wasserth/TotalSegmentator
TotalSegmentator
TotalSegmentator is a medical image segmentation tool and AI-driven organ segmenter designed to isolate anatomical structures from CT scans. It functions as a deep learning anatomy parser and quantitative radiomics analyzer, providing a framework for identifying diverse body tissues and bones to create precise anatomical masks.
The system distinguishes itself through a comprehensive medical analysis suite that includes patient biometric estimation for demographics such as age, sex, weight, and height. It further provides specialized clinical index calculations and modality and phase classification to ensure appropriate processing of medical scans.
The project covers a broad capability surface including automated medical imaging workflow preprocessing, custom model training and evaluation pipelines, and quantitative anatomical analysis. It also provides utilities for anatomical body cropping, segmentation mask aggregation, and the generation of 3D segmentation previews for visual verification.
The tool supports offline image processing through local model weight management, enabling execution in air-gapped environments.
Features
- Medical Image Segmentations - Isolates organs and tissues in medical scans using neural networks to create precise anatomical masks for analysis.
- Medical Image Segmentations - Provides a comprehensive tool for isolating organs and anatomical structures from CT scans using deep learning models.
- Anatomy Parsers - Identifies and delineates diverse body tissues and bones for clinical analysis using neural networks.
- Automated Medical Imaging Pipelines - Automates medical imaging preprocessing by removing background noise and identifying contrast phases.
- High-Resolution Sub-segmentation - Runs specialized models on cropped image regions to extract high-resolution details for specific tissues.
- Organ Segmentation Frameworks - Provides a framework for automated multi-organ segmentation that supports custom model training and evaluation.
- Segment Quantitative Analysis - Computes volume and mean intensity from segmented masks to derive quantitative anatomical measurements and clinical indices.
- Segmentation Model Training - Provides frameworks for executing the training of instance segmentation models on custom medical datasets.
- Model Training Pipelines - Implements end-to-end workflows to convert medical imaging datasets and train custom organ segmentation models.
- Segmentation Evaluation Metrics - Calculates performance metrics by comparing predicted anatomical labels against ground truth medical data.
- Biometric Predictions - TotalAI estimates weight, size, age, sex, and mass indices from medical scans.
- Patient Demographic Analysis - Predicting patient age, sex, weight, and height directly from medical imaging data using automated anatomical analysis.
- Medical Modality Classification - Analyzes median intensity of key structures to determine imaging modality and contrast phase before processing.
- Radiomics Feature Extraction - TotalAI computes volume, mean intensity, and radiomics features to quantify anatomical characteristics.
- Demographic Estimation - Uses neural networks to predict human attributes such as age, gender, weight, and height from imaging data.
- Inference Performance Optimization - Optimizes inference by using a low-resolution model to isolate regions of interest before high-resolution processing.
- Cascaded Resolution Inference - Optimizes memory and compute by using a low-resolution model to locate target anatomical areas before high-resolution processing.
- Multi-Stage Sub-segmentations - Runs specialized models on cropped image regions to extract high-resolution details for specific tissues.
- Medical Imaging Training Frameworks - Provides a pipeline for developing and evaluating new segmentation models using specialized medical datasets.
- Segmentation Visualizations - Creates 3D renderings of identified anatomical classes for rapid visual verification of segmentation accuracy.
- Medical Imaging and Analysis - Classifies imaging modalities and contrast phases to ensure the correct segmentation pipeline is applied.
- Anatomical Statistics Reports - Produces detailed reports containing runtime statistics, output file lists, and anatomical voxel counts.
- Health Metric Calculators - Computes health-related indices, such as mass and surface area, based on predicted patient biometrics.
- Anatomical Image Cropping - Removes background noise and non-body areas from scans to focus processing on patient anatomy.
- Contrast Phase Identification - Analyzes scans to determine which contrast phase was used during image acquisition.
- Cardiac Segmentation - Delineates heart chambers in high-resolution scans using both contrast and non-contrast data.
- Medical Imaging Modality Identification - Determines scan types and imaging modalities by analyzing the characteristics of the image.
- RF Phase Analysis - Classifies images into specific contrast phases by analyzing the median intensity of anatomical structures.
- Clinical Index Calculations - TotalAI computes a specific ventricular enlargement index from skull images.
- Clinical Measurement Tracking - Computes quantitative clinical measurements such as vessel diameter and ventricular enlargement indices.
