Cough

Due to swelling of the nasal mucosa, which makes breathing difficult, patients begin to breathe their mouths. As a result, the mucous membrane of the upper respiratory tract dries up, a sore throat occurs, followed by a dry cough. At the same time, the flow of mucus along the back wall of the pharynx leads to the development of a wet cough. As a rule, an acute cough resolves within 1-2 weeks.

Diagnosis of diseases by coughing

To avoid complications at different stages of the disease, it is important to correctly determine the nature of the cough - dry (unproductive) or wet (productive) - and the causes of its occurrence (ARVI, allergies, chronic diseases of the upper and lower respiratory tract, gastroesophageal reflux, heart disease, etc.).

Up to 90% of diseases of adults and children with cough are due to a viral infection. At the same time, a combination of various pathogens is possible, including the addition of a bacterial infection.

2025: AI model released that recognizes different types of cough and indicates diseases

In mid-October 2025, American researchers at North Carolina State University announced the development of a new model of artificial intelligence to detect various types of cough. The system makes it easier to monitor chronic diseases and predict health risks such as asthma attacks.

As noted by one of the authors of the work, professor of electrical engineering and computing systems Edgar Lobaton (Edgar Lobaton), cough serves as an important biomarker for tracking various states of the body. For example, it may indicate the development of a respiratory disease. Against the background of the spread of modern wearable medical devices, the ability to register and analyze coughs is expanding.

soruce=Freepik

The researchers say that conventional machine learning algorithms are able to distinguish cough from background noise with high enough accuracy. However, when processing similar sounds, such as sneezing, coughing or moaning, efficiency drops sharply.

To solve the problem, experts suggest combining two sources of data from wearable devices. These are the actual sounds taken by the built-in microphone, as well as the readings of the accelerometer - a sensor capable of detecting changes in movement, tilt and vibration. According to the authors of the work, movement alone cannot be used to detect coughs, since it gives limited information about the source of sound. Various actions, such as laughter and coughing, can give similar readings from the accelerometer. But the combination of sounds and motion increases the accuracy of AI algorithms.

This is a significant step forward. We have learned very well to distinguish cough from human speech, and the new model distinguishes it significantly better from non-verbal sounds, says Lobaton.[1]

Notes