A hybrid, artificial intelligence stethoscope that is effective in detecting cardiac and respiratory diseases | Image Credit: yodiyim – dodigiyam – stock.adobe.com.
The development of deep learning analytics that can analyze human body sounds for clinical purposes is a constant process in which the applications of these technologies are being studied. An article published in Sensors suggests that incorporating AI technology into the stethoscope has several benefits, including the ability to automatically provide and produce digital audio records and diagnostic results.
The effectiveness of intervention must be determined by early identification of cardiac and respiratory diseases. Although the stethoscope has been an effective tool for auscultation, it can still be subjective since doctors have had limited experience with auscular testing. Additionally, the process involves identifying and interpreting sounds in various heart and lung diseases, which could lead to misdiagnosis or mistreatment.
Efforts are being made to develop automated means of detecting sounds in clinical settings. The researchers have highlighted that previous studies have focused on training an independent model for diagnosing lung or heart sounds, but it is essential to have a model that can simultaneously detect abnormal lung and heart noises.
The study authors reported that investigators created a hybrid model to classify cardiac and respiratory dysfunction using convolutional neural network (CNN) and best discrepancy forest (BDF). They conducted experiments and found that the hybrid version was more effective than the current methods. Additionally, they developed eSync, enabling the use of NI products in developing countries.
Research revealed that the model can diagnose 11 types of heart and lung diseases with “satisfactory” performance and is capable of being deployed on a low-cost single-board computer. Two publicly accessible datasets were used for the experiment, one at the International Conference on Biomedical Health Informatics (ICBHI) 2017 in Barcelona, Spain. The dataset included audio recordings ranging from 10 seconds to 90 seconds, while the other contained 1000 records of different frequencies including normal, mitral, undulam, and MVR. Experimental results showed that this method could predict the accuracy
Although the hybrid model can detect both lung and heart diseases, researchers noted that problems with classification caused by many classes using unbalanced datasets are more likely to arise when fewer classes are used. The proposed hybrid version could perform better than models for 11 classes, but the study has not been tested in hospitals. They hope that additional funding will be necessary to produce more digital stethoscopes and collaborate with hospital staff for model testing and data acquisition.
Referring to:
Zhang M, Li M., Guo L. and Liu J. Developed a low-cost artificial life machine (AI-powered stethoscope) and lightweight prototype for detecting respiratory and cardiac diseases from lung and heart auscultation sounds. Sensors 2023,23-2591. doi: https://doi.org/10.3390/s23052591; Zhihong X. 2012.