Spinal cord injury (SCI) is a profound trauma that significantly impacts the capability of affected individuals to generate functional standing and locomotor movements. Unfortunately, modern medicine has yet to find a definitive cure for SCI, leaving many patients dependent on wheelchairs for the rest of their lives.

Epidural Electrical Stimulation: Rising Hope for SCI Patients

Recently, spatio-temporal epidural electrical stimulation (EES) - a neuromodulation technique empirically validated effective for pain relief - has emerged as a promising therapeutic approach for SCI rehabilitation. By precisely targeting the posterior (sensory) nerve roots and activating individual dorsal roots in a timed sequence, spatio-temporal EES can naturally mimic the activation patterns of the spinal cord during the gait cycle, thereby improving functional outcomes.  By modulating specific motor neuron pools, the therapy demonstrates superior rehabilitation performance.

Lumbosacral spine MRI dataset: Gaining Insights into Spinal Nerve Root Structure

For spatio-temporal EES to be effective, accurate identification and localization of individual spinal nerve roots are crucial. Moreover, translating this technique into clinical practice is often challenged by the variability of the human spinal cord structure across individuals. Recognizing the scarcity of detailed datasets in this area, we introduce an open-access lumbosacral spine MRI dataset, meticulously capturing spinal nerve root structures.

This dataset, collected from 14 healthy adult volunteers (2 females and 12 males; Age: 23.21 ± 0.89  years; Height: 175.43 ± 8.22 cm; Weight: 71.14 ± 11.72  kg) at the Zhangjiang International Brain Imaging Center of Fudan University, was acquired using a 3T whole-body MRI system (MAGNETOM Prisma, Siemens Healthineers, Erlangen, Germany). Three MRI sequences were employed to visualize different spinal structures: the T2-TSE sequence defined the spinal cord contour, the DESS sequence highlighted ganglion localization and the CISS sequence distinctly captured the distribution of spinal nerve roots in the lumbosacral region. The complete imaging protocol, including preparation and localization, lasted approximately 1 hour.

Based on the acquired MRI images, a derivative annotation-driven lumbosacral spine model was constructed for each subject, incorporating manual adjustments to rectify any intersection issues. The Blender script used for modeling can be accessed via GitHub (https://github.com/Joshua-M-maker/SpineNerveModelGenerator).

Visualizing Human Lumbosacral Models: A Step Toward Advancing SCI Rehabilitation

The dataset is openly accessible on Figshare (https://doi.org/10.6084/m9.figshare.c.7372564). It includes T2-TSE, DESS, and CISS MRI sequences, annotation markers detailing the trajectories of spinal cord roots, ganglion localizations, and complete lumbosacral models.

This dataset lays the groundwork for innovative research in SCI rehabilitation. The annotated lumbosacral model can support tailored EES therapy simulations and provide insights into the variability of the human lumbosacral spine. Moreover, the detailed annotations of nerve roots in each MRI slice can facilitate the development of machine learning and deep learning models for automated nerve root detection. The reconstructed 3D spine models also offer potential for training end-to-end deep learning models that can directly reconstruct 3D models from MRI data.

Notably, the dataset features data collected from healthy subjects. As demonstrated in prior studies,  the overall morphology of the spinal cord in healthy subjects does not substantially differ from that of an SCI patient. Therefore, this dataset serves as a resource for evaluating advanced algorithms on healthy subjects, with the expectation that well-generalized algorithms will perform effectively on patient data.

We invite researchers and clinicians to explore this dataset, contribute to the advancement of SCI treatment, and join us on this journey to help individuals walk again.