SSF had the pleasure of helping Revenio Group in developing their new asthma diagnosis product, Ventica. Ventica uses a novel non-invasive approach for identifying airway obstructions in even very young patients. It works by collecting respiration data from the patient while they are sleeping.
Revenio had concentrated their efforts in research and development of an algorithm, which extracted diagnosis criteria from human selected sections of the respiration data. The operator selection of data, however, would not have been feasible in the final product. SSF was thus commissioned to develop an automated data selection algorithm.
Revenio had developed their algorithm in MATLAB, which required the data selection algorithm to also be developed in MATLAB. This allowed using built in signal processing functions, but also posed challenges in terms of performance and memory use.
The measurements performed by Ventica are based on skin contact electrodes, which means that the signal integrity may be easily disrupted. Common causes are for instance the patient turning in bed, coughing or talking. It is also possible that an electrode becomes disconnected during the measurement. All these disruptions needed to be detected by the algorithm and removed from the analysis data.
During development, SSF also implemented a fast raw data parser in MATLAB for importing the raw data from the measurement hardware. MATLAB is known for its poor performance in control structures, which makes efficient data parsing difficult – in this case a speed up factor of about 100 was obtained over a naive solution.