Accelerating medical research with a sophisticated, human-in-the-loop toolkit for rapid and standardized annotations.
Automated download of PET/CT & MRI DICOM scans from XNAT servers and file sniffing.
Images are processed (NIFTI conversion) and fed into multiple AI models like TotalSegmentator & nnU-Net.
AI-generated labels (CT, T1/T2, Lesions) are combined with original scans to build a complete scene in 3D Slicer.
Clinicians review, refine, and approve the AI-suggested segmentations directly within the 3D Slicer interface.
Leverages a cascade of open-source models (TotalSegmentator, nnU-Net, etc.) to apply the best tool for each specific task, from rough outlines to fine-grained refinement.
This pipeline assists, not replaces, the human expert. It provides intelligent suggestions that users can quickly review, edit, and approve, ensuring clinical accuracy.
Seamlessly connects 3D Slicer to a high-performance external GPU server, enabling the use of complex, resource-intensive semi-supervised algorithms without burdening the local machine.
Built directly on the powerful and widely-used 3D Slicer platform, providing a familiar and accessible environment for clinicians and researchers without a steep learning curve.
Includes a suite of useful shortcuts and streamlined actions within 3D Slicer, minimizing clicks and further accelerating the review and annotation process for maximum efficiency.
By automating initial segmentation, the pipeline improves the standardization and reproducibility of annotations, a crucial requirement for clinical trials and large-scale research.
Dramatically speeds up the process of delineating anatomical structures and pathologies compared to manual methods.
Improves standardization and reproducibility, leading to higher quality annotations for clinical and research use.
Enhances the quality and efficiency of clinical workflows without compromising the expert's final authority and control.