Biosignal ML & MedicalField engagement

An accuracy number inflated by subject-level leakage is worse than no number, because it drives a claim that does not hold in the field.

Architecture and delivery lead.
subject-disjoint
the split that matters
3
modalities fused
defer
when quality is poor
training distributiondeployment distributionunseen, where it fails
What was at stake

Biosignal systems built end to end, from raw PPG, BCG, and camera-based rPPG through leakage-safe evaluation into deployment and monitoring. The field is full of one specific failure: models that look accurate on a random split and collapse on people they have never seen. In non-invasive physiological estimation that inflated number drives a product or a regulatory claim that does not hold in the field.

The constraint

The felt need is a bigger or foundation model. The real problem is whether the accuracy you already have is real. A model predicting near the population mean can clear consensus error-grid zones and post a low error while its variance explained against a mean baseline is negative. Confirm the number is trustworthy before adding capacity, because capacity added on top of a leaking evaluation just produces a more confident wrong claim.

The fork

The reflex, and the fix.

Road not taken

Report accuracy on a random split

Pull

It is the default, it is what the tooling does, and the number it produces is flattering.

Why not

A random split keeps the same people on both sides of the line, so it measures whether the model recognizes those individuals rather than whether it does the task.

Road taken

Prove it under leave-some-subjects-out

Accepted

A grouped, subject-disjoint evaluation with a held-out validation set kept off the test set, plus overlap and duplicate detection.

Bought

An accuracy figure that survives contact with people the system has never seen.

Decision

Establish that the number is real before adding capacity, because leakage does not announce itself.

How it was built
01Acquire
02Quality gate
03Represent
04Fuse
05Validate
06Promote
01

PPG representation learning

Self-supervised temporal modeling, masking, and contrastive learning into reusable transferable embeddings, served through a shared inference backbone with versioned model and preprocessing and reusable downstream heads.

02

PPG and BCG fusion

Cross-modal temporal alignment, cross-attention, and pulse-transit-time coupling, with missing-modality support, modality gating, and graceful degradation when a channel drops out.

03

Leakage-safe validation, the lead facet

Grouped subject-disjoint evaluation, overlap and duplicate detection, calibration, and reproducible benchmarking, wired to promotion gates so a model cannot reach production on a number that was never real.

04

A signal-quality model

Learned artifact representation, weak supervision, and quality calibration, with runtime gating, defer behavior, and reacquisition, so the system declines to answer rather than answering from noise.

05

rPPG and PPG fusion, the harder facet

A full computer-vision front end doing frame validation, face tracking, and ROI stabilization, fused with wearable PPG under real-time modality selection across operating modes, with camera-held-out domain validation.

How it was measured

A single headline number hides where a system fails. This work was scored on the dimensions that actually decide whether it holds in production, measured on real, held-out cases rather than the demo path.

Error under a subject-disjoint splitVariance explained against a mean baselineBehavior with a modality missingDefer rate under poor signal quality
figures

No self-reported accuracy figure is claimed, deliberately. Validation method is the claim.

What it produces
Without this discipline

A strong number on a random split, a product claim built on it, and a collapse the first time the system meets people it has never seen, discovered after the claim is public.

This system

Biosignal systems that report accuracy honestly under the split that matters, degrade gracefully when a modality is missing, and defer when signal quality is poor, with a shared backbone and reusable heads across tasks.

subject-disjoint by defaultmodality gatingdefer on poor qualitypromotion gates
The operating envelope

What it owns, and what it hands to a person.

Handled with confidence
PPG, BCG, and camera rPPG
Graceful degradation with a modality missing
Runtime quality gating and defer
Flagged for review
Low signal quality
Domains outside the validated set
Out of scope by design
Claims validated only on a random split
Regulatory claims without the disjoint evidence
The honest limit

The honest position here is that leakage-safe validation is the product rather than a headline number, so no self-reported accuracy figure is claimed on this page. A figure produced under a weaker split would be larger and would mean less.

What it generalizes to

Combining the most information does not win. The strongest system knows which information to trust under current conditions and degrades gracefully when it cannot. The same discipline carries to any estimation problem with per-subject or per-site structure in the data, which is most of them.

How we engage

You have a system like this one.
Tell us where it stands.

Whether it is failing, not yet built, or about to meet a scale it has never seen, we can tell you what we see.

Start a conversation
mostafa@opulion.dev · Response within 24 hours · By inquiry