Workflow Box

Workflow Box is an automated software system designed to route DICOM-compliant image data and generate anatomical contours using deformable image registration and machine learning algorithms for medical image processing. Its primary application is to streamline pre-processing steps in clinical workflows such as radiation therapy planning, with the crucial requirement that all automatically generated contours are subsequently reviewed and edited using separate visualization software by trained medical professionals.

Workflow Box is a software system designed for medical image processing, primarily focused on automating the routing of DICOM-compliant data and the generation of anatomical contours. It supports various imaging modalities, including CT, MR, and RTSTRUCT data, and includes components such as DLCExpert™, Embrace:CT™, Embrace:MR™, and Re:Contour™. The system employs advanced techniques such as deformable image registration for atlas-based contouring and re-contouring of patient images, alongside machine learning-based algorithms for automated contour delineation. It operates without direct user interaction during the contouring process and lacks its own image visualization interface, instead relying on an administration interface for workflow configuration and data routing rules within a hospital network. The generated contours are intended for use as inputs in clinical workflows, notably in radiation therapy treatment planning. Workflow Box must be used in conjunction with separate image visualization software, such as a treatment planning system, where trained medical professionals can review and edit the automatically generated contours. The system is not designed to automatically detect lesions. Its functionality is described as a subset of its predicate device, Mirada RTx, specifically excluding manual image manipulation and visualization tools but retaining core automated contouring technologies. Validation testing confirmed its performance, functional, and algorithmic capabilities meet user needs and its intended use, adhering to the DICOM standard.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. What specific machine learning models are employed for contouring in Workflow Box, and how are they trained and validated across diverse patient populations and image qualities?; How does Workflow Box handle challenging cases, such as images with significant artifacts or unusual patient anatomy, in its automated contouring processes?; What are the quantitative performance metrics (e.g., Dice similarity coefficient, Hausdorff distance) of Workflow Box's automated contours compared to expert manual contours for various anatomical sites and imaging modalities?; Can the automated workflows and data routing rules within Workflow Box be customized or adapted by institutions to suit specific clinical protocols or preferences beyond the standard configuration?; What mechanisms are in place to ensure seamless data integrity and traceability when routing DICOM data and contours between Workflow Box and external review or treatment planning systems?