The optimal group to review this topic is Computational Radiologists / Biomedical Engineers specializing in Volumetric Data Analysis.
Abstract:
This technical demonstration details the integration and application of TotalSegmentator, an open-source, artificial intelligence-powered segmentation tool, within the 3D Slicer platform for automated volumetric analysis of medical imaging data. Developed by Jacob Wassal and his team in 2022, TotalSegmentator enables the rapid, fully automated delineation of over 100 anatomical structures (including organs, bones, and muscles) from standard CT scans, significantly reducing the manual effort previously required. The high computational cost is emphasized, necessitating a robust hardware configuration, specifically recommending a powerful dedicated GPU for timely execution. The workflow involves installation through the 3D Slicer extension manager, execution of the ‘total’ segmentation task on DICOM inputs, visual refinement via smoothing, and mandatory post-segmentation classification (e.g., grouping structures into Thorax or Abdomen) to manage the large output volume effectively. The presentation notes that while highly effective for static structures, the tool currently exhibits limitations in accurately segmenting complex, highly variable structures such as blood vessels.
TotalSegmentator + 3D Slicer: Rapid, AI-Driven Volumetric Segmentation
0:04 Segmentation Revolution: TotalSegmentator is introduced as an AI-driven tool that has superseded previous manual and semi-automatic segmentation methods within 3D Slicer due to its fully automated capability.
0:32 Core Functionality: The tool, available since 2022, provides fully automated segmentation of over 100 anatomical structures, including organs, bones, muscles, and blood vessels, from a single CT scan.
0:44 Open Source Origin: The project is open-source, developed by Jacob Wassal and team at the Institute of Computational Biomedicine in Germany, benefiting from continuous community contributions.
1:04 Hardware Requirements: The tool is computationally demanding. Optimal performance requires a powerful multi-core CPU and a dedicated GPU (an NVIDIA RTX 4090 is cited for extremely fast runtime). Users with slower computers may use the 'fast' setting, potentially sacrificing accuracy.
1:47 Installation Protocol: Installation is conducted simply by accessing the 3D Slicer Extension Manager, searching for and installing 'TotalSegmentator,' and restarting the application.
2:14 Segmentation Workflow: After loading the CT scan, the user accesses the TotalSegmentator module, verifies the input volume, selects the 'total' segmentation task (or 'fast' for efficiency), and initiates the process by clicking 'Apply.'
2:53 Processing Speed: The demonstration confirms the tool executes the complex, full-body segmentation in approximately three minutes.
4:25 Post-Processing Visualization: Segmented results are rendered in 3D. The user is instructed to move to the Segment Editor module and increase the smoothing factor to enhance visibility and clarity of the segmented structures.
4:56 Data Management Necessity: Due to the large number of resulting segments (over 100), the creation of additional, classified segmentations (e.g., Bones and Muscles, Thorax, Abdomen) via the 'Copy and Move Segments' function is demonstrated as a practical method for organizing the output data.
7:35 Current Limitations: The tool currently does not handle blood vessel segmentation effectively. This limitation is attributed to the extreme anatomical variation of vascular structures, making the task significantly challenging for the AI model.