A deep unrolled neural network for real-time MRI-guided brain intervention

Accurate navigation and targeting are critical for neurological interventions. Real-time image guidance improves surgical planning, and MRI is ideal for pre- and intra-operative imaging. Balancing spatial and temporal resolution is challenging for real-time interventional MRI (i-MRI). This study proposes LSFP-Net, a deep unrolled neural network, for real-time i-MRI reconstruction. Integrating LSFP-Net with an MR-compatible interventional device in a 3T MRI scanner creates a real-time MRI-guided brain intervention system. Phantom and cadaver studies evaluated the system, achieving 2D/3D real-time i-MRI with temporal resolutions of 80/732.8 ms, latencies of 0.4/3.66 s, and 1 × 1 mm² in-plane spatial resolution. The method enables real-time monitoring of remote-controlled brain intervention and shows potential for integration into diagnostic scanners for image-guided neurosurgery.

Zhao He, Ya-Nan Zhu, Yu Chen, Yi Chen, Yuchen He, Yuhao Sun, Tao Wang, Chengcheng Zhang, Bomin Sun, Fuhua Yan, Xiaoqun Zhang, Qing-Fang Sun, Guang-Zhong Yang, Yuan Feng