AVIEW RT ACS

AVIEW RT ACS is a medical image processing system utilizing deep learning technology for automated segmentation of organs at risk from CT scans, designed to support radiotherapy planning. Its efficacy and generalization capabilities were established through rigorous performance testing against expert-derived gold standards and a predicate device, using metrics like Dice Similarity Coefficient and Hausdorff Distance on diverse patient datasets and imaging parameters.

AVIEW RT ACS, by Coreline Soft Co., Ltd., is a medical image processing system designed for auto-segmentation of organs using deep learning. It processes CT images (non-contrast and contrast) to assist in radiotherapy planning by delineating structures such as the lungs, heart, eyes, brain, mandible, kidneys, liver, femurs, and bladder. The system is intended for use by trained medical professionals, including radiation oncology physicians. The device's auto-segmentation performance was validated through a comparative study against a predicate device and a robust gold standard. This gold standard was meticulously developed by three radiation oncology physicians through a sequential editing process. The validation utilized a dataset of 120 CT cases, including both Korean and U.S. populations with diverse ethnic backgrounds. Evaluation metrics included the Dice Similarity Coefficient (DSC) and 95% Hausdorff Distance (HD), with sub-group analyses performed for different CT scanner vendors, slice thicknesses, and reconstruction kernels to assess the robustness and generalization of the segmentation algorithms across varied conditions.

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 deep learning architecture is employed in AVIEW RT ACS for organ segmentation, and how was it trained (e.g., dataset size for training, diversity of training data, annotation process for training data)?; How do the quantitative DSC and 95% HD values for AVIEW RT ACS on specific organs compare to inter-observer variability among human experts segmenting the same structures?; What is the typical processing time for auto-segmentation of a full CT dataset for each anatomical region (e.g., head & neck, pelvis), and what are the computational requirements (e.g., GPU, CPU, memory)?; Were there any specific organ types or image characteristics (e.g., post-contrast artifacts, presence of metallic implants, low image quality) where the auto-segmentation performance was notably degraded or required more significant manual correction?; Beyond the listed vendors, slice thicknesses, and kernels, what other factors of image acquisition or patient variability were assessed for their impact on segmentation performance?