Ceevra Reveal 3

Ceevra Reveal 3 is a medical imaging system that processes CT and MR scans to create interactive 3D visualizations for preoperative surgical planning and intraoperative display. It employs machine learning and computer vision algorithms to generate preliminary segmentations of normal anatomy in adult patients, with performance validated by quantitative metrics against expert segmentations across diverse datasets.

Ceevra Reveal 3 is a software-based medical imaging system designed to enhance surgical planning and intraoperative visualization. Its primary function is to process multi-dimensional digital images acquired from CT or MR scans, allowing for comprehensive review, analysis, communication, and media interchange. The system facilitates the generation of three-dimensional (3D) images from patient scans, which clinicians can view and interact with via dedicated mobile and desktop applications, or on external displays including virtual reality (VR) headsets. Key interactive features include rotating, zooming, panning, and selectively displaying or hiding individual anatomical structures. Ceevra Reveal 3 is intended for use by healthcare professionals to assist in preoperative surgical planning and for the intraoperative display of these 3D images, supporting clinicians in making final patient management decisions. A core component of Ceevra Reveal 3 is its use of machine learning and computer vision algorithms to generate preliminary segmentations of normal anatomy, specifically for adult patients aged 22 and over; for younger or unknown-age patients, 3D images are created without employing these machine learning algorithms. The system incorporates four distinct machine learning models, which were verified using a dataset of 141 actual CT or MR imaging studies, ensuring independence between training and testing data at the institutional level. Performance evaluation focused on diverse patient demographics and various scanner manufacturers. The accuracy of the ML-generated segmentations was quantified by comparing them against segmentations performed by medical professionals, using metrics such as the Sørensen–Dice coefficient (DSC) and the Hausdorff distance at the 95th percentile (HD-95). Reported DSC values for structures like prostate, bladder, kidney, arteries, and veins ranged from 0.81 to 0.90, while neurovascular bundles showed a 7.8 mm HD-95.

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 performance of the machine learning segmentation, as indicated by DSC and HD-95 values, correlate with improvements in surgical planning efficiency or clinical outcomes?; What are the specific anatomical structures, beyond those listed with performance metrics, that Ceevra Reveal 3 can segment using its machine learning models, and what is their validated accuracy?; How is the 'preliminary' nature of the ML-generated segmentations communicated to clinicians to ensure they maintain final decision-making authority, and what is the typical workflow for reviewing and potentially modifying these segmentations?; What are the computational resource requirements and average processing times for generating 3D models from CT/MR scans, both with and without the machine learning algorithms, particularly for complex anatomical regions?; Considering the age restriction (22+ years) for machine learning algorithm use, what are the primary differences in the 3D image generation process and resulting visualization quality for pediatric patients or those of unknown age, and are there plans to validate the ML models for younger populations?