Raman spectroscopy for minimally invasive spinal nerve detection

Abstract

Minimally invasive spine surgery (MISS) provides both patients and surgeons with huge benefits. However, at the same time, it is still experiencing obstacles, such as confined working space, the absence of force feedback information and lack of visualization etc. The spinal cord and spinal nerves connect the brain with the rest of the body. They are quite vulnerable and small damages on them may lead to severe body dysfunction. This paper presents that using Raman spectroscopy as an analytical tool to identify spinal cord and spinal nerves from surrounding tissues for MISS purpose. Ex vivo Raman experiments is done on swine backbone samples with a fiberoptic Raman sensing system to examine the viability of this work. A total number of 750 raw spectra of bone, fat, muscle and spinal cord are obtained and data pre-processing procedures are employed before multivariate analysis. Principal Component Analysis (PCA) are then applied for data dimensionality reduction. Finally, classification model (together with 10-fold cross validation) using linear discriminant analysis (LDA) for multiple classes is built based on the PCA results and an overall accuracy of 93.1% is reached.