Now onboarding institutional research cohorts
Science — the method
Wearable EEG, PPG and respiration are read together, quality-checked before anything is scored, and compared with the practitioner’s own baseline. This is how the measures are built — and where their limits are.
Session profile
Three streams, checked before scoring.
Wearable EEG
Attention
PPG
Regulation
Respiration
Breath
What the record can show
Attention, regulation and recovery ask different questions of the same recording. Each is reported with its own uncertainty, and none is compressed into a single score for the whole sit.
Attention is the central claim of any practice, and the hardest thing to verify from the inside. It is easy to finish a session believing attention was steady when the record says otherwise.
Wearable EEG gives the session a second opinion. DeepMindex reads the neural record for how attention behaved across the sit, rather than compressing it into one quality score for the whole session.
Features drawn from the research base are treated as candidates and developed as candidates. Nothing is presented as a measure until it has survived validation.
Physiological regulation describes how the autonomic system responds across a session: whether the body settles, and how quickly it recovers after effort or distraction.
PPG-derived heart rate and pulse-rate variability carry that information. Working from pulse-rate variability rather than ECG matters here, because it is the signal a wearable can realistically record.
Read alongside the EEG, regulation becomes part of the same session profile instead of a separate health metric kept in another app.
One session tells you very little. Recovery is about whether regulation between sessions is improving, which only becomes visible across a series.
DeepMindex therefore anchors every reading to the practitioner’s own baseline, built from their earlier sessions, and reports the trend rather than a single figure detached from it.
That also means the platform deliberately avoids ranking one practitioner against another. A useful comparison is with your own history, not with the person on the next cushion.
How a reading earns trust
Any sensor placed on a person will produce a line. The question worth asking of an assessment platform is whether that line can be trusted.
Every session is checked before it is scored, rather than after the fact.
Segments that fail the quality threshold are excluded, not averaged into the result — so a reading that is nearly noise is not dressed up as a finding.
Population norms describe populations. Applied to one person they can mislead, because an entirely ordinary session can sit well outside a range built from other people.
Progress is measured against the practitioner’s own earlier sessions, and one practitioner is never ranked against another.
Neural and autonomic streams are modelled together, because attention and regulation move together.
The output reads as one session profile rather than two disconnected dashboards.
Every reading carries how confident the platform is in it.
Where a trend is not yet meaningful, that is said rather than drawn as a line.
What still has to be proven
Candidate features drawn from existing recordings still have to hold up under wearable conditions, across experience levels and across sites.
That is the work of independent validation: reliability and generalisation, assessed rather than asserted.
We would rather state the limits of the method plainly than imply a maturity it has not reached.
The autonomic read-out
PPG-derived pulse-rate variability is the autonomic signal a wearable can record continuously. It is informative, and it is easy to over-read.
The interval between heartbeats is never perfectly even. It tends to shorten slightly on inhalation and lengthen on exhalation, and the size of those fluctuations reflects the balance between sympathetic and parasympathetic influence on the heart.
That balance is one of the clearest available windows onto physiological regulation, which is why the platform reads it alongside the neural record rather than on its own.
Conventional variability measures are derived from ECG. PPG reads changes in blood volume at the skin instead, which is what a wearable can record continuously and unobtrusively — and pulse-rate variability is what PPG yields.
The trade-off is real: PPG is more exposed to movement artifact. That is precisely why signal quality is assessed first, and why poor segments are excluded rather than averaged into the result.
Variability measures describe autonomic activity. They are not a diagnosis, and they shift with sleep, posture, caffeine, illness and time of day.
DeepMindex therefore reads them against the practitioner’s own baseline, taken at comparable times, and reports the trend with its uncertainty rather than as a single ranked value.
Open questions
These are open research questions rather than completed studies. Each one will be published when it is through review.
In progress · Validation programme under way
Research foundation
DeepMindex builds on long-standing research in EEG and meditation, and on an existing body of EEG and PPG recordings that includes three deep-jhana meditation recordings.
Those recordings capture states that are rarely measured in a laboratory at all. They guide candidate-feature development, subject to the applicable ownership, consent and data-use permissions.
Further data is collected using standardised protocols across experience levels, and independent validation assesses reliability and generalisation.
Dr. Junling Gao
Founder
Going further
The method draws on published work in EEG, autonomic physiology and meditation research. These are the gateways we point people to, rather than rehosting the literature.
Collaboration
For research access, collaboration on validation, or data-use enquiries, we welcome contact from academic groups and institutions.