Teramisu

An Intuitive, End-to-End Platform for AI-Powered Medical Image Segmentation

Empowering medical researchers and clinicians, regardless of coding expertise, to build custom segmentation models, conduct insightful analyses, and create 3D visualizations.

Download Teramisu Explore Features

About Teramisu

The Challenge: Identifying specific structures in medical images like MRIs or CTs is vital for diagnosis, treatment planning, and research. While AI has revolutionized this field by enabling high throughput, automated segmentation, developing these segmentation models typically requires significant programming skills, creating a barrier for many domain experts.

The Solution: Teramisu (Toolkit for Robust Medical Image Segmentation with U-Net) is a standalone desktop application that breaks down these barriers. It provides a seamless, user-friendly graphical interface that guides users through the entire lifecycle of a segmentation project: from raw image data to trained AI models, rigorous testing, insightful performance analysis, and compelling 3D visualizations. Teramisu demystifies complex deep learning workflows, making advanced image analysis accessible and efficient for everyone, regardless of coding background.

Its integrated, no-code approach offers a unique, all-in-one solution, significantly accelerating research and development in medical imaging.

Key Features

Teramisu simplifies the complex, end-to-end AI segmentation pipeline into manageable, user-friendly modules.

Intuitive Data Annotation

Easily load NIfTI or DICOM images and create segmentation masks slice by slice using a variety of features. Advanced tools include a smart adaptive brush with a live preview for more efficient annotation.

Simplified Model Training

Creating a custom U-Net++ model is as simple as importing training images and their corresponding segmentation masks.

Advanced Customization

Modify every parameter in the model construction process, including backbone type, number of epochs, and loss function weights, all via an intuitive interface.

Robust Model Testing

Load trained models, select NIfTI files, and perform automated segmentation. Visualize the output segmentations directly in the app, and quantitatively evaluate model performance through a variety of statistical tools.

Interactive 3D Reconstruction

Generate, evaluate, and download 3D surface meshes from 2D segmentations. Dual-view panels allow for side-by-side comparison.

Truly No-Code Operation

All functionalities are purely GUI-driven. Complex algorithms are abstracted into user-friendly controls with constant visual feedback and guided workflows.

Your End-to-End Segmentation Workflow

1

Data Preprocessing

Convert raw images into formats suitable for machine learning model development by adjusting their size and normalizing pixel values.

2

Build AI Model

Create segmentation masks directly in the app using our comprehensive annotation tools. Customize the training parameters, and let Teramisu train your U-Net++ model.

3

Test & Validate

Load your trained model. Input new NIfTI files and see automated segmentations. Utilize test-time augmentation for enhanced accuracy and visualize results instantly.

4

Assess Performance

Rigorously evaluate your model's output against ground truth masks. Calculate various 2D and 3D metrics like Dice, IoU, and Hausdorff Distance, visualized in comprehensive plots.

5

Reconstruct in 3D

Transform your 2D segmentations into interactive 3D surface models. Customize appearance, compare reconstructions, and save meshes for presentations or 3D printing.

Get Started with Teramisu

Download the Teramisu toolkit today and revolutionize your medical image segmentation workflow.

Download Teramisu.exe

Current Version: 1.1.3 | For Windows