GitHub Repository
Created On
Updated On
SAM-Med2D is a specialized adaptation of the Segment Anything Model (SAM) specifically fine-tuned for medical images. This advanced tool has been rigorously trained on over 4.6 million images, encompassing 10 distinct medical imaging modalities such as CT, X-Ray, and Ultrasound, 31 major organs, and their corresponding anatomical structures. With its impressive dataset, SAM-Med2D sets itself apart in the field of medical image segmentation with its remarkable accuracy and user-friendly interface. Designed for both seasoned professionals and novices, the tool simplifies the segmentation process. Users can effortlessly upload images directly or via URLs and navigate through the segmentation process with ease, choosing from modes that include point specification or bounding boxes to accurately delineate areas of interest. SAM-Med2D's performance is excellent, showcasing a significant leap in accuracy on the MICCAI2023 dataset compared to the base SAM model. For instance, when a user marks just one spot on the image to mark the area of interest, SAM-Med2D achieves 83.41% accuracy, a stark contrast to the 48.08% accuracy achieved by SAM. This leap in performance underscores the tool's capability to interpret user inputs with precision and provide detailed segmentations that enhance medical analysis and diagnostic processes. With SAM-Med2D, medical image analysis is more accessible, precise, and efficient, offering an unparalleled tool in the advancement of medical diagnostics.

More details

Use Cases Limitations Evidence Owner's Insight
  • Clinical Diagnosis Enhancement: SAM-Med2D equips clinicians with a powerful image segmentation tool that expedites the identification of critical regions within diverse medical imaging formats. By streamlining this aspect of the diagnostic workflow, the tool assists in pinpointing areas of interest for further examination, potentially reducing diagnostic times and increasing accuracy.
  • Research Analysis: SAM-Med2D is an invaluable asset for researchers specializing in medical imaging. Its segmentation abilities can help researchers focus on specific parts of interest in the image, facilitating medical image analysis and the discovery of new medical insights.
  • Educational Demonstration: SAM-Med2D can be integrated into medical education to enhance the learning experience by providing interactive, practical experience with medical imaging and the segmentation process. This helps to prepare students for both the technical and analytical aspects of medical diagnostics.

The model's accuracy may vary across different imaging modalities and anatomical structures. For more information, see the app’s “Model Information” page.SAM-Med2D cannot directly segment 3D images, although it can segment 2D slices of them. Currently, it only accepts “.png”, “.jpg”, and “.jpeg” images.The tool should be used as a supplementary aid and not as a definitive source for medical diagnosis, as it cannot replace expert clinical judgment.

This application is based on the model outlined in the SAM-Med2D paper ( and the associated GitHub repository ( The model’s performance has been validated on a dataset of over 4.6 million medical images with a diverse range of imaging modalities and anatomical structures.The SAM-Med2D model (with the adapter layer dropped during the test phase) has demonstrated robust generalization capabilities on the MICCAI2023 datasets, consistently outperforming the base SAM model and scoring a weighted average Dice score of 90.12%.  For more information, see the “Model Information” page after launching the application.

I'm Venkata Chengalvala, the primary developer of this app. As an AI consultant for Health Universe, I port high-quality peer-reviewed AI models helpful to clinicians, patients, and/or researchers to Health Universe's platform. I have a Bachelor of Science in Molecular, Cellular, and Developmental Biology (MCDB) and Computer Science from the University of Michigan-Ann Arbor. During my education, I engaged in medical research, co-authoring a literature review on tumor-derived exosomes, cited by 30+ people so far:

At Health Universe, we prioritize the development of tools that are not just at the forefront of technology but are also intuitive and accessible to all users, ranging from seasoned medical professionals to those still in training. While the current focus is on maintaining the high standard of SAM-Med2D, I’m open to enhancing the application based on your feedback. If you have any feedback or thoughts about the app, feel free to leave a comment below. While my current focus is on maintaining the high standard of SAM-Med2D, I’m open to adding new features or otherwise enhancing the app based on your suggestions.

Peer reviewed

Warning: This application or model has been peer reviewed, but still may occasionally produce unsafe outputs.

  • Favorites: 7
  • Executions: 237

  • Diagnostics & Imaging


Venkata Chengalvala

Member since