JevESMO
Spanish oncology decision-support prototype that walks ESMO guideline trees and uses Jev only for judgment calls.
Overview
JevESMO is a clinical decision-support research prototype for oncology. Starting from a patient's clinical record, it walks the ESMO guideline decision tree for 43 tumour types across 11 specialties and uses Jev for decisions that require clinical judgment, returning an explainable recommendation that shows what was asked, what was answered and with what confidence. It is never autonomous: if required data is missing the app asks instead of computing, and if confidence is low a review by an oncologist is mandatory.
A generic engine reads one JSON spec per tumour. Required fields are checked first, deterministic code derives subtypes and risk groups, Jev layers one and two give judgments in grey zones as context, deterministic rules list the options ESMO allows, Jev layer three chooses among valid options with expected benefit, and Jev layer four checks safety and contraindications that can block options. A confidence gate sends results below 60%, cases with no safe option, or ECOG 3-4 non-candidates to mandatory review. The README states it is not a certified medical device and does not replace clinical judgment. Documentation is in Spanish.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which ESMO-permitted treatment option best fits this patient? | — |
| Yes / no | Is there a contraindication that blocks this option? | — |
| Other | Grey-zone clinical judgments as Choice, Score or Noul per tumour spec | — |
How it works
- Check required patient data; ask if missing
- Derive subtypes and risk groups deterministically
- Jev judges grey zones, chooses among permitted options and checks safety
- Confidence gate forces oncologist review when needed
Strengths
- Hard guideline knowledge in deterministic rules
- Full explainability of questions and probabilities
- Never autonomous; mandatory review gates
Caveats
- Research prototype, not a certified medical device
- Guideline encoding accuracy must be validated
- Spanish documentation
At a glance
- Type
- Project
- Status
- research
- Language
- Python
- Runtime
- Web app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Apoyo a la decision clinica oncologica: guias ESMO (43 tumores) + Jev System One, con explicabilidad y evaluacion. Prototipo de investigacion.
hellogumbo/awesome-jev
Frequently asked questions
Does it make treatment decisions on its own?
No, it never acts autonomously and low-confidence cases require oncologist review.
Where does guideline knowledge live?
In deterministic per-tumour JSON specs; Jev only handles judgment calls.
Is it a certified medical device?
No, it is a research prototype.
Related projects
TLA+-verified consensus kernel around Jev that may escalate but never returns a confident wrong verdict, tested on 1,680 pharmacy cases.
ProjectresearchRustMedplum demo where a bot uses Jev to compare a provider's visit note with a discharge summary and flag agreement or potential conflict.
AppprototypeTypeScriptFrench library ranking ADEME emission factors for an activity label, splitting code rules from Jev's semantic match.
ProjectprototypeJavaScriptSAP S/4HANA proof of concept that sends anonymised purchase-order risk indicators to Jev and classifies orders to release, approve or block.
ProjectprototypeABAPSmall web demo that makes clinic triage decisions with Jev, falling back to heuristics when no key is set.
DemoprototypeTypeScriptChinese-language agent skill that offloads classify, screen, score and compliance checks to Jev via OpenCode Zen's free tier.
PluginbetaPython