Release Gates
Research beta 0.2 / 31 October 2026
Suitable scope: clinician evaluation using teaching or authorized deidentified research data. Not a production medical device, clinical planning system or patient-record service.
| Area | This Release | Evidence or Remaining Gate |
|---|---|---|
| Teaching datasets | Two licensed CT cases, supplied masks and source attribution | Source checksums, mask/mesh transforms and display-grid checks; no independent anatomy review |
| DICOM import | Bounded local uncompressed little-endian, 16-bit, conventional single-frame CT | Known-value phantoms and negative cases; original scanner diversity not established |
| NIfTI import | Scalar 3D, axis-aligned qform/sform, finite supported data | Unspecified units disable distances; time series, oblique/sheared and unsupported data rejected |
| 3D / MPR | Surfaces, CT intensity MIP, linked planes, image-space annotations | Engineering registration and browser checks, not lesion-sizing or surgical-margin validation |
| Recovery | Asset integrity verification, bounded retries, graphics recovery and CT-only fallback | Fault-injection tests; no uptime SLA or exhaustive device coverage |
| Review files | In-memory sessions and passphrase-encrypted local review export | AES-GCM round-trip/tampering checks; no key recovery or durable patient-record backend |
| Activity | Local hash-linked operator history | Editable/exportable, not authenticated, signed or immutable clinical auditing |
| Clinical feedback | Structured research feedback and gated review stage | No clinician feedback is prefilled or claimed; human validation still required |
| AI segmentation | Not implemented | Bundled masks are dataset annotations; imported scans have no inferred masks |
| Hospital workflow | Not implemented | Institutional identity, roles, records, access policies, retention, integration and operational governance |
| Clinical release | Blocked | Defined intended use, risk management, representative reference data, clinical evaluation and applicable authorization |
Current Engineering Evidence
Machine-readable test evidence lists the checks actually run. A test count is not a clinical readiness percentage or a security certification. Clinical outcomes, independent penetration testing, regulatory authorization and representative scanner validation remain unevaluated.
Next Clinical-Development Gates
- Agree a narrowly defined intended use and hazard analysis with cardiothoracic and radiology advisers.
- Obtain lawful, appropriately governed, deidentified original DICOM studies and independently reviewed reference annotations.
- Validate source geometry, calibration, series selection, protocol/contrast coverage and segmentation before evaluating clinical utility.
- Establish institutional security, identity, permissions, audit, retention, backup and incident-response requirements for any patient-data service.
- Complete the applicable software-quality, usability, cybersecurity and jurisdiction-specific regulatory work before patient-care deployment.