AI-powered remote monitoring of brain responses to clear and incomprehensible speech via speckle pattern analysis

• Published in: Journal of biomedical optics Vol. 30; no. 6; p. 067001
• Main Authors: Segal, Natalya, Kalyuzhner, Zeev, Agdarov, Sergey, Beiderman, Yafim, Beiderman, Yevgeny, Zalevsky, Zeev
• Format: Journal Article
• Language: English
• Published: United States Society of Photo-Optical Instrumentation Engineers 01.06.2025; SPIE
• Subjects: Adult; Artificial Intelligence; Brain; Brain - diagnostic imaging; Brain - physiology; Brain research; Computational linguistics; Deep Learning; Female; Humans; Image Processing, Computer-Assisted - methods; Imaging systems; Injuries; Language processing; Male; Natural language interfaces; Neural Networks, Computer; Patient monitoring equipment; Sensing; Speech - physiology; Young Adult; United Kingdom; Sweden; Wernicke’s area; AI-driven neuroimaging; speech response detection; laser speckle patterns; remote brain monitoring; photonic brain sensing; noninvasive brain analysis
• ISSN: 1083-3668; 1560-2281; 1560-2281
• DOI: 10.1117/1.JBO.30.6.067001

Functional magnetic resonance imaging provides high spatial resolution but is limited by cost, infrastructure, and the constraints of an enclosed scanner. Portable methods such as functional near-infrared spectroscopy and electroencephalography improve accessibility but require physical contact with the scalp. Our speckle pattern imaging technique offers a remote, contactless, and low-cost alternative for monitoring cortical activity, enabling neuroimaging in environments where contact-based methods are impractical or MRI access is unfeasible. We aim to develop a remote photonic technique for detecting human brain cortex activity by applying deep learning to the speckle pattern videos captured from specific brain cortex areas illuminated by a laser beam. We enhance laser speckle pattern tracking with artificial intelligence (AI) to enable remote brain monitoring. In this study, a laser beam was projected onto Wernicke's area to detect brain responses to a clear and incomprehensible speech. The speckle pattern videos were analyzed using a convolutional long short-term memory-based deep neural network classifier. The classifier distinguished brain responses to a clear and incomprehensible speech in unseen subjects, achieving a mean area under the receiver operating characteristic curve (area under the curve) of 0.94 for classifications based on at least 1 s of input. This remote method for distinguishing brain responses has practical applications in brain function research, medical monitoring, sports, and real-life scenarios, particularly for individuals sensitive to scalp contact or headgear.