Hemispheric Aims to Turn Brain Scans Into 'Blood Tests' With $52 Million Boost

The promise of AI in healthcare often oscillates between revolutionary and speculative, and the latest venture from Hemispheric is positioning itself at the center of that tension. According to reporting from Wired, the startup—co-founded by Gidi Littwin and Hagai Lalazar—has raised $52 million to develop a frontier AI model designed to diagnose cognitive disorders by decoding electrical activity in the human brain.

Littwin, who previously worked at Apple as a co-inventor of FaceID and worked on hand-tracking for the Vision Pro, is applying the same massive data-collection philosophy to neurology that he used for biometric security. Wired reports that Hemispheric gathered a dataset consisting of 250,000 hours of brain data from 100,000 paid volunteers across Boston, Tel Aviv, and Asia. These subjects engaged in game-like activities designed to activate various brain regions, providing the raw material for deep-learning models to analyze brain function statistically, similar to how large language models process text.

From a clinical perspective, the goal is to replace the subjective nature of behavioral observations and questionnaires currently used to diagnose conditions such as Parkinson’s, Alzheimer’s, and depression. The proposed workflow involves a patient wearing a lightweight EEG headset for approximately 15 minutes while interacting with a tablet app. Hemispheric claims this AI-driven approach will allow clinicians to make diagnoses, predict the most effective interventions, and monitor patient progress. Lalazar told Wired he envisions the technology becoming as cheap and accessible as a blood test, distributed across hospitals, mental health clinics, and psychologists' offices.

While the company claims the model made accurate deductions when tested on subsets of people with depression, schizophrenia, and PTSD, the technology is still far from general clinical use. Wired notes that Hemispheric is currently conducting a clinical study to determine if the model can predict and diagnose Alzheimer’s. The company intends to submit its first product, focused on PTSD, to the FDA for approval in early 2027, with a potential public rollout later that year.

Beyond the software, the company is developing proprietary brain scanners. Littwin told Wired that traditional EEG devices were not built for deep learning, necessitating hardware specifically designed for machine learning data acquisition.

Funding for the venture comes from individual investors, including early Uber backer Howard Morgan, as well as Israeli and American venture capital firms. The company plans to use the capital to expand its U.S. workforce, pursue partnerships with pharmaceutical firms and governments, and scale its dataset to include millions of people.