Three products, one clinical standard
Clinical AI for Indonesian hospitals
Imaging, documentation and quality assurance — three tools clinicians use every day. Data storage and LLM inference both sit in the Jakarta Region. Nothing enters the record without a clinician's signature.
Segmentation complete
Lung
- Organ
- Lung
- Window WW / WL
- 1500 / −600
- Threshold
- 0.50
Segmentation complete
- 3
- Organ models
- 14
- Note templates
- 7
- Finding types
- ap-southeast-3
- Jakarta Region
Lung, colon, liver
SOAP, ED, obstetrics, paediatrics, psychiatry
Typos through clinical context
Data and LLM inference
Products
One tool for each weak point in the workflow.
All three stand alone and can be adopted separately. What they share is one rule: the clinician signs, not the model.
Eden
ImagingDICOM viewer with automatic lung, colon and liver segmentation.
- Upload a DICOM series or a NIfTI file — UNet segmentation runs server-side and comes back as a probability map.
- Axial, coronal and sagittal views with linked crosshairs and per-organ window presets.
- Segmentation overlay with an adjustable threshold, so the clinician sees what every contour rests on.
EdenScribe
DocumentationAI medical scribe: listens to the consultation, drafts the clinical note.
- Ambient Indonesian or English transcription with speaker labels and partial results.
- 14 note templates, referral letters, sick notes, discharge summaries and ICD-10 suggestions.
- Patient consent gate before recording, per-facility audio retention, and an append-only audit log.
Eden QA
Quality assuranceRadiology report proofreader — catches mistakes before the report goes out.
- Seven finding types: typo, laterality, inconsistency, measurement, terminology, clinical and patient context.
- OCR for photos, scans, PDF and Word — processed in memory, nothing is stored.
- Learns from rejections: a suggestion rejected with a reason becomes a house rule for the next report.
Data Residency
Where your data actually sits.
Compliance is not a badge in the footer. Data storage and model inference run in the Jakarta Region — because your DPO is going to ask.
Stored in Jakarta
Database, audio recordings, user identities and logs live in ap-southeast-3. The 25-year record retention requirement is honoured through soft deletes.
LLM inference in Jakarta
Kimi K2.5 on Amazon Bedrock is served in-Region: the request is signed, accepted and executed in Jakarta. Clinical text never crosses a border to be processed by the model.
See all three running on your own cases.
We run the demo on sample data, then work through the DPIA, retention policy and integration with the systems your facility already runs.
admin@edensehat.id