TotalSegmentator is a self-configuring segmentation model built on the nnU-Net architecture, trained on a diverse dataset of whole-body CT images. It automatically delineates major organs, bones, vessels, and muscles without manual intervention, outputting precise pixel-level masks for each structure. The model handles the inherent variability in patient anatomy and imaging protocols through nnU-Net's automated preprocessing, topology adaptation, and post-processing pipeline, making it a reliable off-the-shelf tool for quantitative imaging workflows.
Glossary
TotalSegmentator
A pre-trained nnU-Net model capable of fully automatically segmenting over 100 anatomical structures in whole-body CT scans, widely used as a baseline for organ volumetry.
WHOLE-BODY CT SEGMENTATION
What is TotalSegmentator?
TotalSegmentator is a pre-trained deep learning model based on the nnU-Net framework that performs fully automated segmentation of over 100 anatomical structures in whole-body CT scans, serving as a robust baseline for organ volumetry and radiological analysis.
Widely adopted as a benchmark in medical image analysis, TotalSegmentator enables rapid organ volumetry, radiation dose planning, and population-level morphometric studies. It generates standardized DICOM Segmentation Objects compatible with clinical software, and its open-source availability has accelerated research in semantic segmentation, radiomics feature extraction, and federated learning across institutions. The model's consistent performance across heterogeneous CT acquisitions demonstrates the power of self-adapting deep learning frameworks in diagnostic support.
WHOLE-BODY CT SEGMENTATION
Key Features of TotalSegmentator
TotalSegmentator is a pre-trained nnU-Net model that performs fully automatic segmentation of over 100 anatomical structures in whole-body CT scans. It serves as a robust baseline for organ volumetry, surgical planning, and population-level imaging studies.
01
104 Anatomical Structures in a Single Pass
TotalSegmentator segments 104 distinct anatomical structures across the entire body from a single CT volume, including:
- Major organs: liver, spleen, kidneys, pancreas, lungs, heart
- Vascular structures: aorta, inferior vena cava, iliac arteries and veins
- Skeletal system: individual vertebrae (C1-L5), ribs, hip bones, sacrum
- Gastrointestinal tract: stomach, duodenum, small bowel, colon
- Brain substructures: gray matter, white matter, ventricles
This comprehensive coverage eliminates the need for multiple task-specific models, dramatically reducing inference pipeline complexity.
104
Structures Segmented
< 2 min
Inference Time (GPU)
03
Trained on a Large-Scale Multi-Institutional Dataset
The model was trained on 1,204 whole-body CT scans from multiple institutions, with ground truth labels generated through a combination of:
- Atlas-based segmentation propagation for initial label generation
- Manual correction by trained radiographers to refine boundary accuracy
- Multi-rater consensus review for challenging anatomical regions
The diverse training data spans varying scanner vendors, acquisition protocols, and patient populations, contributing to the model's strong generalization across clinical settings. Publicly available CT datasets were also incorporated to expand anatomical coverage.
1,204
Training CT Scans
Multi-site
Data Provenance
05
Robust Performance Across Anatomical Regions
TotalSegmentator achieves strong Dice similarity coefficients across diverse anatomical structures, with performance varying by organ complexity and boundary definition:
- High-performing organs (Dice > 0.90): lungs, liver, spleen, kidneys, heart
- Moderate performance (Dice 0.80–0.90): pancreas, adrenal glands, individual vertebrae
- Challenging structures (Dice 0.70–0.80): small bowel, duodenum, thin vascular branches
The model's performance degrades gracefully on structures with high anatomical variability or low soft-tissue contrast, providing transparent quality indicators for downstream clinical workflows. Quantitative validation was performed against expert radiologist annotations using both Dice score and Hausdorff distance metrics.
> 0.90
Dice on Major Organs
0.70–0.90
Dice on Challenging Structures
06
Subdivision into Task-Specific Variants
Beyond the full 104-structure model, TotalSegmentator offers specialized sub-models for focused clinical applications:
- TotalSegmentator MRI: adaptation for whole-body magnetic resonance imaging with Dixon sequences
- Lung lobe segmentation: fine-grained pulmonary lobe delineation for surgical planning
- Cardiac substructure segmentation: individual heart chambers and great vessels
- Vertebral body segmentation: individual vertebra labeling with anatomical numbering
- Body composition analysis: segmentation of adipose tissue compartments (subcutaneous, visceral, intramuscular)
These task-specific variants achieve higher accuracy on their target structures by reducing the label space and focusing model capacity.
TOTAL SEGMENTATOR
Frequently Asked Questions
Addressing common technical and clinical queries regarding the deployment and mechanics of the TotalSegmentator model for whole-body CT analysis.
TotalSegmentator is a pre-trained deep learning model based on the self-configuring nnU-Net framework designed to perform fully automatic segmentation of over 100 anatomical structures in whole-body CT scans. It works by ingesting a volumetric CT image and processing it through a convolutional neural network that has been trained on a diverse, multi-institutional dataset. The model predicts a pixel-level label map, assigning each voxel to a specific organ or bone class. Unlike manual segmentation, which can take hours, TotalSegmentator generates a complete whole-body segmentation in under a minute on modern GPU hardware, making it a robust baseline for organ volumetry and surgical planning.
Enabling Efficiency, Speed & Accuracy
Intelligent Analysis, Decision & Execution
We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.
Talk to Us
Search across company data
Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.
Useful when people spend too long searching or get different answers from different systems.
Enterprise searchRAGPermissions
Read more
Automate internal workflows
Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.
Useful when repetitive work moves across multiple tools and teams.
AI agentsWorkflow automationGovernance
Read more
Add AI to products and internal tools
Build assistants, guided actions, or decision support into the software your team or customers already use.
Useful when AI needs to be part of the product, not a separate tool.
AI integrationDecision supportModel routing
Read moreCOMPARATIVE ANALYSIS
TotalSegmentator vs. Other Segmentation Models
Feature and performance comparison of TotalSegmentator against general-purpose and medical-specific segmentation architectures for whole-body CT analysis.
| Feature | TotalSegmentator | nnU-Net (Custom) | SAM / MedSAM | Mask R-CNN |
|---|---|---|---|---|
Primary Task | Whole-body multi-organ semantic segmentation | User-configured segmentation (any task) | Promptable general-purpose segmentation | Instance segmentation with bounding boxes |
Pre-trained Organs (CT) | 104+ structures | 0 (requires training) | 0 (zero-shot, no CT specificity) | 0 (requires training) |
Self-Configuring Pipeline | ||||
Training-Free Deployment | ||||
3D Volumetric Native Support | ||||
Average Dice Score (Abdominal Organs) | 0.93 | 0.91–0.95 (task-dependent) | 0.78–0.85 (zero-shot) | 0.85–0.90 (trained) |
Inference Time (Full-Body CT) | < 60 sec (GPU) | Varies by configuration | < 10 sec (2D slices) |
|
DICOM Segmentation Object Output |
About the author
Prasad Kumkar
CEO & MD, Inference Systems
Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.
His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
How We Work
Custom AI workflows for your Business
One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us