the problem
Peripheral artery disease is common, under-diagnosed and easiest to treat early. Near-infrared spectroscopy can read tissue oxygenation non-invasively, but the raw signal is noisy, subject-dependent and hard to turn into a decision a clinician can trust.
what we built
NIRS signal-processing pipelines that clean, segment and normalise tissue-oxygenation recordings into features that mean the same thing across subjects. On top of them, a machine-learning diagnostic model that classifies peripheral arterial disease from those features. The work is grounded in general-purpose signal-processing R&D that Bones applies across biomedical, industrial and communications domains.
what it changed
A non-invasive measurement becomes a candidate screening tool. The pipeline makes the signal reproducible; the model makes it decision-grade; both are built so clinicians can see why a result was produced.