thymia Documentation

When a Helios model run completes with status COMPLETE_OK, the response contains a results object with biomarker scores for the recording. This page explains the structure of those results, how scores are calculated, and how to use them responsibly.

Results structure

Results are returned as an array of sections, each covering a timed segment of the recording. Currently, Helios always returns a single section covering the full recording.

1 {
2   "results": {
3     "sections": [
4       {
5         "startSecs": 0,
6         "finishSecs": 28.6,
7         "transcript": "The North Wind and the Sun had a quarrel...",
8         "mentalStrain": { "value": 0.42 },
9         "distress": { "value": 0.33 },
10         "stress": { "value": 0.33 },
11         "exhaustion": { "value": 0.66 },
12         "sleepPropensity": { "value": 0.66 },
13         "lowSelfEsteem": { "value": 0.0 },
14         "uniformDistress": { "value": 0.48 },
15         "uniformStress": { "value": 0.39 },
16         "uniformExhaustion": { "value": 0.61 },
17         "uniformSleepPropensity": { "value": 0.58 },
18         "uniformLowSelfEsteem": { "value": 0.21 }
19       }
20     ]
21   }
22 }

Each section contains:

Field	Description
`startSecs`	Time in seconds from the start of the recording where this section begins
`finishSecs`	Time in seconds from the start of the recording where this section ends
`transcript`	Transcription of speech within this section

Biomarker scores

Biomarker	Description
`mentalStrain`	Overall measure of mental wellness — the composite signal across all dimensions
`distress`	Levels of worry, nervousness or feeling on edge
`stress`	How anxious or wound up the user may feel; ability to relax
`exhaustion`	Physical and mental tiredness resulting from manual or cognitive labour
`sleepPropensity`	Likelihood of falling asleep if given the opportunity; linked to sleep quality
`lowSelfEsteem`	How confident and capable the user feels in themselves

Bucketed scores

Five biomarkers — distress, stress, exhaustion, sleepPropensity, and lowSelfEsteem — are returned as one of four discrete values:

Value	Label
`0.0`	Low
`0.33`	Moderately low
`0.66`	Moderately high
`1.0`	High

Bucketed scores are benchmarked against a matched population in thymia’s database — users of the same age, birth sex, and language or accent.

Benchmarking accuracy depends directly on the quality of demographic data supplied when creating the model run. Inaccurate dateOfBirth, birthSex, or language values will result in comparisons against a mismatched population and less reliable scores.

Continuous scores

The same five biomarkers are also returned as continuous values under the uniform prefix — for example, uniformDistress, uniformExhaustion. These can take any value between 0 and 1.

mentalStrain is returned only as a continuous score and has no bucketed equivalent.

Score interpretation guide

Score range	Interpretation
`0.0 – 0.25`	Low — within normal range for matched population
`0.25 – 0.50`	Moderately low — slightly elevated
`0.50 – 0.75`	Moderately high — notably elevated, worth monitoring
`0.75 – 1.0`	High — significantly elevated relative to matched population

Helios scores reflect the user’s state at the time of recording. They are point-in-time measurements and should be interpreted in the context of longitudinal trends where possible.

Longitudinal monitoring

Helios is designed for continuous monitoring. Submitting regular recordings for the same userLabel over time allows you to track changes in a user’s wellbeing state and identify meaningful trends.

For safety-critical applications, we recommend establishing a baseline for each user over several sessions before acting on individual readings.

Responsible use

Helios is a wellness monitoring tool. It is not a regulated medical device and must not be used for clinical diagnosis or as a substitute for clinical judgement. Results should not be presented to users as diagnostic or predictive of any medical condition.

For clinical assessment of depression and anxiety, see Apollo.

Next steps

Error Reference — Handle COMPLETE_ERROR responses correctly in your integration.
Apollo — For clinical-grade assessment of depression and anxiety.
API Reference — Complete endpoint documentation for the Helios API.