medical SAM adapter in the literature. The code and model are available at https://github.com/AlfredQin/DB-SAM. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [3D Medical Image Segmentation on Gallbladder (CT)](): 3D Medical Image Segmentation; Gallbladder (CT); DSC: 90.12. - [3D Medical Image Segmentation on Brain Ventricles MR-T1](): 3D Medical Image Segmentation; Brain Ventricles MR-T1; DSC: 78.95. - [3D Medical Image Segmentation on Brain Ventricles MR-T2](): 3D Medical Image Segmentation; Brain Ventricles MR-T2; DSC: 77.02. - [3D Medical Image Segmentation on Brain Tumor MR-FLAIR](): 3D Medical Image Segmentation; Brain Tumor MR-FLAIR; DSC: 0.9249. - [3D Medical Image Segmentation on Cerebellum MR-T1](): 3D Medical Image Segmentation; Cerebellum MR-T1; DSC: 94.47. - [3D Medical Image Segmentation on Cerebellum MR-T2](): 3D Medical Image Segmentation; Cerebellum MR-T2; DSC: 92.69. - [3D Medical Image Segmentation on Gallbladder MR](): 3D Medical Image Segmentation; Gallbladder MR; DSC: 87.43. - [3D Medical Image Segmentation on Left Ventricle MR](): 3D Medical Image Segmentation; Left Ventricle MR; DSC: 0.9115. - [3D Medical Image Segmentation on Right Ventricle MR](): 3D Medical Image Segmentation; Right Ventricle MR; DSC: 90.51. - [3D Medical Image Segmentation on Liver MR](): 3D Medical Image Segmentation; Liver MR; DSC: 96.17. - [3D Medical Image Segmentation on Pancreas MR](): 3D Medical Image Segmentation; Pancreas MR; DSC: 85.16. - [3D Medical Image Segmentation on Prostate MR-ADC](): 3D Medical Image Segmentation; Prostate MR-ADC; DSC: 93.25. - [3D Medical Image Segmentation on Abdomen Tumor CT](): 3D Medical Image Segmentation; Abdomen Tumor CT; DSC: 0.7831. - [3D Medical Image Segmentation on Head-Neck Tumor CT](): 3D Medical Image Segmentation; Head-Neck Tumor CT; DSC: 76.14. - [3D Medical Image Segmentation on Liver CT](): 3D Medical Image Segmentation; Liver CT; DSC: 96.21. - [3D Medical Image Segmentation on Lung Infections CT](): 3D Medical Image Segmentation; Lung Infections CT; DSC: 0.7851. - [3D Medical Image Segmentation on Pancreas CT](): 3D Medical Image Segmentation; Pancreas CT; DSC: 0.838. - [3D Medical Image Segmentation on Pleural Effusion CT](): 3D Medical Image Segmentation; Pleural Effusion CT; DSC: 75.52. - [3D Medical Image Segmentation on Stomach CT](): 3D Medical Image Segmentation; Stomach CT; DSC (%): 92.54. - [3D Medical Image Segmentation on Head-Neck Tumor PET](): 3D Medical Image Segmentation; Head-Neck Tumor PET; DSC: 84.12. - [Breast Tumor Segmentation on MedSAM 1.0 (test)](): Breast Tumor Segmentation; MedSAM 1.0 (test); DSC: 87.43. - [Gland Segmentation on MedSAM 1.0 (test)](): Gland Segmentation; MedSAM 1.0 (test); DSC: 52.79. - [Heart Segmentation on MedSAM 1.0 (test)](): Heart Segmentation; MedSAM 1.0 (test); DSC: 95.35. - [Instrument Segmentation on MedSAM 1.0 (test)](): Instrument Segmentation; MedSAM 1.0 (test); DSC: 91.89. - [Liver Segmentation on MedSAM 1.0 (test)](): Liver Segmentation; MedSAM 1.0 (test); DSC: 81.1. - [Lungs Segmentation on MedSAM 1.0 (test)](): Lungs Segmentation; MedSAM 1.0 (test); DSC: 97.6. - [Polyp Segmentation on MedSAM 1.0 (test)](): Polyp Segmentation; MedSAM 1.0 (test); DSC: 92.13. - [Retinal Vessel Segmentation on MedSAM 1.0 (test)](): Retinal Vessel Segmentation; MedSAM 1.0 (test); DSC: 6753. - [Vessel Segmentation on MedSAM 1.0 (test)](): Vessel segmentation; MedSAM 1.0 (test); DSC: 72.15. - [3D Medical Image Segmentation on Prostate MR-T2](): 3D Medical Image Segmentation; Prostate MR-T2; DSC: 93.45. ## Original sources and implementation [Original paper]() [PDF]() [Implementation]()