Syngo.Via RT Image Suite

The syngo.via RT Image Suite (SOMARIS/8 VB40) is a medical image analysis software designed to enhance radiation therapy planning through multi-modality image visualization, manipulation, and advanced segmentation of tumors and organs-at-risk. It employs sophisticated image processing, including a novel deep learning-based approach for automatic contouring, to improve workflow efficiency and the accuracy of preparing and assessing radiotherapy treatments.

The syngo.via RT Image Suite, software version SOMARIS/8 VB40, is an advanced image analysis software designed for radiation therapy planning and response assessment. Its primary purpose is to assist medical professionals in viewing, manipulating, and visualizing 3D and 4D medical images from multiple modalities, including CT, PET, MRI, and CBCT. The software facilitates the crucial tasks of segmenting tumors and organs-at-risk (OARs), creating and modifying treatment contours, and managing points of interest such as isocenters. It supports various image processing techniques like multiplanar reconstruction (MPR), minimum intensity projection (MIP), and volume rendering technique (VRT), and allows for both manual and semi-automatic contouring, image registration (rigid and deformable), and multi-modality image fusion. The suite is also capable of generating synthetic CT images from MR acquisitions and managing simple treatment plans, with results exportable to dedicated Treatment Planning Systems. The SOMARIS/8 VB40 version introduces several enhancements over its predicate, focusing on improved workflow and advanced capabilities. Key modifications include a more intuitive user interface for beam placement (now supporting rectangular block fields), simplified reference point management with streamlined breast iso-centering, and improved patient marking workflows. Contouring tools have been upgraded with features like contour interpolation, 3D pan/scale, new shape options, and an improved nudge tool. Significantly, the advanced contouring capabilities now extend to automatic segmentation of thoracic and abdominal regions, powered by a new deep learning-based approach utilizing an adversarial network. Clinical performance data demonstrated that this new version offers improved OAR segmentation, with higher detection rates (100% for evaluated organs like Brain, Liver, Kidneys) and enhanced segmentation quality, as measured by improved Average Symmetric Surface Distance (ASSD) and DICE coefficients compared to manual annotations and the previous software version.

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. How does the new deep learning-based adversarial network for automatic contouring in thoracic and abdominal regions compare in terms of accuracy and processing time to other established auto-segmentation algorithms for these specific anatomical sites?; What specific anatomical landmarks are utilized by the 'region of interest detection' component of the OAR segmentation algorithm, and how robust is this landmark detection across diverse patient anatomies and image qualities?; The document mentions improved Average Symmetric Surface Distance (ASSD) and DICE coefficients for OAR segmentation; what were the specific values for critical structures like the spinal cord or lungs, and what was the inter-observer variability in the manual ground truth used for comparison?; Could the 'simplified workflow for breast iso-centering' be detailed, particularly how it identifies breast boundaries and suggests an isocenter, and how does its accuracy compare to conventional manual methods?; Regarding the generation of synthetic CT from multiple MR acquisitions, what specific MR sequences are required, what is the underlying methodology (e.g., atlas-based, deep learning), and how does the Hounsfield unit accuracy of the synthetic CT compare to actual CT scans for radiotherapy dose calculation purposes?