Breath-phase segmentation tracks a changing respiratory process rather than capture only a moment. For two centuries, the stethoscope has supported auscultation, enabling skilled clinicians to detect wheeze, crackles, and reduced breath sounds with reasonable reliability. The limitation is not the instrument or ear, but intermittency. Auscultation provides a ten-second snapshot. Acoustic biomarkers convert continuous breath sounds into quantifiable, trackable features. This article explains what they are, how they are derived, and why some are clinically meaningful while others remain investigational.
Defining an Acoustic Biomarker
Classically, a biomarker is a measurable biological indicator linked to a physiological or pathological process. An acoustic biomarker is a breath-sound waveform feature associated with particular anatomical or physiological airway states. Unlike imaging or laboratory biomarkers, these features can be measured passively, continuously, and non-invasively, without patient effort or contact with a healthcare setting.
The input is a time-domain acoustic signal recorded by a sensor against the chest wall or at the trachea. Processing occurs in stages. Noise reduction first removes environmental artefact. Breath-phase segmentation then marks the start and end of each inspiratory and expiratory phase. Feature extraction calculates measurable properties for each phase. The resulting acoustic biomarkers include duration ratios, spectral content in specific frequency bands, the timing of acoustic events, and measures of signal regularity.
Four Clinically Informative Acoustic Features
Inspiratory and Expiratory Duration
The inspiratory-to-expiratory phase-duration ratio is the most physiologically grounded acoustic biomarker for COPD. In healthy adults at rest, inspiration takes roughly one-third of the breath cycle. With obstructive disease, dynamic collapse of peripheral airways restricts expiratory flow, so the body lengthens expiration to clear trapped air. As obstruction increases, the I:E ratio moves from approximately 1:2 towards 1:3 or longer. Tidal breath recordings can measure this change without patient effort, and it reflects dynamic hyperinflation rather than one effort-dependent manoeuvre.
Crackle Burden and Distribution
Crackles are discontinuous, non-musical sounds produced when previously closed or collapsed airways open suddenly, or when secretions move. In COPD, coarse crackles are mainly heard in the mid-to-late inspiratory phase in the lower lobes. Their density broadly reflects secretion retention and small airways inflammation. Fine, early-inspiratory crackles have a different clinical meaning and are more associated with pulmonary fibrosis than COPD. This distinction is important, and our models are trained to recognise both distribution patterns.
Crackle density is measured as a count per breath, averaged across monitoring windows. Change from a patient's individual baseline is more informative than an absolute threshold, since crackle characteristics vary substantially with body habitus, position, and baseline disease severity.
Wheeze Scoring
Wheeze is a continuous musical sound caused by airflow through narrowed airways. In COPD, it is usually expiratory, polyphonic, meaning multiple frequencies occur simultaneously, and present during tidal breathing rather than only during forced manoeuvres. The wheeze index is the proportion of each expiratory phase occupied by wheeze frequencies, calculated from signal spectral content in the 100 to 1,000 Hz range. New-onset wheeze, or a rise in the index from baseline, indicates progressive bronchospasm or worsening mucosal oedema.
The wheeze index is not being presented as an independent predictor of exacerbation. It contributes to a composite score. A patient whose index rises by 30% over four days while the I:E ratio also lengthens has a different risk profile from one with an isolated index change. A composite is needed because no single acoustic feature is sufficiently specific.
Overnight Breathing Rate
Respiratory rate is among the most sensitive early indicators of physiological deterioration, yet it remains one of the vital signs most consistently underrecorded in clinical practice. During sleep, it is relatively stable and less affected by the voluntary modulation seen while awake. A sustained overnight increase of four or more breaths per minute above an individual's baseline is a meaningful physiological signal. It indicates that the work of breathing has risen beyond what the body can offset through behavioural adaptation alone.
What These Features Cannot Show
Acoustic biomarkers describe airway dynamics and breathing pattern. They do not show blood gas status, oxygen saturation, or systemic inflammatory markers. A patient may have a clearly abnormal acoustic profile without hypoxaemia, and the reverse is also possible. The acoustic signal does not replace clinical assessment. It prompts clinical assessment.
The features above perform best when COPD is the dominant pathology. Concurrent cardiac disease, neuromuscular conditions that affect respiratory mechanics, or active pleural disease can make the acoustic profile more complex, requiring greater caution in clinical interpretation. We address these limitations in the clinician-facing guidance supplied with the monitoring system.
Turning Features into a Score
The four features above are combined in the RespiroScore, a daily index on a 0 to 100 scale. A score below 40 represents acoustic features within the patient's stable range. Scores between 40 and 65 indicate deviation from baseline and prompt a review flag. Scores above 65 trigger an alert to the monitoring coordinator. In this pilot phase, thresholds are set conservatively to favour sensitivity over specificity. Some review flags will therefore not lead to clinical intervention. This is deliberate: in an early monitoring programme, missing a deteriorating patient is more serious than prompting an unnecessary telephone call.
As the enrolled cohort expands and more labelled outcome data accumulate, we will refine score thresholds and feature weightings. This is standard clinical algorithm development practice. We are transparent that current calibration draws on a small number of monitored patient-months. The score is useful clinically today, but it is not yet the fully optimised instrument it may become with more data.
Acoustic Biomarkers versus Acoustic Diagnostics
Using acoustic features to support ongoing monitoring in a patient with known COPD is different from using them to diagnose COPD or describe its severity. RespiroScreen does the former. The latter requires a diagnostic standard, and current evidence for acoustic COPD diagnosis is not strong enough to justify replacing spirometry. We make no diagnostic claims. For a patient with established COPD, we claim that a sustained change in the individual's acoustic profile is a clinically significant signal worth investigating. These are different claims, and that distinction matters when assessing the technology.