DeepSound AI algorithm diagnoses respiratory diseases in less than a second with 98 percent accuracy

The EMBLE company developed a neural network algorithm for respiratory disease diagnostics

16 September 2020, at 9:00am

The EMBLE company developed the DeepSound AI algorithm for any sound events classification. EMBLE trained the algorithm on Kaggle human respiratory sound database and taught the model to recognize the breathing of sick and healthy people by auscultation. The DeepSound AI neural network has learned to diagnose COPD, bronchiectasis, and bronchiolitis diseases by human breathing in less than a second with 98 percent accuracy, 99 percent specificity, and 98 percent sensitivity. It’s also noise-resistant, hence it does not require an ideal environment.

The EMBLE research team set the goal of mathematically describing the whole space of pulmonary diseases. It will allow an increase in the number of diagnosed pulmonary diseases for the possibility of their timely detection without the use of expensive procedures, thereby minimizing the risks of chronic forms of respiratory diseases.

Such a diagnosis method will be most in-demand in veterinary medicine, where other modern diagnostic methods, such as CT, MRI, X-ray, are often contraindicated for animals with cardiovascular or respiratory diseases.

Now EMBLE is launching a SaaS service for equine respiratory disease diagnostics. DeepSound AI will help veterinarians to perform respiratory system diagnostics of racehorses right in the stables with 100x faster, 30x cheaper, and safe for human and animal.