1M+ healthcare records across X-ray, CT, MRI, pathology, dermatology, oncology, and prescriptions, with diagnostic reasoning and clinical notes providing symptom, disease, and side-effect grounding.
| Record | Modality | Source | Extracted finding | ICD-10 | Drug link | Status |
|---|---|---|---|---|---|---|
| R-1042 | MRI | PACS | L4-L5 disc herniation | M51.16 | linked | ready |
| R-1043 | Rx | EHR | metformin 500mg BID | E11.9 | linked | ready |
| R-1044 | Path | LIS | adenocarcinoma, grade 2 | C18.9 | linked | ready |
| R-1045 | Note | EHR | HbA1c 7.8% trending | E11.65 | - | Grounding... |
| R-1046 | CT | PACS | RLL nodule 6mm, stable | R91.1 | linked | ready |
| R-1047 | Rx | Scan | atorvastatin 20mg QHS | E78.5 | linked | ready |
| R-1048 | MRI | PACS | left meniscal tear | S83.242A | - | Grounding... |
| R-1049 | Path | LIS | chronic inflammation, mild | K63.89 | linked | ready |
| R-1050 | Note | EHR | shortness of breath on exertion | R06.02 | linked | ready |
| R-1051 | CT | PACS | hepatic steatosis, moderate | K76.0 | linked | ready |
| R-1052 | Rx | EHR | lisinopril 10mg daily | I10 | - | Grounding... |
| Capability | Prescription | Diagnostic | Reports | Drug data | Zstate Healthcare |
|---|---|---|---|---|---|
| Structured digitisation output | ✓ | ✕ | ~ | ✕ | ✓ |
| Clinical reasoning tasks | ✕ | ✓ | ~ | ~ | ✓ |
| Radiology interpretation | ✕ | ~ | ✓ | ✕ | ✓ |
| Pathology interpretation | ✕ | ~ | ✓ | ✕ | ✓ |
| Symptoms and disease context | ~ | ✓ | ~ | ✓ | ✓ |
| Side effects and drug metadata | ✕ | ~ | ✕ | ✓ | ✓ |
| Cross-task medication grounding | ~ | ~ | ~ | ✓ | ✓ |
A large healthcare dataset spanning X-ray, CT, MRI, pathology, dermatology, oncology, prescriptions, and diagnostic reasoning. Built to preserve real medical language, messy clinical inputs, and task-specific outputs.
A strong drug dataset tied to symptoms, diseases, side effects, and medication entities. This turns the core corpus from isolated task data into a more connected substrate for training grounded medical systems.
Imaging narratives grounded by specialists, connecting findings and impressions to the clinical picture.
Pathology report coverage for systems that need to reason over findings, impressions, and disease-linked clinical language.
Skin cases with specialist context, so models learn the visual and descriptive language dermatologists actually use.
Cancer care records where staging, course, and medication context have to stay connected to be useful.
Messy prescription inputs turned into structured medication signals, dosage understanding, and normalized entities.
Clinical reasoning that connects symptoms, disease hypotheses, and medication context the way specialists do.
Drug data linked to symptoms, diseases, and side effects that can ground the rest of the clinical corpus and complete the loop.
The corpus is modular. Teams can work with a single specialty or pull linked subsets spanning imaging, reasoning, and grounding into their training pipeline.
We can package a sample around X-ray, CT, MRI, pathology, dermatology, oncology, or prescriptions, plus the linked drug context layer, depending on what your training pipeline needs first.
