The technique uses the brain scans of young adult participants who enter an fMRI over a period of 18 hours to look at pictures from a dataset of 160,000 images for nine seconds each. The images range from everyday buildings to sports activities like baseball or wildlife including giraffes and swans.

The signals are then put through “MinD-Vis”, an AI model, to train it to associate certain brain patterns with particular image features like colour, shape, texture and semantics.

In the final stage, it can recover unseen visual stimuli from new images that participants were shown in the fMRI, based on analysis of their brain activity.

“In other words, Mind-Vis is able to read and reconstruct images from our minds,” said Ms Chen, who said the decoded pictures were consistent. While not reproduced with 100 per cent accuracy, they are recognizable matches and significantly out-performed previous experiments of a similar nature.

“Decoding reaches vital information that plays a role in understanding how our brain processes and interprets the world around us. It helps researchers to visualize and unlock the mystery of the brain and a deeper comprehension of its complex functions,” she said.

Technology could transform life for disabled

The technology within a few years could open up a new world for people with disabilities, allowing them to interact with a more advanced machine that can better mimic human cognitive and decision-making processes in real-time.

“Say, for example, if people are not able to type then they can just imagine a sentence … a sentence that he is thinking about that we are able to decode. That’s the future goal,” she said.

Helen Zhou, an associate professor at the NUS Center for Sleep and Cognition, predicted it could take a decade to finetune the technology to reach this goal.

For the concept to be commercialized, it would also require a shift away from unwieldy equipment like fMRI machines to a portable device, such as an EEG (electroencephalogram) headband, which uses sensors to detect brain activity, she said.

The aim would be to develop a machine capable of real-time decoding that a customer could take home, connect to WiFi, and easily tune it to their own needs. “That’s the dream,” said Ms Zhou.

But she conceded that guardrails were needed to prevent the exploitation of such technological advances.