(A-8) Using Smart Imaging of Micrographs Process and Labeling for Integrated Image Screening and Analysis.

Introduction:: Microscope-based image capture and analysis remains imperative in research and clinical settings. Current methods for quality analysis of micrographs are powerful and have immense value. However, many of the processes lack user functionality and reproducibility. These are some shortcomings where our Smart Imaging of Micrographs Process and Labeling (SIMPL) system aims to improve. SIMPL is an open-source and automatic image capture software that promotes reproducibility and automatic reporting for high throughput applications.

Materials and Methods:: Leveraging Jupyter Notebook and the R program on a Raspberry Pi, SIMPL is designed to perform automatic image capture to strengthen image quality and take micrographs of full samples. An Arducam camera sensor and Zeiss 40x objective lens constitute the optical portion of SIMPL. A stepper motor and custom Arduino firmware automatically locates and adjusts the sample stage to the best z-position to provide the most optimal image quality for segmentation and analysis. This is done by using a z-stack approach to capture images at each unit position and calculating Laplacian variance to measure the focus quality of the image. The resulting quality scores and corresponding z-positions are processed to locate the focal point of the microscope. From the best position height, stepper motors move the stage in the x- and y-direction via GRBL controller software on the Arduino. This allows for SIMPL to complete a full-sample image capture. The captured images are automatically stored into a folder. Next, the images from the image folder are put through a process involving object segmentation, and object data analysis. This is done using the Bioimage Analysis Statistic, and Comparison using Machine Learning (BASIN-ML) software written in R. BASIN-ML provides the user an easy-to-read analysis report on the data that was extracted for simple interpretation.

Results, Conclusions, and Discussions::
Results









So far, SIMPL can automatically find the best height to maximize image quality, move the stage, and capture images at various positions. However, there is still a gap in automation for transferring the folder images into BASIN-ML without manual user intervention. The current SIMPL workflow is shown on the process diagram. It will find the best z-position, scan the sample for full imaging, and process the sample images through BASIN-ML.









Conclusion









Here we have shown SIMPL’s ability to automatically focus and capture images. While the automatic analysis aspect leaves room for improvement, the current system still provides value. Having ran analyses on various cell types, it is shown that the software can give reproducible and automatic sample screening in research applications. The reproducibility of the process can be beneficial in research settings while allowing for a framework for easy automation.









Discussion









Moving forward, we aim to improve SIMPL’s processing power for better high throughput image analysis. It would also be beneficial for SIMPL to take steps towards even further automation. A current goal for SIMPL is to be able to perform real-time image capture and reporting for biofilms being grown in well plates. Since it is open-source software, the possibilities can stretch even further than currently planned.




Acknowledgements (Optional): :

References (Optional): :