Roche Molecular Systems, Inc. · Class II · Cleared May 30, 2025
| K-number | K243863 |
| Device name | Opulus Lymphoma Precision |
| Applicant | Roche Molecular Systems, Inc. |
| Product code | QIH |
| Device class | Class II |
| Decision date | May 30, 2025 |
| Decision | Substantially Equivalent |
| Regulation | 892.2050 |
Opulus™ Lymphoma Precision is an AI/ML software tool that assists physicians in quantifying disease burden in patients already diagnosed with FDG-avid lymphomas. It automatically segments and visualizes lymphoma lesions in whole-body FDG-PET/CT scans and calculates total metabolic tumor volume (TMTV). A radiologist must review and make the final interpretation of the annotated images.
Both the subject device and predicate (NS-HGlio) are AI/ML algorithms that segment disease-related tracer/contrast uptake, generate volumetric measurements, and overlay segmentation masks on input images. The subject device processes FDG-PET/CT scans for lymphoma, while the predicate processes multi-sequence MRI for high-grade gliomas. Both produce reports with image overlays and were validated against ground truth established by specialty-trained expert readers.
IEC 62304:2006/AC:2015 (Medical device software – Software life cycle processes) and FDA Guidance 'Content of Premarket Submissions for Device Software Functions' (June 14, 2023). Performance was evaluated using absolute agreement metrics (cubic root transformation) and Dice Similarity Coefficient (DSC).
Both devices employ the same AI/ML methodology for automated segmentation and volumetric quantification of disease burden from radiological imaging, producing report and image overlay outputs validated against expert ground truth. Although they target different anatomical diseases (lymphoma vs. glioma) and imaging modalities (PET/CT vs. MRI), the core technological approach, validation methodology using reference standards from board-certified specialists, and output format establish substantial equivalence. The performance metrics demonstrate acceptable agreement with ground truth (mean DSC 0.70) comparable to predicate validation standards.
View the full FDA submission: accessdata.fda.gov