There is a need to monitor patients with cancer of the head and neck postradiation therapy, as diminished swallowing activity can result in disuse atrophy and fibrosis of the swallowing muscles. This paper describes a flexible strain sensor comprising palladium nanoislands on single-layer graphene. These piezoresistive sensors were tested on 14 disease-free head and neck cancer patients with various levels of swallowing function: from nondysphagic to severely dysphagic. The patch-like devices detected differences in (1) the consistencies of food boluses when swallowed and (2) dysphagic and nondysphagic swallows. When surface electromyography (sEMG) is obtained simultaneously with strain data, it is also possible to differentiate swallowing vs nonswallowing events. The plots of resistance vs time are correlated to specific events recorded by video X-ray fluoroscopy. Finally, we developed a machine-learning algorithm to automate the identification of bolus type being swallowed by a healthy subject (86.4%. accuracy). The algorithm was also able to discriminate between swallows of the same bolus from either the healthy subject or a dysphagic patient (94.7% accuracy). Taken together, these results may lead to noninvasive and home-based systems for monitoring of swallowing function and improved quality of life.
Keywords: dysphagia; graphene; machine learning; nanoislands; strain sensor; wearable sensor.