Dream Decoder DreamScope Achieves 85% Accuracy: Sleep Brain Signals First Translated in Real-Time

Japan's RIKEN institute published a paper on April 5 describing its DreamScope system, which captures brain signals during REM sleep through a 64-channel EEG cap and decodes them into low-resolution visual images. In double-blind tests with 200 subjects, the system achieved 85.3% accuracy in identifying object categories appearing in dreams (such as faces, buildings, and natural landscapes).

DreamScope's core is a visual reconstruction model trained on large-scale fMRI-EEG data, learning the mapping between visual stimuli and brain signals during waking states, then transferring this mapping to sleep states. The research team acknowledges the current system can only identify broad visual categories and cannot restore specific details or emotional content from dreams.

Paper corresponding author Professor Yukiyasu Kamitani stated that potential applications include PTSD dream analysis and sleep quality assessment, but the team explicitly opposes using the technology for "mind reading" or unauthorized consciousness monitoring. The ethics review committee required all experimental data to be destroyed immediately after the study concludes.

The paper was published in the April issue of Nature Neuroscience, sparking intense debate in both neuroscience circles and privacy advocacy organizations.