RUS

RUS is a medical imaging software designed to assist clinicians in treatment and surgical planning by processing iodine contrast-enhanced abdomen CT images to generate interactive 3D anatomical models. It leverages DICOM standards, a modular software architecture (h-Server, h-Space, RUS Stomach Planning), and machine learning algorithms for image segmentation and analysis, providing advanced visualization and measurement tools to support clinical decision-making.

RUS is a medical imaging software suite intended to provide trained medical professionals with tools to aid in reading, interpreting, reporting, and treatment planning for patients using iodine contrast-enhanced abdomen CT images. The system processes DICOM compliant images and offers functionalities such as 2D and 3D viewing, volume and surface rendering, image fusion, measurements, reporting, and image management. A key feature is its capability for surgical planning via 3D modeling derived from patient CT data, which serves as an auxiliary tool for clinicians who retain sole responsibility for final patient management decisions. The software is not intended for the primary diagnostic interpretation of mammography images. The RUS system architecture includes three main components: h-Server for data management including anonymization and pseudonymization of DICOM images; h-Space for 3D reconstruction, evaluation of patient anatomy, and creation of 3D models and anatomical atlases; and RUS Stomach Planning for utilizing these models in therapy or surgical planning. The workflow involves transferring patient CT data from PACS to the h-Server, processing in h-Space to generate a 'hu3D model' (typically within 72 hours), which is then used in RUS Stomach Planning. The software employs machine learning models for specific tasks: CADD U-NET for organ segmentation, 3D U-NET for vessel segmentation, and linear regression for pneumoperitoneum detection. Performance validation on 60 CT studies showed Dice Similarity Coefficients of 0.927 for organ segmentation and 0.920 for vessel segmentation, and a Mean Absolute Error of ±0.972 mm for pneumoperitoneum detection. Length measurements were validated to be accurate within a mean difference of ±10%.

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 72-hour turnaround time for 'hu3D model' generation impact its utility in urgent surgical planning scenarios?; What is the performance of the organ and vessel segmentation algorithms on CT scans with significant motion artifacts or atypical contrast enhancement patterns?; Beyond DICOM compliance, what are the specific technical interoperability capabilities of RUS with diverse Picture Archiving and Communication Systems (PACS) and Electronic Health Record (EHR) systems?; How does the system's pseudonymization and anonymization process for patient data in h-Server and h-Space ensure compliance with various data privacy regulations while maintaining data integrity for model generation?; What were the specific characteristics of the 60 imaging studies used for machine learning model validation, particularly concerning patient demographics, pathological diversity, and CT scanner variations, and how do these reflect the intended use population?