Artemis II public-proxy method review

Real public data, clear limits.

This page shows the current ConductSignal Artemis demo run. The numeric rows come from public NHANES 2017-2018 and PhysioNet MMASH data. It is not Artemis II astronaut data, and the current run does not include numeric Inspiration4 immune, RNA, or microbiome rows.

Proxy observations

231

Reference rows

Ranked findings

Figure 1

Source Record

NHANES 2017-2018

Numeric rows used

Standard Measures reference context and example physiological rows

https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2017

PhysioNet MMASH

Numeric rows used

ARCHeR-like wearable, sleep, activity, and salivary time-series proxy

https://physionet.org/content/mmash/1.0.0/

Inspiration4/SOMA public supplementary sources

Context only in this run

Human spaceflight biology source inventory and access-boundary context

https://www.nature.com/articles/s41586-024-07639-y

Run audit

Reproducibility

Public repository

https://github.com/ConductScience-Foundation/artemis-ii-conductsignal-submission

Demo version

0.2.5

Generated UTC

2026-05-28T12:21:41+00:00

The run records input and output hashes. The page uses the same files listed in `method_provenance.json`, so reviewers can compare generated files against the public repository.

Figure 2

MMASH Subject Trace

Method: the line is built from `artemis/data/proxy_observations.csv`; the marker is from `artemis/output/change_point_table.csv`. The selected trace is MMASH_003 / steps per minute.

Figure 3

Reference Anchors

MMASH_001 · cortisol norm87.5 percentile

MMASH_001 · lying ratio91.7 percentile

MMASH_001 · mean hr19.4 percentile

MMASH_001 · melatonin norm100.0 percentile

MMASH_001 · sleep efficiency80.0 percentile

MMASH_001 · sleep minutes20.0 percentile

MMASH_001 · steps per minute2.8 percentile

MMASH_001 · vector magnitude2.8 percentile

MMASH_001 · waso40.0 percentile

MMASH_002 · cortisol norm25.0 percentile

MMASH_002 · lying ratio100.0 percentile

MMASH_002 · mean hr2.8 percentile

Each bar is an empirical percentile from `reference_anchor_table.csv`. NHANES rows provide terrestrial reference context; MMASH rows provide public proxy context for wearable and sleep features.

Figure 4

Recovery Review

MMASH_001 · sleep minutes · recovering52.0 effect (52.00)

MMASH_003 · steps per minute · returned_to_baseline9.1 effect (9.14)

MMASH_002 · steps per minute · recovering4.8 effect (4.79)

MMASH_001 · sleep efficiency · recovering4.8 effect (4.75)

MMASH_001 · waso · recovering4.0 effect (-4.00)

MMASH_001 · steps per minute · recovering4.0 effect (-3.97)

MMASH_001 · vector magnitude · returned_to_baseline4.0 effect (-3.96)

MMASH_004 · lying ratio · returned_to_baseline3.9 effect (3.94)

MMASH_002 · lying ratio · recovering3.9 effect (3.91)

MMASH_004 · steps per minute · returned_to_baseline3.9 effect (-3.88)

This view comes from `recovery_table.csv` and keeps subject-level direction visible instead of averaging four people into one claim.

Figure 5

Ranked Findings

Rank	Subject	Finding	Score	Source
1	MMASH_003	MMASH_003 showed its largest public MMASH-derived transit-window change in steps_per_minute, with cross-modal support. steps_per_minute: d=2.94 [2.194, 3.694]; mean_hr: d=0.05 [-0.696, 0.804]	0.888	PhysioNet MMASH
2	MMASH_004	MMASH_004 showed its largest public MMASH-derived transit-window change in lying_ratio, with cross-modal support. lying_ratio: d=1.60 [0.852, 2.352]; mean_hr: d=-0.35 [-1.097, 0.403]	0.720	PhysioNet MMASH
3	MMASH_002	MMASH_002 showed its largest public MMASH-derived transit-window change in lying_ratio, with cross-modal support. lying_ratio: d=-1.32 [-2.068, -0.568]; mean_hr: d=0.16 [-0.587, 0.913]	0.685	PhysioNet MMASH
4	MMASH_001	MMASH_001 showed its largest public MMASH-derived transit-window change in vector_magnitude, with cross-modal support. vector_magnitude: d=-0.66 [-1.405, 0.095]; mean_hr: d=-0.19 [-0.944, 0.556]	0.602	PhysioNet MMASH

Figure 6

Hash Trail

Inputs

api/artemis_findings.json

f99299eb3dc7...

artemis/data/proxy_observations.csv

021431821916...

artemis/data/reference_slices.csv

0b9c3a536839...

artemis/data/source_manifest.json

3bcfdff1675a...

Outputs

artemis/output/baseline_table.csv

136886b7453c...

artemis/output/change_point_table.csv

cdca7bfc3fe0...

artemis/output/concordance_table.csv

a07e134815b8...

artemis/output/crew_state_summary.csv

8d6da49b39d7...

artemis/output/crew_state_summary.svg

499959b6e8a3...

artemis/output/finding_atlas.csv

e855894e74a0...

artemis/output/missingness_report.csv

4940e67d92a9...

artemis/output/recovery_table.csv

8e561780bb02...

Conclusion

What This Means

The method is real-run-backed: current tables and figures are generated from public NHANES and MMASH files with recorded hashes.

The scientific claim is narrow: ConductSignal can organize small-cohort, many-measurement data into individual review priorities with uncertainty visible.

The limitation is explicit: these are public proxy rows, not Artemis II astronaut outcomes, and protected human spaceflight values remain outside this public demo.