Task

To statistically analyze the data from randomized clinical trials related to sleep apnea and sleep quality. To create a machine learning workflow for snoring sound classification.

Challanges

• The data was multi-modal (Apnea-Hypopnea Index (AHI), oxygen levels, bio-impedance, audio recordings, etc.), requiring re-synchronization and containing significant noise and several experimental errors.
• We had to work within an existing complex web of systems and databases.
• We needed to create a solid machine learning workflow for snoring sound classification.
• We assisted in creating a large dataset of annotated snoring sounds.

How we helped

• Performed rigorous statistical analysis of the clinical study data (t-tests, p-values, ANOVA, etc.).
• Assisted our client, Signifier Technologies, in sanitizing their clinical trial data and extracting valuable findings.
• Developed a best-practice machine learning workflow that included data sanitization, data annotation, model training, and validation.

Results

The outcomes of our investigations helped our client gain a better understanding of their clinical study results. We also contributed to numerous improvements and simplifications of the client’s internal software stack.