Interactive tools checked against the reference codes

What is compared

• LQ and LQL dose (LQL-Equiv): biologically effective dose per course before proliferation, against lqlequiv.model._lql_dose_term of LQL-Equiv-web (Python), for the rectum, bladder and spinal cord, 1 to 25 fractions of 1 to 20 Gy.
• Thyroid protocol explorer (RAIR-Sim): thyroglobulin every month over 80 months, against the symbolic solution of the public MATLAB demonstrators (Demonstrateur98f.m, Demonstrateur666f.m), refractory and responding profiles, 1 and 8 sessions of 3.7 GBq every 3 months. The browser integrates the thyroglobulin equation with a fourth-order Runge–Kutta scheme.
• NICE^k (guide): against nice_metrics.m, with simple persistence as the reference (the only reference of NICE).
• BLEND (guide): weights per phase and RMSE, against predict_P_BLEND_LS.m, horizons 1, 3 and 6.
• sCV and F (guide): against scv.m and forecastability.m, horizons 1 and 3.
• ARTU coefficients (worked example): K and α interpolated in the tables M(R), against K_alpha.m of the ARTU repository, for the four values of R, at the example of the article and at five other points.
• γ-index (guide): γ map, γ̄, γ_max, GPR and RMSE on four synthetic cloud scenes (displacement, delay, bias, combination), against a literal MATLAB transcription of Eq. (1) of the article; the purely spatial GPR also matches computeGammaMetrics.m of the repository.

The reference values were produced on 23 October 2026 with MATLAB R2025b and Python, from the archived codes; the inputs are a synthetic hourly series of 30 days (test vectors, JSON). Scripts: make_vectors.m and make_vectors_lql.py.