Researchers at Weill Cornell Medicine have successfully utilized artificial intelligence (AI) to map the visual functions in the human brain. By employing AI-selected natural images and AI-generated synthetic images, the team used these neuroscientific tools to investigate the visual processing areas of the brain. The study, published in Communications Biology, involved volunteers who viewed various images chosen or generated based on a model of the human brain. Through functional magnetic resonance imaging (fMRI), the researchers recorded the brain activity of the participants and discovered that the images indeed activated the target areas more effectively than control images.
Additionally, the researchers demonstrated that they could fine-tune their vision model for individual volunteers by using the image-response data. This adjustment ensured that the images created specifically for a particular individual resulted in better activation compared to images generated based on a general model. Dr. Amy Kuceyeski, a professor at Weill Cornell Medicine, expressed optimism about this novel approach to studying the intricacies of vision.
To conduct their research, the team employed a vast dataset comprising tens of thousands of natural images, along with corresponding fMRI responses from human subjects. They trained an artificial neural network (ANN), a type of AI system, using this dataset to model the visual processing system of the human brain. The model was then utilized to predict which images from the dataset would optimally activate targeted vision areas of the brain. In addition to this, the researchers utilized an AI-based image generator combined with the model to produce synthetic images that achieved the same objective.
The study involved six volunteers, whose fMRI responses to the images were recorded. The primary focus was on the participants’ responses in various visual processing areas. The results indicated that both the natural and synthetic images predicted as maximal activators significantly activated the targeted brain regions more than a control set of images that were selected or generated to be only average activators.
The team’s general objective was to systematically and impartially map and model the visual system, potentially using images that individuals might not encounter in their everyday lives. By combining AI and neuroscientific techniques, this research opens up new possibilities for understanding the complexities of vision.
In conclusion, researchers at Weill Cornell Medicine have made significant strides in utilizing AI to map visual functions in the human brain. Their study showcased the effectiveness of AI-selected natural images and AI-generated synthetic images as tools for investigating the visual processing areas. By recording brain activity through fMRI, the researchers established that these images successfully activated the target areas. Furthermore, the team demonstrated the ability to tailor the vision model for individual participants, resulting in even better image activation. This groundbreaking research offers a promising approach to studying the mechanics of vision.
