In an Aging society, it is common for people and their parents do not live together. However, if their parents have an accident at home, they often cannot provide assistance upon knowing the accident immediately. Therefore, we develop CarePLUS. The CarePLUS encrypts the personal information and uploads image to the cloud through the network. After analyzed by the AI model on the cloud, the AI model returns the result to the device and interacts with the user.
Correct diagnosis of voice disorders require specific endoscopy performed by experience doctors. With the progress of computation capability, various machine learning algorithms have been applied to interpret complex clinical and laboratorial data. Our research group combined expert of phoniatrics, electric engineering, signal processing and artificial intelligence. We published the first research using deep learning approach to detect pathological voice in 2018. Our results exceeds all the existing classifiers, reached an accuracy of 99.14%. We also hosted the FEMH Voice Disorder Detection Competition on 2018 IEEE International Conference on Big Data, which was the first of its kind. More than 100 teams from 27 different countries participated in our competition. To breakthrough current research limitations of classification of voice disorders, we proposed multi-modal approaches to integrate demographic feature, acoustic signals, and clinical symptoms. Our results achieved an accuracy of 87.26% for the classification between phonotraumatic lesions, glottic neoplasm, and vocal palsy.
NTU Medical Genie Lifestyle and Environment Platform
The product is mainly composed of wearable devices, IoT environmental sensors, deep learning technology and healthcare platform. It can collect and monitor the user's lifestyle and environment automatically, and assist medical staff in making decisions when an emergency occurs. Currently, chronic obstructive pulmonary disease (COPD) is the first success story. COPD is the fourth leading cause of death in the world. The frequency and severity of acute exacerbation are highly correlated with mortality. Our AECOPD prediction model can predict the possibility of acute exacerbation in the next 7 days based on daily physical activity data and environmental data. We have achieved 93.5 % on accuracy for the task of predicting whether a patient will suffer an acute exacerbation within the next 7 days.