Search

In the coming decade, artificial intelligence systems will continue to improve and revolutionise every industry and facet of human life. Designing effective, seamless and symbiotic communication paradigms between humans and AI agents is increasingly important. This paper reports a novel method for human-AI interaction by developing a direct brain-AI interface. We discuss a novel AI model, called MindSpeech, which enables open-vocabulary, continuous decoding for imagined speech. This study focuses on enhancing human-AI communication by utilising high-density functional near-infrared spectroscopy (fNIRS) data to develop an AI model capable of decoding imagined speech non-invasively. We discuss a new word cloud paradigm for data collection, improving the quality and variety of imagined sentences generated by participants and covering a broad semantic space. Utilising a prompt tuning-based approach, we employed the Llama2 large language model (LLM) for text generation guided by brain signals. Our results show significant improvements in key metrics, such as BLEU-1 and BERT P scores, for three out of four participants, demonstrating the method's effectiveness. Additionally, we demonstrate that combining data from multiple participants enhances the decoder performance, with statistically significant improvements in BERT scores for two participants. Furthermore, we demonstrated significantly above-chance decoding accuracy for imagined speech versus resting conditions and the identified activated brain regions during imagined speech tasks in our study are consistent with the previous studies on brain regions involved in speech encoding. This study underscores the feasibility of continuous imagined speech decoding. By integrating high-density fNIRS with advanced AI techniques, we highlight the potential for non-invasive, accurate communication systems with AI in the near future.

In the coming decade, artificial intelligence systems are set to revolutionise every industry and facet of human life. Building communication systems that enable seamless and symbiotic communication between humans and AI agents is increasingly important. This research advances the field of human-AI interaction by developing an innovative approach to decode imagined speech using non-invasive high-density functional near-infrared spectroscopy (fNIRS). Notably, this study introduces MindGPT, the first thought-to-LLM (large language model) system in the world.

Targeted light delivery into biological tissue is needed in applications such as optogenetic stimulation of the brain and in vivo functional or structural imaging of tissue. These applications require very compact, soft, and flexible implants that minimize damage to tissue. Here, we demonstrate a novel implantable photonic platform based on a high-density, flexible array of ultracompact (30 ${\mu}$m x 5 ${\mu}$m), low-loss (3.2 dB/cm at ${\lambda}$ = 680 nm, 4.1 dB/cm at ${\lambda}$ = 633 nm, 4.9 dB/cm at ${\lambda}$ = 532 nm, 6.1 dB/cm at ${\lambda}$ = 450 nm) optical waveguides made of biocompatible polymers Parylene C and polydimethylsiloxane (PDMS). This photonic platform features unique embedded input/output micro-mirrors that redirect light from the waveguides perpendicularly to the surface of the array for localized, patterned illumination in tissue. This architecture enables the design of a fully flexible, compact integrated photonic system for applications such as in vivo chronic optogenetic stimulation of brain activity., Comment: 20 pages, 7 figures, submitted to Nature Microsystems and Nanoengineering

We show that significant water wave amplification is obtained in a water resonator consisting of two spatially separated patches of small-amplitude sinusoidal corrugations on an otherwise flat seabed. The corrugations reflect the incident waves according to the so-called Bragg reflection mechanism, and the distance between the two sets controls whether the trapped reflected waves experience constructive or destructive interference within the resonator. The resulting amplification or suppression is enhanced with increasing number of ripples, and is most effective for specific resonator lengths and at the Bragg frequency, which is determined by the corrugation period. Our analysis draws on the analogous mechanism that occurs between two partially reflecting mirrors in optics, a phenomenon named after its discoverers Charles Fabry and Alfred Perot., Comment: 8 pages, 6 figures, published in Phys. Rev. E 92, 043015

Conventional optical lenses have been used to focus light from outside without disturbing the medium. The focused spot size is proportional to the focal distance in a conventional lens, resulting in a tradeoff between depth of penetration in the target medium and spatial resolution. We have shown that virtual ultrasonically sculpted gradient-index (GRIN) optical waveguides can be formed in the target medium to guide and steer light without disturbing the medium. Here, we demonstrate that such virtual waveguides can relay an externally focused beam of light through the medium beyond the focal distance of an external physical lens to extend the penetration depth without compromising the spot size. Moreover, the spot size can be tuned by reconfiguring the virtual waveguide. We show that these virtual GRIN waveguides can be formed in transparent as well as turbid media to enhance the confinement and contrast ratio of the focused beam of light at the target location. This method can be extended to realize complex optical systems of external physical lenses and in situ virtual waveguides to extend the reach and flexibility of optical methods., Comment: 11 pages, 6 figures

Cell detection and segmentation is fundamental for all downstream analysis of digital pathology images. However, obtaining the pixel-level ground truth for single cell segmentation is extremely labor intensive. To overcome this challenge, we developed an end-to-end deep learning algorithm to perform both single cell detection and segmentation using only point labels. This is achieved through the combination of different task orientated point label encoding methods and a multi-task scheduler for training. We apply and validate our algorithm on PMS2 stained colon rectal cancer and tonsil tissue images. Compared to the state-of-the-art, our algorithm shows significant improvement in cell detection and segmentation without increasing the annotation efforts.

We present a novel approach for achieving tightly concentrated optical field by a hybrid photonic-plasmonic device in an integrated platform, which is a triangle-shaped metal taper mounted on top of a dielectric waveguide. This device, which we call a plasmomic light concentrator (PLC), can achieve side-coupling of light energy from the dielectric waveguide to the plasmonic region and light focusing into the apex of the metal taper(at the scale ~10nm) at the same time. For demonstration, we numerically investigate a PLC, which is a metal (Au) taper on a dielectric (Si3N4) waveguide at working wavelengths around 800nm. We show that three major effects (mode beat, nanofocusing, and weak resonance) interplay to generate this light concentration phenomenon and govern the performance of the device. By coordinating these effects, the PLC can be designed to be super compact while maintaining high efficiency over a wide band. In particular, we demonstrate that under optimized size parameters and wavelength a field concentration factor (FCF), which is the ratio of the norm of the electric field at the apex over the average norm of the electric field in the inputting waveguide, of about 13 can be achieved with the length of the device less than 1um for a moderate tip radius 20nm. Moreover, we show that a FCF of 5-10 is achievable over a wavelength range 700-1100nm when the length of the device is further reduced to about 400nm., Comment: 14 pages, 9 figures