uOmnispace.MI

uOmnispace.MI is a comprehensive medical image post-processing software platform designed to enhance diagnostic capabilities by enabling advanced viewing, analysis, and quantification of multi-modality images (PET, CT, MR, SPECT) across oncology, neurology, and cardiology. It employs sophisticated image processing techniques, including deep learning-based automated anatomical labeling and specialized modules for lesion segmentation, image fusion, and dynamic functional assessment, to support detailed clinical evaluation and reporting.

uOmnispace.MI is an advanced medical image post-processing software solution designed for comprehensive viewing, manipulation, evaluation, and analysis of PET, CT, MR, and SPECT images within healthcare settings. It integrates several specialized applications to support clinical decision-making across various medical disciplines. These include MM Fusion for multi-modality image fusion and reporting in oncology, neurology, and cardiology; MM Oncology for follow-up analysis of cancer patients, featuring image registration, lesion segmentation, and statistical tools like PERCIST and RECIST; and Dynamic Analysis for evaluating PET data with anatomical correlation, including lesion segmentation and time-activity curve generation. The platform also supports third-party applications like NeuroQ for quantitative brain PET analysis, including amyloid uptake assessment, and Emory Cardiac Toolbox for detailed cardiac image reconstruction, perfusion, and functional analysis. The software leverages sophisticated image processing techniques, including manual and automatic image registration, various lesion segmentation methods (fixed threshold, percentage threshold, adaptive threshold), and 3D visualization. A key technological advancement highlighted is the integration of deep learning-based algorithms for automated spine and rib labeling on CT images, which are optimized from previous versions. Performance validation of these deep learning algorithms demonstrated high accuracy: the spine labeling algorithm achieved an average score of 4.951 out of 5 (exceeding a 92% mean identification rate equivalent) on 286 CT scans, and the rib labeling algorithm achieved a perfect average score of 5 out of 5 (exceeding a 91.7% mean identification rate equivalent) on 160 CT scans. Both algorithms showed robust performance and good generalization across diverse patient demographics and imaging equipment.

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 deep learning-based spine and rib labeling algorithms compare when applied to images with significant anatomical variants or pathological conditions not extensively represented in the testing dataset?; What is the specific impact of the "optimized layout protocols" in the MM Fusion application on workflow efficiency and diagnostic accuracy in complex oncology, neurology, and cardiology cases compared to standard layouts?; Beyond the listed PERCIST and RECIST criteria, does the MM Oncology module support, or plan to support, other emerging quantitative oncologic response criteria (e.g., iRECIST, mRECIST for specific cancers)?; For the NeuroQ application, what is the source and demographic composition of the "AC database" used for normal brain region comparisons, and how does this affect its applicability across diverse patient populations?; Considering the integration of multiple advanced applications, what are the computational requirements and processing times for typical complex cases, such as a whole-body PET/CT oncology follow-up with multiple lesions requiring segmentation and quantitative analysis?