Israeli startup looks to harness AI to decode the ‘language’ of the brain

Israeli startup Hemispheric has unveiled Descartes, an artificial intelligence platform designed to decode the brain’s electrical activity and offer new ways to assess neurological and psychological conditions such as depression and post-traumatic stress disorder (PTSD). Developed by co-founders Gidi Littwin and Hagai Lalazar, both accomplished scientists with backgrounds in computational neuroscience and machine learning, the technology represents a major leap forward in understanding and quantifying brain function through non-invasive means. Littwin, who previously worked on deep learning solutions for Apple’s augmented reality device Vision Pro, left the company in 2020 to join forces with Lalazar. Together, they aimed to create a machine learning model capable of inferring brain function from electrical activity recorded outside the skull. Their goal was to eliminate the need for invasive procedures such as electrode implants, which are typically used in clinical settings to monitor brain activity. However, they faced a critical obstacle: the lack of sufficient high-quality brain data to train such a model effectively. To overcome this challenge, the duo established a vast global network of research sites spanning Asia, Israel, and Boston, recruiting over 100,000 paid volunteers. These participants contributed more than 250,000 hours of electroencephalogram (EEG) data, which was used to train the Descartes AI platform. The model incorporates 6 billion parameters, allowing it to analyze patterns in brainwave activity with unprecedented precision. This approach enables the system to identify subtle variations in neural responses, offering insights that traditional methods often fail to capture. Electroencephalography, or EEG, has long been the standard tool for measuring brain activity. It involves placing small electrodes on the scalp to record electrical impulses generated by neurons. While widely used, EEG has limitations in terms of spatial resolution and consistency across individuals. Brain signals can vary significantly from one person to another, making it difficult to translate raw data into meaningful clinical assessments. The Descartes platform aims to address these issues by leveraging advanced AI techniques to normalize and interpret brain activity with greater accuracy. According to Littwin, the key breakthrough came from recognizing that large-scale variability could actually be modeled. By analyzing data from thousands of participants, the team discovered that patterns emerge at a macro level, enabling the creation of a standardized framework for interpreting brain function. This allows clinicians to obtain reliable, reproducible measurements without requiring invasive interventions. The result is a system that can provide actionable insights in just 15 minutes, using a simple, wearable EEG headset that resembles a bicycle helmet. Users interact with the Descartes platform via a mobile application, engaging in tasks designed to elicit specific neural responses. As the system processes the data, it generates detailed reports that highlight potential indicators of mental health conditions. This information can then be used by healthcare professionals to guide diagnoses and tailor treatments more effectively. The startup emphasizes that the technology is not intended to replace existing diagnostic tools but rather to complement them by providing objective, data-driven assessments. With its focus on accessibility and non-invasiveness, the Descartes platform holds promise for transforming how brain-related disorders are understood and managed. As the startup continues to refine its algorithms and expand its dataset, the potential impact on both clinical practice and scientific research grows ever more profound.