Blepharospasm (BSP) is a focal dystonia marked by involuntary, excessive, and repetitive contractions of the muscles around the eyes. Traditionally, diagnosis of this condition has relied on manual clinical observation or video-based detection. Here, we present a class of technologies that combine artificial intelligence with soft wearable bioelectronics to enable automatic, quantitative detection of abnormal facial movements in BSP. This wearable system integrates soft membrane sensors for electrophysiological signal acquisition and automated pathological event detection using deep learning algorithms. The digital healthcare system enables robust detection of BSP-related pathological events across multiple behavioral categories, achieving classification accuracies of 82.5% for periocular spasm, 82.0% for blinking, and 84.4% for mouth movement behaviors. Event-level classifications from clinical studies indicate that BSP severity can be quantitatively estimated as an index of pathological event burden. Collectively, such an artificial intelligence-integrated wearable system may facilitate more objective and continuous detection of various diseases, advancing population medicine and public health. • A soft, wearable, portable biosensor system enables automated detection of pathological facial movements in BSP. • Hybrid 1D CNN–BiLSTM models discriminate periocular spasms, abnormal blinking, and mouth movements. • The wearable-AI system achieves over 80% accuracy in detecting pathological events. • Event-level detection may enable quantitative estimation of BSP symptom severity.
Choi et al. (Wed,) studied this question.