Severity classification
Move beyond binary screening toward clinically meaningful severity groups.
MAJOR PROJECT / FRONTEND RESEARCH STORYDOCUMENTATION MODE / 2026
SHWASA / Project narrative
The sound of
breathing,
made legible.
An interactive visual account of a respiratory sound research project — from the question we asked to the evidence we measured.
RESPIRATORY SOUND / SIGNAL 001+INPUT
We are not presenting a diagnosis. We are making the research process visible: the data, transformations, architecture, and evaluation choices behind an experimental screening workflow.
◌
CRACKLEQUIETRAW RESPIRATORY TRACE / 08.4 SEC
01
WHY THIS MATTERS · COPD-EFF
Can a breath carry a clinical signal?
COPD screening often begins with equipment, appointments, and effort. This project explores whether the sound of breathing can become a more accessible research signal.
INPUTA short respiratory recording
→OUTPUTA measurable acoustic trace
01 / 04
Evidence, not decoration · COPD-EFF
A model is only
one part of the story.
Performance is read alongside data quality, patient-level separation, and the operating threshold. Research screening signal only. Not a clinical diagnosis.
87.73%Accuracy
88.38%Sensitivity
83.80%Specificity
92.51%F1 score
Where this can go next
From prototype
to evidence.
Broader validation
Test across cohorts, devices, environments, and patient populations.
Clinical partnership
Translate an academic prototype into a carefully validated research tool.