Live Cell Imaging of Cellular Dynamics in Poplar Wood Using Computational Cannula Microscopy

We report Computational Cannula Microscopy (CCM) for imaging living cells within poplar tree stems, with the long-term goal of observing cell-wall differentiation and functions. CCM overcomes depth limitations faced by conventional methods, employing a fiber-optic cannula to transport light, and machine-learning for image reconstruction. We demonstrate imaging of live cells, including cellular dynamics with an approximate spatial resolution of 3 to 5 𝜇m at a depth of 200 𝜇m to 500 𝜇m below the surface of the stem in a young living poplar branch with field-of-view of 55 𝜇m x 55 𝜇m using a cannula of core diameter = 50 𝜇m. The small cannula diameter offers the potential for minimal tissue damage upon insertion. We use the pix2pix model for image reconstructions and investigate its advantages and disadvantages over prior efforts using the U-Net. Image reconstruction at 0.29s/frame is obtained with a maximum average error in the reconstructed image 0.07. Lastly, we introduce a method to ascertain the confidence in the reconstructed images when no ground-truth is available.