The 4th Affiliated Hospital of Zhejiang University
Shatin Caring Centre of the Evangelical Lutheran Church of Hong Kong
Hong Kong has a high prevalence of heart failure (HF), with low public awareness of their symptoms and challenges in providing HF-related services in real-world settings, increasing the healthcare burden on society. Heart failure also carries high mortality with more than 50% of the patients dying at 5 years after being diagnosed to have heart failure. In addition, heart failure imposes a significant burden on our healthcare service with more than 20,000 hospitalizations related to heart failure each year. Heart disease is difficult to discover because the symptoms are not obvious. Regular testing is one of the best actions to provide people with early warning. However, regular testing is not feasible for ordinary citizens because most of them are not willing to undergo such testing regularly because they are unwilling to incur money and time to go to clinical sites. Thus, this project aims to develop a point-of-care and an intelligent screening method to provide early, non-invasive warning of potential heart failure and serve the grassroots in a continuous healthcare process. Also, with the success of this project, this method can provide timely assessment and better management of patients suspected of HF in primary care and hope to reduce the public health burden. This project will first extract Korotkoff-Sounds features and develop an acoustic features filter to decrease the number of features and increase the model accuracy. Secondly, an audio transformer with an acoustic (time-frequency) attention module will be developed to achieve more optimized audio modelling. In addition, since the audio transformer's acoustic attention mechanism allows the model to focus on different parts of the audio signal at different times and frequencies, it will reduce the effect of the coupled noise in capturing Korotkoff sounds mentioned in previous researches.