SAFE-Gate

Overview

Safety-first gated triage ensemble with conservative risk merging and formal safety checks.

This repository is part of an eight-repository clinical decision-support research portfolio. Current status: manuscript or component package in preparation. The repository role is manuscript and supplementary.

Standard Repository Layout

Path	Purpose
src/	Package source code: safe_gate
tests/	Unit, smoke, and behavior checks
scripts/	Reproducibility and export scripts
examples/	Runnable examples and demonstrations
figures/, visualizations/, outputs/, results/	Generated visual and result artifacts
data/, models/, evaluation/	Dataset, model, and evaluation assets when used by this repo
FIGURE_MANIFEST.csv	Curated figure inventory for manuscript or component evidence
pyproject.toml, setup.py, requirements.txt, pytest.ini	Python package and test configuration

Architecture Flow

flowchart LR
    A[Input data or scenario] --> B[Core package logic]
    B --> C[Safety and quality checks]
    C --> D[Metrics and audit outputs]
    D --> E[Curated figures and result artifacts]

Loading

Core Logic

• Evaluate six safety gates.
• Merge gate outputs conservatively.
• Generate risk tier and abstention decision.
• Export safety, baseline, and per-class figures.

Key Formulas And Rules

• Merged risk: R = max(G1, G2, G3, G4, G5, G6)
• Abstention rule: abstain if uncertainty >= theta_u or data_quality < theta_q
• Safety certificate: valid if all gate invariants pass

Data, Results, Charts, And Graphs

The curated visual set is controlled by FIGURE_MANIFEST.csv and currently lists 6 figure entries. The manifest links figure IDs, roles, source scripts, source data, captions, sections, timestamps, and export DPI.

ID	Role	PNG	PDF
SAFE-F1	manuscript	evaluation\manuscript_figures\safety_performance.png	evaluation\manuscript_figures\safety_performance.pdf
SAFE-F2	manuscript	evaluation\manuscript_figures\baseline_comparison.png	evaluation\manuscript_figures\baseline_comparison.pdf
SAFE-F3	manuscript	evaluation\manuscript_figures\confusion_matrix.png	evaluation\manuscript_figures\confusion_matrix.pdf
SAFE-F4	manuscript	evaluation\manuscript_figures\per_class_metrics.png	evaluation\manuscript_figures\per_class_metrics.pdf

Reproduce

cd D:\PhD-NU\Manuscript\GitHub\SAFE-Gate
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -e .
python -m pytest -q

If figure-generation scripts are present, run the matching script listed in FIGURE_MANIFEST.csv from the repository root.

Verification Criteria

• Root metadata and package files are present.
• Source paths follow src/<package>/... where the package shape allows it.
• Tests pass with python -m pytest -q.
• Curated figures are listed in FIGURE_MANIFEST.csv rather than inferred from every raw image file.
• Manuscript status wording stays conservative: in preparation, implementation, supplementary, or reproducibility/component evidence as appropriate.
• No local manuscript path, external assistant wording, or software metadata block is kept in the repository text.

Portfolio Relationship

Repository	Role
BASICS-CDSS	Beyond-accuracy evaluation methodology
TRI-X	Framework-level package
ORASR	Routing and safety-action component
DRAS-5	Dynamic risk-state component
SAFE-Gate	Safety-gated ensemble framework
SynDX	Synthetic validation and explainability evidence
SURgul	SRGL/governance reproducibility component
TRI-X-CDSS	Integration and implementation package

Contact

Chatchai Tritham
Department of Computer Science and Information Technology, Faculty of Science, Naresuan University, Phitsanulok 65000, Thailand
Email: chatchait66@nu.ac.th
ORCID: 0000-0001-7899-228X

Chakkrit Snae Namahoot
Department of Computer Science and Information Technology, Faculty of Science, Naresuan University, Phitsanulok 65000, Thailand
Email: chakkrits@nu.ac.th
ORCID: 0000-0003-4660-4590