Natural-language case-finding over 20+ years of NHS pathology
Describe the patients. We find them.
Ask in plain English. hepatoSIGHT compiles your intent into an auditable, editable structured query, then scans synthetic test results across patients — powered by precomputed derived metrics, not a black box.
Try:
Saved searches
Favourites and recent cohorts — click to load into the builder and run live.
hepatoSIGHT · query planner
Illustrative of the tool calls the assistant makes
()
Interpreted intent
Compiled structured query
Plain language in → auditable, editable rules out. The rules execute — not the text.
Match rows on
Mode
Nested group ·
—
—
edit any cell → re-runs the same optimized SQL, no NL round-trip
Summary
Cohort
top by
| NHS # | Age | Sex | Peak ALP | AMA | FIB-4 | CLDI | Risk |
|---|---|---|---|---|---|---|---|
| No patients match this cohort. | |||||||
Risk stratification
Live histogram over the cohort — drag the threshold.
High-risk threshold
patients ≥ threshold
high-risk band
Ingest a new county in hours, not weeks.
Derived metrics — AUC, trends, FIB-4, CLDI — are precomputed in parallel at ingest, so every query is a scan over ready-made columns. Same clinical logic PHI already trusts, rebuilt to scale across ICBs.
0.886
CLDI C-statistic
<1s
per cohort scan