Evidence before inference
We distinguish observation from interpretation, and interpretation from action.
Hardware-independent signal analysis
Understand yourcardiac signals.
ASH helps teams assess recordings from different devices: what is usable, what may be noise, and what needs closer review.
Explore the approach ↓Research principles shared publicly · Implementation kept private
Before interpreting a signal
Check how it was recorded, where data is missing, and how much noise it contains.
01 / Philosophy
What we check.
Our analysis accounts for how the data was collected and what it can support.
Recording conditions
We consider how collection conditions and variation affect the analysis.
Reproducible results
A result should be traceable to its source and possible to reproduce.
Limits of the evidence
We state what the data supports and where it leaves uncertainty.
02 / Approach
From recording to analysis.
We review the recording and its quality before interpreting patterns.
Inspect the recording
Review the original signal and the available information about its capture.
Check quality
Identify noise and missing data that could affect the result.
Consider the context
Examine patterns alongside the conditions in which they were recorded.
Explain the result
Document the findings, the evidence behind them, and what remains uncertain.
03 / Research
Research questions.
These questions guide our research. We share our approach while keeping implementation details and sensitive data private.
Question / 01How signal quality affects interpretation
Question / 02How to validate changes to an analysis
Question / 03What is lost when recording context is removed
04 / How we partner
How we partner.
We work with teams that collect or analyze physiological signals. Together, we define the question, the data available, and what we will deliver.
Compare technical options
We investigate a specific question and deliver a comparison, a recommendation, and the evidence behind it.
Assess the recordings
We review noise, artifacts, recording conditions, and traceability, then document how they affect the analysis.
Test an analysis pipeline
We build a small prototype that takes a signal through analysis to a documented result the team can inspect and reproduce.
05 / Roent
Evidence and privacy.
We distinguish correlation from causation, state when evidence is incomplete, and protect the privacy of the data we work with.