We believe in open, reproducible science. All software and datasets below are freely available under permissive licenses. If you use our work, please cite the associated paper where applicable.
Source
Open-source code, datasets, and tools released by NIMGroup.
NIM-BCI
Software
A PyTorch-based open-source toolbox for EEG-based brain-computer interface research, covering signal preprocessing, feature extraction, neural decoding, and closed-loop experiment control. Includes reference implementations for SSVEP, P300, and motor imagery paradigms.
FedMedSeg
Software
Federated learning framework for heterogeneous multi-site medical image segmentation. Handles non-IID scanner protocols and label distributions across institutions without sharing raw patient data. Achieves <2% Dice gap vs. centralized training on brain tumor benchmarks.
NIMNav
Software
Real-time brain shift compensation library for intraoperative neuronavigation. Fuses intraoperative ultrasound with preoperative MRI via a deformable registration network, achieving sub-millimeter target registration error (<0.71 mm) on clinical neurosurgical cases.
NIMPath
Software
Reinforcement learning environment for patient-specific neurosurgical path planning. Generates simulation scenes from segmented MRI/CT volumes and trains trajectory policies that avoid critical vascular and eloquent cortex structures. Compatible with OpenAI Gym API.
THINGS-NIM-EEG
Dataset
Large-scale EEG dataset for image-concept decoding research. Contains 64-channel EEG recordings from 20 healthy participants viewing 10,000 object images drawn from the THINGS concept database. Includes preprocessed epochs, ICA components, and CLIP semantic embeddings.