Concept illustration of the Sentinel sample path

Scroll to run the clock

We compute where each organ is going.

How the estimate is made

A laboratory that fits beside the bed.

Sentinel is being designed to read blood biochemistry continuously, through a microfluidic sample path, at the bedside.

Concept illustrationWhat Sentinel is designed to do

All ten organ systems on one map.

The further an organ sits from the centre, the more stress it is under, which makes the one drifting toward the red boundary the one that is running out of time.

Will the drug arrive before the damage does?

MOISS, our Multi-Organ Intervention State Space, sets the time this organ has left against the time the therapy needs to take effect, and grades the intervention by the margin between the two.

How the grading works
Each dot is a lab draw, and each dashed line is the forecast made at that moment.

Why this exists

The problem we are solving

A decision in intensive care is still made from snapshots: a laboratory result from some hours ago, a reading at the bedside and whatever the clinician can hold in mind. Fields that manage systems which change quickly, from aviation to power grids, rest on a layer that measures continuously and computes what happens next, and medicine does not have one.

Aethryva is building that layer, starting in intensive care. The three scenes above show its parts at work, with the trajectories as what it computes, the sample path as how it would measure, and the intercept as how it times a therapy.

About the company

Measure
Continuously and at the bedside, without waiting on the laboratory.
Compute
Where each organ system is heading, and how long there is left to act.
Carry
To the whole care team, as protocols a computer can run and a picture everyone shares.

Behind the trajectories

How we estimate between lab results

Most ICU biomarkers are measured once every 6 to 24 hours, which leaves long stretches in which an organ can deteriorate without a single new number to show for it.

We close those stretches with a modified Kalman filter that weighs each incoming result by how reliable it is, so that a trustworthy result moves the estimate further than a doubtful one, and that carries on estimating through the hours when no result arrives at all.

Because the filter follows how fast each biomarker is moving as well as where it stands, it can say where that biomarker will be an hour from now, and that forecast is what everything else on this page is built on.

See the estimate at work

Position
Where the biomarker stands at this minute, including the minutes in which no result has come in.
Velocity
How fast it is rising or falling, which is what turns an estimate into a forecast.
Confidence
How far the estimate can be trusted. It tightens when a result arrives and loosens as time passes without one.

Behind the sample path

PriyAI Sentinel

Sentinel is the bedside instrument we are designing to give those estimates a data source of their own. It is intended to analyse blood biochemistry continuously through a microfluidic sample path so that it no longer waits on intermittent draws, to collect what the ventilator, infusion pumps, dialysis machine and monitor already know, and to turn that combined stream into a trajectory and a time left to intervene for each organ system.

The instrument is at the design stage, which means that no unit has been built or tested yet and that it is not available for clinical use.

Follow one sample through Sentinel

Concept rendering of the PriyAI Sentinel bedside instrument

Concept rendering.

Behind the intercept

How intervention timing is graded

For any organ and any therapy a clinician is weighing, MOISS (Multi-Organ Intervention State Space) works out two durations: how long the organ has before its trajectory reaches the critical threshold, which is the remaining distance divided by the current velocity, and how long the therapy takes to reach its effect.

The difference between the two is the margin, and its size relative to the therapy's own onset time places the intervention in one of six grades.

So far the grading has been tested retrospectively, against records from public critical care databases, and a prospective study at the bedside is still ahead of us.

See how a margin becomes a grade

We review each request and reply by email.

PROPHYLACTICThe therapy takes effect with more than its own onset time to spare.
ON_TIMEIt takes effect before the organ reaches the threshold.
PARTIALIt takes effect shortly after the threshold is crossed.
MARGINALIt takes effect later still, in the period where bridging therapy may help.
FUTILEIt takes effect well past the threshold, with little left to recover.
TOO_LATEIt cannot take effect before the damage is irreversible.

The layer, part by part

  • SentinelMeasures blood biochemistry continuously at the bedsideIn design
  • Estimation and timingComputes where each organ is heading and how long there is to actTested retrospectively
  • MOISSCodeEncodes clinical protocols so that a computer can run themOpen source
  • tachyDx TeamsKeeps an intensive care team on the same pictureLive
  • The SororiumTrains clinicians to work alongside bedside AIOpening soon

Contact

Get in touch

  1. CareersOpen roles and applicationsApply
Aethryva Deeptech

Paper request

Request the paper

Fill in the form below. We review each request and send the paper by email.

Aethryva Deeptech Pvt Ltd, Hyderabad. Your details are handled under our privacy policy.