Capture brain signals and convert them into electrical signals to achieve information transmission and control

Enabling industrial manufacturing, establishing cross-task algorithms for brain state monitoring of construction personnel; enabling military aviation, establishing cross-subject algorithms for cognitive assessment of firefighters; enabling healthcare and health, establishing cross-site algorithms for auxiliary diagnosis of brain diseases.

Figure 1: Brain-Computer Interface Research

1.1 Empower national infrastructure

Based on brain-computer interface technology and feature selection algorithms, a digital personnel safety supervision management system is established to monitor and warn the brain states such as concentration and fatigue of construction personnel for key national projects.

1.2 Empower healthcare

Collaborating with multiple hospitals including Tiantan Hospital and Xuanwu Hospital, quickly extracting functional neurology biomarkers, empowering multiple auxiliary diagnosis platforms, and currently serving as an auxiliary diagnostic tool for nearly ten thousand patients' diseases.

1.3 Empower national defense

The laboratory undertakes projects for units such as the Commission of Science and Technology of the Central Military Commission, applies the brain function operation mechanism to the selection and assessment of special position talents, and has served thousands of pilots and firefighters. The system has served thousands of firefighters in many provinces of our country, with an accuracy rate of cognitive state recognition ≥90% and an accuracy rate of state warning ≥99%.

Figure 2: Empower national defense

Inspired by the brain's neural mechanism and cognition, machine intelligence realized through the collaborative implementation of hardware and software

Inspired by neural diversity, design new neural structures; inspired by the consolidation mechanism of sleep memory, design new network module training strategies; inspired by decision-making mechanisms, design new network module training strategies.

Figure 3: Brain-inspired intelligence research

2.1 Cognitive function decoding research

Determine the distribution of brain functional networks, spatial network distribution, and multi-scale brain network distribution; explore the spatiotemporal evolution of brain function, decode spatiotemporal co-variation features, decode task-specific features; aggregate multi-modal brain function representations, redundant feature selection, multi-modal intelligent fusion.

Figure 4: Cognitive function decoding research

2.2 Cross-Site Online Disease Diagnosis Research

Insufficient data from a single site limits the performance of diagnostic models, the cost of collecting neuroimaging data is high, and it is difficult to obtain batch data in online diagnostic scenarios; the existing cross-site disease diagnosis covers various differences such as equipment and environment, different scanning equipment, protocols, and population distribution, resulting in significant differences in the distribution of cross-site data; existing methods require obtaining all data in the target domain, but in clinical diagnosis, most target samples arrive sequentially in an online form, and only a small number of target domain samples can be obtained in advance.

Through cross-site online disease diagnosis research, focusing on the cross-site online disease diagnosis scenario, and solving the problem of small sample size in the target domain.