Deep learning method helps to improve research into spinal ganglia
2025
University Hospital Würzburg, Wuerzburg, Germany
Dorsal root ganglia (DRG) are involved in processing sensation and pain, but their exact function is not fully understood. New magnetic resonance imaging (MRI) techniques allow for the study of DRG within the body. Certain DRG measurements in MRI, such as volume and T2w signal, have been identified as potential indicators (biomarkers) that may be related to biochemical and genetic factors as well as neuropathic pain. To better utilize these biomarkers, automated methods for evaluating DRG images are needed, as DRG are currently mostly marked manually on the images. In this study, such an automated method based on deep learning was developed. A computer network (CNN) was trained with the nnU-Net software to identify DRG on detailed 3D MRI images (220 DRGs). The automated markings of the DRG were similarly accurate to the manual ones (accuracy of 0.89 vs. 0.87) and 10 times faster. The method was tested on patients with Fabry disease, a disease in which the DRG change. The computer network was able to detect the known changes of the DRG in this disease. Thus, an automated method for marking the DRG in MRI images was developed, and it was shown that it can be used to study DRG changes, e.g., in Fabry disease.
Automated segmentation of the dorsal root ganglia in MRI
Magnus Schindehütte
Added on: 05-02-2025
[1] https://www.sciencedirect.com/science/article/pii/S1053811925001910?via%3Dihub
