AI-Based Spatiotemporal Analysis Reveals Distinct Neuronal Markers of Morphine Response during Chronic Pain in Mice.

Treatment of chronic pain often relies on long term use of analgesics, such as opioids like morphine. However, the habit-forming natures of such substances can pose a significant risk. Furthermore, since chronic pain is developed through a maladaptive plastic change in neuronal networks, the effect of morphine on patients with chronic pain may be different than its effect on healthy patients. We previously developed and used an advanced complementary metal-oxide-semiconductor (CMOS) sensor to record video of ventral tegmental area (VTA) activity in response to acute pain and pain chronification using artificial intelligence (AI) in the form of a Vector Quantized Variational Autoencoder (VQ-VAE). The VQ-VAE (validation mean squared error 0.00732) identified distinct spatiotemporal motifs relating to noxious stimulus response in the mouse VTA. Here we extend this pre-trained network to investigate VTA response to this stimulus pattern on subsequent days during administration of either saline (first day) or morphine (on the following day) and quantify these changes in terms of discrete motifs. These motifs show distinct alteration in both spatial locale and time-response in both chronic pain model animals and sham animals. Furthermore, the spatiotemporal impact of morphine/administration day on the VTA is substantially altered in the animals with chronic pain. This difference is stratified across the different motifs, which show organized, spatially localized responses to the stimulus and how they change based on the experimental conditions. These motifs show how AI can fit complex signaling patterns to an ordered latent representation, and how analysis of these AI-based representations allows for nuanced interpretation of underlying pain-response dynamics.