Software as a Medical Device
OpticFlow™ v2.1
Automated screening software for detection of glaucoma suspect features in color fundus photographs of adults aged 40 and above.
Intended purpose
OpticFlow™ is intended to be used by trained healthcare professionals (ophthalmologists, optometrists, orthoptists, and qualified screening technicians) as a decision-support tool in screening programmes and community eye care settings.
The software analyses one or two macula-centred color fundus images per eye and provides an automated refer or no refer recommendation for further glaucoma assessment based on optic nerve head morphology — including cup-to-disc ratio, disc haemorrhages, and rim notching patterns.
Important: OpticFlow™ does not provide a diagnosis of glaucoma. All refer outputs require clinical examination including tonometry and visual field testing where indicated. Not intended for paediatric or non-adult populations.
How it works
- Image acquisition Standard 45° or 30° color fundus camera. Minimum resolution 2048×1536 px. JPEG or DICOM input.
- Quality assessment Automated QC checks for focus, illumination, and field of view before inference.
- AI inference Ensemble of convolutional neural networks (EfficientNet-V2 backbone) produce probability scores and Grad-CAM heatmaps.
- Clinical output Refer / no refer result with confidence score, exported to EMR or printable PDF report.
Technical specifications
| Target regulatory class | EU MDR Class IIa (Rule 11) — submission in preparation |
|---|---|
| Software version | OpticFlow™ 2.1.4 (build 20260214) |
| Input | Color fundus photograph (DICOM Ophthalmic Photography or JPEG) |
| Output | Refer / No refer + confidence + heatmap overlay |
| Inference time | < 4 s per bilateral study (GPU) · < 12 s (CPU) |
| Deployment | On-premise (Docker/Kubernetes) or Retinotech Cloud (EU-West) |
| Integrations | DICOM C-STORE/C-FIND, HL7 FHIR R4, REST API |
| Supported OS | Linux (Ubuntu 22.04+), Windows Server 2019+ |
| Languages | Polish, English, Czech, German (UI) |
Comparison with general retinal AI screeners
| Capability | OpticFlow™ | General DR screeners |
|---|---|---|
| Primary indication | Glaucoma suspect triage | Diabetic retinopathy |
| Optic disc C/D analysis | Dedicated models | Secondary / limited |
| EGS-aligned refer criteria | Yes | No |
| Non-diabetic populations | Intended use | Out of scope |
| Explainability heatmaps | Per-inference | Varies |
| Regulatory clearance | Pathway in progress | Varies by product |