Scanner-side inference for urgent findings
Deploy Vision directly at the imaging edge using Holoscan for low-latency triage of critical findings — stroke, PE, pneumothorax — before images reach PACS.
Imaging AI
Mergic Vision
Real-time GPU radiology and pathology AI for detection and segmentation across CT, MRI, X-ray, and whole-slide images.
Product capabilities
Real-time detection and segmentation across every imaging modality.
Mergic Vision connects scanners, PACS, and workstations into a GPU-accelerated detection loop with sub-second inference.
MONAI segmentation workflows
Pre-trained and fine-tunable models for organs, lesions, and pathology regions with validated concordance.
Cross-modal context with Atlas
Link imaging findings to genomics, pathology, and EHR timelines for richer decision support.
Priority queue routing
Surface urgent findings at the top of radiologist work queues with confidence scores and evidence overlays.
Governed deployment via Citadel
VPC or on-prem with audit trails, model versioning, and PHI boundaries.
Continuous learning loop
Radiologist feedback flows back into Atlas for model retraining and performance compounding.
NVIDIA healthcare stack
Built natively on Clara, MONAI, and Holoscan.
NVIDIA Clara
GPU-accelerated medical imaging pipelines for detection, segmentation, and reconstruction.
MONAI
Open-source deep learning framework for medical imaging with pre-trained model zoo.
Holoscan
Low-latency sensor processing for operating rooms, imaging suites, and edge inference.
Performance metrics
Vision by the numbers.
Sub-second inference for urgent CT findings with Holoscan edge deployment
Sensitivity on stroke, PE, and pneumothorax detection tasks with validated models
Imaging modalities supported: CT, MRI, X-ray, ultrasound, pathology slides
Radiologist capacity increase without new hires via priority queue and triage automation
Audit trail coverage for model predictions, escalations, and human signoff
Use cases
Where Vision fits in your workflow.
Emergency triage for time-sensitive findings
Surface stroke, PE, pneumothorax, and acute fractures at the top of radiologist queues with confidence scores and anatomical overlays. Reduce time-to-read for critical cases from hours to minutes.
Tumor detection and measurement
Segment lesions across CT, MRI, and PET for RECIST measurement and treatment monitoring.
Whole-slide pathology screening
Detect tumor regions, quantify biomarkers, and prioritize slides for pathologist review.
Remote reading with GPU-backed triage
Distribute imaging workload across distributed radiologists with AI-powered urgency scoring and automatic case assignment. Integrate with existing PACS and RIS infrastructure via DICOM and HL7.
Imaging biomarker extraction
Quantify imaging endpoints for trial cohorts with reproducible, GPU-accelerated pipelines.
How it works
From scanner to signoff in seconds.
Vision connects to your imaging infrastructure and applies GPU-accelerated models at every stage of the diagnostic pipeline.
Ingest
Pull DICOM from scanners, PACS, or VNA via Holoscan edge or API
Inference
Run MONAI models on NVIDIA GPUs for detection, segmentation, classification
Triage
Route urgent findings to priority queues with confidence scores
Context
Link to Atlas for genomics, pathology, EHR timeline context
Review
Radiologist confirms, edits, or escalates via workbench or PACS overlay
Learn
Feedback flows to Atlas for model retraining and performance improvement
Integration
Vision connects to your existing imaging stack.
DICOM ingest and push
Bi-directional DICOM connectivity to every major PACS vendor
Overlay and worklist APIs
Inject findings into radiologist reading workflows via HL7, FHIR, or REST
Results delivery
Push structured reports and critical alerts to Epic, Cerner, or FHIR endpoints
Related products
Compose Vision with the rest of the Mergic platform.
Mergic Helix
Whole genomes from a day to under an hour with Parabricks-backed secondary analysis.
Mergic Forge
Generative AI for protein structure and molecular design with BioNeMo.
Mergic Lumen
The clinical-decision agent that sees across modalities — imaging, genomics, EHR.
Mergic Citadel
HIPAA-compliant deployment with full audit, model versioning, and PHI boundaries.
Mergic Atlas
The consented, de-identified data layer that powers the Mergic flywheel.
FAQ
Questions before your demo.
Answers for radiology, IT, and compliance teams evaluating Vision.
Vision provides AI-assisted detection and triage for radiologist review. It is not a standalone diagnostic device. Final interpretation and signoff remain with licensed clinicians. Regulatory classification depends on deployment scope and intended use.
CT, MRI, X-ray, ultrasound, mammography, PET, and whole-slide pathology images. We support DICOM ingest and can fine-tune models for custom anatomies or protocols.
Yes. Vision runs in VPC, on-prem DGX, or hybrid configurations with full audit trails and PHI boundaries via Mergic Citadel.
Bi-directional DICOM connectivity, HL7 worklist integration, and REST APIs for overlay and structured reporting. We support all major PACS vendors.
Get started
See Vision in action with your imaging workflows.
Bring a radiology or pathology use case. We'll map the GPU architecture, PACS integration, and first measurable outcome.