Open-source software to calculate the static sciatic index automatically

TL;DR

This paper introduces an open-source tool that automatically calculates the static sciatic index in rat nerve injury studies, improving consistency and reducing labor.

Contribution

An open-source machine learning tool for automatic and consistent calculation of the static sciatic index in nerve injury research.

Findings

The model's outputs showed a nerve regeneration profile comparable to manual measurements.

The model outperformed manual methods with a much tighter standard deviation.

The software was tested on two datasets with different experimental setups.

Abstract

The static sciatic index is commonly used in rat models of nerve crush injury to quantify functional recovery from new therapies under evaluation. However, it is challenging to standardize these measurements across different investigations, and the process is labor intensive.

A new machine learning method was previously developed that performs these measurements automatically and consistently. Here, the approach is tested using two data sets that use different experimental setups, and end-user requirements are evaluated.

The model’s outputs presented a nerve regeneration profile comparable to the manual measurements and outperformed the latter by having a much tighter standard deviation (± 5- ± 10 compared to ± 10 - ± 50).

An inexpensive automatic tool that can perform functional analysis for nerve repair research was developed and tested. The software is available open source to facilitate its dissemination and use in quantifying recovery from peripheral nerve crush injury.

The rat sciatic nerve crush injury model is a widely used animal model for testing new therapies under development for treating nerve injuries. A commonly used metric to evaluate functional recovery is the static sciatic index (SSI), which is a weighted ratio of distances between the rat’s digits. However, it requires careful measurement of the distance between digits to ensure consistency. This must be done for several rats over numerous time points, becoming a labor-intensive task. High accuracy and consistency in these measurements are important to ensure that the treatment is properly assessed. Here, a new automatic method is proposed that makes use of a machine learning architecture to label the digits and perform the SSI measurements from hundreds of video frames. The detailed computational development of the algorithm has previously been published. Here, an accessible guide for installing and using the model is presented. Additionally, the model is evaluated using two data sets from different setups, illustrating the necessary care required in the data acquisition protocol to ensure the model performs as intended. Furthermore, criteria to avoid common errors are presented. The model is available as an open-source tool, which we hope facilitates its dissemination and uptake to quantitatively assess functional recovery post nerve crush injury.

Figures