RESEARCH INTEGRITY & SCIENTIFIC RIGOR

Evidence and limits.
In clinical context.

Review reported results, evaluation context, and the limits of each model.

Medical research workspace with brain scans on digital displays
Where medical imaging meets AI research.
MODEL EVALUATION REGISTRY

Reported benchmark metrics.

Team-reported metrics from the September 2026 model overview. See each model card for source limitations and evaluation scope.

ModelFamilyEvaluation MetricReported PerformanceClinical Status
Wrist Fracture DetectionOnyxOverall diagnosis AUROC0.974PREVIEW
ChestVisionOnyxMean AUROC across 14 conditions0.805PREVIEW
SightDxOpalQuadratic weighted kappa0.8965PILOT
BoneScan XOnyxInternal validation AUROC0.878PREVIEW
TB Audio ScreeningSonaPreliminary research accuracy~84%RESEARCH
Brain MRI ReconstructionFluxNot publishedNot publishedRESEARCH

AUROC and kappa measure different tasks. These results are not directly comparable.

STATISTICAL METHODOLOGY

Evaluation Metrics

AUROC evaluates sensitivity across all decision thresholds, avoiding threshold-tuning bias. Quadratic weighted kappa accounts for ordinal distances between diabetic retinopathy severity stages (0–4).

HOSPITAL PILOT STUDIES

Sachdeva Netralaya Validation

SightDx was evaluated in a team-reported multi-reader pilot at Sachdeva Netralaya, Jammu, measuring agreement with experienced ophthalmologists on fundus photographs.

OPEN-SOURCE BENCHMARKS

Stanford MURA Disclosures

BoneScan X was evaluated on Stanford's MURA musculoskeletal dataset, achieving 0.878 AUROC in team validation compared to the published 0.929 baseline. Independent multi-center validation details will be published with our technical report.

Original Stanford MURA Research Paper ↗
SYSTEM SAFETY PROTOCOL

Scope. Uncertainty. Review.

Three research priorities for model development and future healthcare deployment.

01 / BOUNDARIES

Explicit scope definitions

Each model has defined input eligibility criteria. The planned platform is designed to reject out-of-scope studies and return them for clinician review.

02 / UNCERTAINTY

Calibrated confidence

Planned outputs include calibrated confidence alongside findings, so ambiguous cases can be flagged for senior review.

03 / ATTRIBUTION

Concept heatmaps

Attention heatmaps are designed to show which image regions contributed to a finding, such as a fracture line or haemorrhage. They support review and do not prove causation.

Help evaluate medical AI.

Discuss Research Collaboration