# Modeling strategies in non-invasive spinal stimulation: perspectives on state-of-the-art

## TL;DR

This paper reviews current computational models for non-invasive spinal stimulation and highlights the need for personalized approaches to improve treatment effectiveness.

## Contribution

The paper provides an overview of current computational modeling strategies for tsDCS and identifies gaps in personalization using MRI-based models.

## Key findings

- tsDCS modulates spinal reflexes and sensorimotor responses as shown through electromyography.
- Current modeling strategies rely on template models but lack personalization from MRI-based segmentation.
- Future work should focus on developing semi-automatic pipelines for realistic spinal cord modeling.

## Abstract

Non-invasive Spinal Stimulation (NISS) is of increasing interest for clinicians addressing spinal dysfunctions, such as spasticity, chronic pain and hypotonia. NISS can be an alternative when surgically-implanted stimulators or pharmacological therapy are not compatible nor viable. Trans-spinal direct current stimulation (tsDCS) is one NISS strategy that delivers direct currents (DC) of low intensity (1–4 mA) through large electrodes (8–25 cm2) placed over selected targets in the vertebral column. Since 2008, tsDCS researchers have build-up evidence regarding modulation of spinal reflexes and sensorimotor responses measured by electromyography. Biophysical constructs based on computational numerical methods provide a well-grounded framework to determine the most effective protocols designed according to each patient’s needs. Additionally, models can work as theoretical labs to investigate how the EF profile induced by stimulation can relate with the changes observed in spinal responses. The accuracy of predictions in tsDCS biophysical constructs strongly rely on how realistic are the digital twins of the spine. The main strategy used is to adapt accurate template models to have a fine description of spinal structures down to the millimeter resolution. However, this strategy lacks the personalized approach of MRI-based realistic models. This is due to the fact that development of pipelines for semi-automatic segmentation of the spinal cord is still in its early stages. This work aims to discuss the current state-of-the art regarding computational constructs of tsDCS, what is known on its effects on spinal networks, based on combined modeling-experimental approaches, and what lies ahead for a more targeted and personalized application.