Figure 4.
Autonomous microscopy and AR powered using Jetson TX2 and LattePanda allows extensive and descriptive screens to be performed with ease. (A–D) Schematic illustrating the AMCA working with a Jetson TX2 development board and AR optics. (A) The AMCA will control the automated acquisition on the microscope and will move the slide, imaging at different positions in sequence. If cells are detected in a location, the image will be stored and then the location optically sectioned in the “z” dimension until no more cells are detected. Once all the regions containing cells have been detected in a stage location, they are processed using a contiguous volume detection to find the 3D regions that encapsulate the individuals cells (see G). (B) Photo of the Jetson TX2 development board (left) on which the AMCA algorithm, object detection, hardware control, and image acquisition are running. On the right-hand side is the LattePanda Windows computer, which runs Windows-specific hardware control software and communicates with the Jetson TX2 via an ethernet cable. (C) AR allows the user to see the outputs of the analysis algorithm when viewing the sample down the microscope. The ROI generated from cellular detection can be visualized through the AR system as the microscope acquires images online or subsequently offline, when the user views areas of the sample that have already been processed. (D) View down the binocular eyepiece of the microscope where the AR graphics are overlaid with the light emitted from the sample (scale bar, ∼20 µm). (E–H) Preliminary screen of C127 cells stained with DAPI. (E) Low-resolution overview image of C127 cells acquired during screen. Scale bar, 200 µm. (F) Zoom area shown by green rectangle in E. Image areas are shown with detected cells bounded by colored boxes. Cells touching the image area boundaries are excluded in this analysis. Scale bar, 200 µm. (G) Left: Example image with bounding boxes representing discrete cellular classification from object detection algorithm and the color represents track linking with the contiguous volume detection (colored rectangles). Depiction of cell classifications tracked through the different z-slices; color border represents cells in G. (H) Summary violin plots calculated over the mean intensity values for the cells acquired during the screen. Three independent slides were screened and analyzed (green, red, and blue). The median value (3,216, 3,905, and 2,487, solid line), the lower quartile (2,891, 3,404, and 2,227, left dashed line) and the upper quartile (3,923, 4,825, and 3,055, right dashed line) for each slide (1–3), respectively.

Autonomous microscopy and AR powered using Jetson TX2 and LattePanda allows extensive and descriptive screens to be performed with ease. (A–D) Schematic illustrating the AMCA working with a Jetson TX2 development board and AR optics. (A) The AMCA will control the automated acquisition on the microscope and will move the slide, imaging at different positions in sequence. If cells are detected in a location, the image will be stored and then the location optically sectioned in the “z” dimension until no more cells are detected. Once all the regions containing cells have been detected in a stage location, they are processed using a contiguous volume detection to find the 3D regions that encapsulate the individuals cells (see G). (B) Photo of the Jetson TX2 development board (left) on which the AMCA algorithm, object detection, hardware control, and image acquisition are running. On the right-hand side is the LattePanda Windows computer, which runs Windows-specific hardware control software and communicates with the Jetson TX2 via an ethernet cable. (C) AR allows the user to see the outputs of the analysis algorithm when viewing the sample down the microscope. The ROI generated from cellular detection can be visualized through the AR system as the microscope acquires images online or subsequently offline, when the user views areas of the sample that have already been processed. (D) View down the binocular eyepiece of the microscope where the AR graphics are overlaid with the light emitted from the sample (scale bar, ∼20 µm). (E–H) Preliminary screen of C127 cells stained with DAPI. (E) Low-resolution overview image of C127 cells acquired during screen. Scale bar, 200 µm. (F) Zoom area shown by green rectangle in E. Image areas are shown with detected cells bounded by colored boxes. Cells touching the image area boundaries are excluded in this analysis. Scale bar, 200 µm. (G) Left: Example image with bounding boxes representing discrete cellular classification from object detection algorithm and the color represents track linking with the contiguous volume detection (colored rectangles). Depiction of cell classifications tracked through the different z-slices; color border represents cells in G. (H) Summary violin plots calculated over the mean intensity values for the cells acquired during the screen. Three independent slides were screened and analyzed (green, red, and blue). The median value (3,216, 3,905, and 2,487, solid line), the lower quartile (2,891, 3,404, and 2,227, left dashed line) and the upper quartile (3,923, 4,825, and 3,055, right dashed line) for each slide (1–3), respectively.

This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal