Portable AI cough detector that predicts COVID-19
Researchers at the University of Machachussets (UMASS) Amherst have developed an AI cough detector that records coughs and sneezes of people present at public places and makes predictions whether a person is suffering from a Flu or COVID19. It is a non-speech body sound detector, its predictions are based on the sample size, where people congregate in huge numbers, such as banks, hospitals, malls, office complexes, etc. Currently, it is being used only in hospitals and the creators have envisioned it only for hospitals.
The research has been carried out by Professor Tauhidur Rehman and Forsad Al Hossain. The device is named “FluSense”. The tool is equipped with a microphone array, a thermal imaging camera with a Raspberry Pi and a neural network computing engine, the only concern that is left is of an individual’s security, but, the creators assist that it will not record any speech or image as speech or image data. Hence, the patients at the hospital can be well aware of their privacy concerns. It will record only the non-speech sounds from a sample population in data present in the healthcare premises. The size of FluSense is equivalent to the size of a large dictionary.
The microphone will record the non-speech data, while the thermal imaging camera will help in the recording of bodily temperatures. The thermal imaging camera is also used by Firefighters also to detect heat through smoke darkness. A neural network will help in the prediction, classification, pattern recognition, etc of the data that will be stored, this will help in identifying the sickness trends in multiple hospitals and due to which it will help make certain decisions on the demand and supply of medical instruments, such as predicting the demand for vaccine production, the number of beds needed in a hospital, hospital staffing. FluSense can be termed as one of the dire needs of hospitals in the current scenario of a global crisis caused by the pandemic COVID19.
The FluSense is currently working for 8 months and has analyzed more than 350,000 thermal image samples and a whopping 21 million non-speech audio samples from the public waiting areas from December 2018 to July 2019. As said by Prof Rehman that during the initial validation they have defined that there is a strong correlation between coughing and influenza-related illness. The next step should be to test FluSense in other public areas and other geographic locations.
According to a business point of view, AI in health care is one of the booming industry and expected to have a cumulative growth. AI in health care is going to be a multi-billion dollar industry. The only concern related to AI is data breach and privacy but FluSense does not record any speech or visual related data apart from coughing. Overall a good initiative as well as a product by the research team of UMASS, Amherst, another good invention and innovation coming from the USA. Watch this space for more.