Search for: All records

Deep-brain microscopy is strongly limited by the size of the imaging probe, both in terms of achievable resolution and potential trauma due to surgery. Here, we show that a segment of an ultra-thin multi-mode fiber (cannula) can replace the bulky microscope objective inside the brain. By creating a self-consistent deep neural network that is trained to reconstruct anthropocentric images from the raw signal transported by the cannula, we demonstrate a single-cell resolution (< 10μm), depth sectioning resolution of 40 μm, and field of view of 200 μm, all with green-fluorescent-protein labelled neurons imaged at depths as large as 1.4 mm from the brain surface. Since ground-truth images at these depths are challenging to obtain in vivo, we propose a novel ensemble method that averages the reconstructed images from disparate deep-neural-network architectures. Finally, we demonstrate dynamic imaging of moving GCaMp-labelledC.elegansworms. Our approach dramatically simplifies deep-brain microscopy.

We experimentally demonstrate a camera whose primary optic is a cannula/needle ($\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{m}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{r}=0.22\mathrm{m}\mathrm{m}$and$\mathrm{l}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}=12.5\mathrm{m}\mathrm{m}$) that acts as a light pipe transporting light intensity from an object plane (35 cm away) to its opposite end. Deep neural networks (DNNs) are used to reconstruct color and grayscale images with a field of view of 18° and angular resolution of$\sim <\#comment/>{0.4}^{\circ <\#comment/>}$. We showed a large effective demagnification of$127\times <\#comment/>$. Most interestingly, we showed that such a camera could achieve close to diffraction-limited performance with an effective numerical aperture of 0.045, depth of focus$\sim <\#comment/>16\text{µ<\#comment/>}\mathrm{m}$, and resolution close to the sensor pixel size (3.2 µm). When trained on images with depth information, the DNN can create depth maps. Finally, we show DNN-based classification of the EMNIST dataset before and after image reconstructions. The former could be useful for imaging with enhanced privacy.

Computational cannula microscopy (CCM) is a high-resolution widefield fluorescence imaging approach deep inside tissue, which is minimally invasive. Rather than using conventional lenses, a surgical cannula acts as a lightpipe for both excitation and fluorescence emission, where computational methods are used for image visualization. Here, we enhance CCM with artificial neural networks to enable 3D imaging of cultured neurons and fluorescent beads, the latter inside a volumetric phantom. We experimentally demonstrate transverse resolution of ∼6µm, field of view ∼200µm and axial sectioning of ∼50µm for depths down to ∼700µm, all achieved with computation time of ∼3ms/frame on a desktop computer.

Computational cannula microscopy is a minimally invasive imaging technique that can enable high-resolution imaging deep inside tissue. Here, we apply artificial neural networks to enable real-time, power-efficient image reconstructions that are more efficiently scalable to larger fields of view. Specifically, we demonstrate widefield fluorescence microscopy of cultured neurons and fluorescent beads with a field of view of 200 µm (diameter) and a resolution of less than 10 µm using a cannula of diameter of only 220 µm. In addition, we show that this approach can also be extended to macro-photography.