Autofluorescent imprint of chronic constriction nerve injury identified by deep learning

• Published in: Neurobiology of disease 2021-12, Vol.160, p.105528-105528, Article 105528
• Main Authors: Gosnell, Martin E., Staikopoulos, Vasiliki, Anwer, Ayad G., Mahbub, Saabah B., Hutchinson, Mark R., Mustafa, Sanam, Goldys, Ewa M.
• Format: Article
• Language: English
• Subjects: Allodynia; Animals; Autofluorescence imaging; Chronic constriction injury (CCI); Chronic pain; Constriction; Deep Learning; Female; Fluorescent Antibody Technique; Hyperalgesia - metabolism; Male; Mice; Nerve injury; Optical Imaging; Peripheral Nerve Injuries - metabolism; Sciatic Nerve - injuries; Sciatic Nerve - metabolism; Spinal cord
• ISSN: 0969-9961; 1095-953X
• DOI: 10.1016/j.nbd.2021.105528

Our understanding of chronic pain and the underlying molecular mechanisms remains limited due to a lack of tools to identify the complex phenomena responsible for exaggerated pain behaviours. Furthermore, currently there is no objective measure of pain with current assessment relying on patient self-scoring. Here, we applied a fully biologically unsupervised technique of hyperspectral autofluorescence imaging to identify a complex signature associated with chronic constriction nerve injury known to cause allodynia. The analysis was carried out using deep learning/artificial intelligence methods. The central element was a deep learning autoencoder we developed to condense the hyperspectral channel images into a four- colour image, such that spinal cord tissue based on nerve injury status could be differentiated from control tissue. This study provides the first validation of hyperspectral imaging as a tool to differentiate tissues from nerve injured vs non-injured mice. The auto-fluorescent signals associated with nerve injury were not diffuse throughout the tissue but formed specific microscopic size regions. Furthermore, we identified a unique fluorescent signal that could differentiate spinal cord tissue isolated from nerve injured male and female animals. The identification of a specific global autofluorescence fingerprint associated with nerve injury and resultant neuropathic pain opens up the exciting opportunity to develop a diagnostic tool for identifying novel contributors to pain in individuals. •First validation of hyperspectral imaging as a tool to differentiate with chronic constriction injury vs non-injured animals.•Technique is fully biologically unsupervised.•The auto- fluorescent signals associated with chronic constriction injury formed specific microscopic size regions.•We identified that the area of these regions could differentiate spinal cord tissue isolated from male and female animals, both with chronic constriction injury.•The autofluorescence fingerprint discovered here may help identify novel contributors to chronic pain.