The significance of denoising biomedical signals, particularly in crucial life-saving interventions like cardiopulmonary resuscitation (CPR), cannot be overstated, as the accuracy and reliability of CPR data can be a matter of life and death.

Despite advancements in technology, traditional methods of signal denoising using filters often fall short of effectively removing noise while preserving the integrity of underlying signals. However, a high level of precision is crucial considering CPR as a matter of human life and death.

In this context, the capability of ML to capture the complex underlying patterns is proven in the biomedical domain.

Although there are some noted ML methods for denoising CPR signals, these are supervised approaches.

The difficulty and unavailability of getting labeled clean signals that correspond to the noisy signals reduces the potential for these ML methods to be effective in real-life contexts.

To solve this problem, a dedicated unsupervised ML method for CPR signals is required in the domain, which can process multiple CPR signals concurrently without labeled data.

Read More