Fundamental study on a method for evaluating swallowing function using swallowing pattern images

Abstract

Recently, dysphagia associated with aging has become more a serious concern. The current study therefore sought to develop a novel method for evaluating swallowing function that focused on changes in the movement of swallowing-related organs, which slightly occur depending on swallowing conditions, using surface electromyography (sEMG) signals from the suprahyoid and infrahyoid muscles. A total of 15 elderly individuals without any history of dysphagia and 15 healthy young individuals were included herein. Initially, using two 22-channel electrodes, sEMG signals from suprahyoid and infrahyoid muscles were measured during the following four swallowing conditions, which combines two bolus volumes and two techniques: 1-mL normal swallow (NS1), 6-mL normal swallow (NS6), 1-mL effortful swallow (ES1) and 6-mL effortful swallow (ES6). sEMG signals from the suprahyoid and infrahyoid muscles were individually converted into swallowing pattern images. After extracting 8,192-dimensional feature vector from both images using the pretrained convolutional neural network AlexNet, feature vector was dimensionally compressed to three dimensions using kernel principal component analysis. Finally, we calculated the normalized Euclidean distance between each feature vector under different swallowing conditions and compared the differences in swallowing pattern changes between young and elderly individuals. Accordingly, our results found that significant differences in the normalized Euclidean distance NED(NS1, NS6) for normal swallowing at varying bolus volumes and in the normalized Euclidean distance NED(NS1, ES1), NED(NS6, ES6) at varying swallowing strengths. These significant differences between young and elderly individuals indicate differences in their ability to change swallowing patterns depending on swallowing conditions, and they may reflect an age-related decline in swallowing function. The evaluation of swallowing function using the proposed method could be applied to the early detection of reduced swallowing function or a state of frailty in aged individuals because it is noninvasive and can be performed at the bedside.

Keywords

• swallowing
• image recognition
• multichannel semg signals
• suprahyoid muscles
• infrahyoid muscles