semi-automated object-segmentation matlab script  (MathWorks Inc)

Journal: Frontiers in Cellular Neuroscience

Article Title: A novel automated morphological analysis of Iba1+ microglia using a deep learning assisted model

doi: 10.3389/fncel.2022.944875

Figure Lengend Snippet: Comparison of MATLAB and Aiforia ® in the quantification of microglia morphology. (A) Brain region specific comparison of area/perimeter ratio values between methods, suggesting similar levels of accuracy. N’s represent the number of cells quantified in each method. Legend includes the duration of researcher time needed to acquire, process and analyze the complete dataset. n = 5 mice, using 3 serial tissue sections per brain region. With AIforia an entire brain region per tissue section was quantified, compared with 5–7 user manually identified cells from 3 representative 60x bright field images per tissue section in MATLAB. (B) A cell specific comparison of area/perimeter values between methods, suggesting that the significant ( p < 0.0001) difference in output values is not a product of sampling biases. Connecting lines indicate differences in values for a single cell between methods. Example images showing how each analysis method segments individual cells to determine the area/perimeter values, and the duration of researcher time need to acquire, process and analyze the complete dataset. (C) Comparison of the AIforia microglia model performance against five researchers experienced in microglia histopathology (80 validations regions; 14 images) with no significant differences, suggesting that the AI is performing to the same standard as human researchers. F-measure group labels apply to all histograms by color code.

Article Snippet: To evaluate how well our new microglia detection model quantifies change in microglia reactivity we compared area/perimeter ratios between our model and a custom semi-automated object-segmentation MATLAB script that had been previously published ( , ).

Techniques: Comparison, Sampling, Histopathology

Journal: Frontiers in Cellular Neuroscience

Article Title: A novel automated morphological analysis of Iba1+ microglia using a deep learning assisted model

doi: 10.3389/fncel.2022.944875

Figure Lengend Snippet: Schematic of a side-by-side comparison between the workflows for the quantification of microglial morphology using MATLAB or the Aiforia ® platform. While the MATLAB methodology requires investigators to spend relatively long hours using a common brightfield microscope, the use of a slide scanner in the Aiforia ® methodology requires only the amount of time that it takes to load the slides and set up the scanning parameters. During the sampling phase the MATLAB methodology can be prone to some sample bias, using 3 high magnification images (60×) per section and 3 sections per brain region per animal. During the data acquisition phase, the MATLAB script requires some dedicated hardware and the software license, while the Aiforia ® platform requires a subscription and Internet connection. Finally, data analysis in Aiforia ® is a faster method particularly when using large datasets and quantifies more morphology parameters (created with BioRender.com ).

Article Snippet: To evaluate how well our new microglia detection model quantifies change in microglia reactivity we compared area/perimeter ratios between our model and a custom semi-automated object-segmentation MATLAB script that had been previously published ( , ).

Techniques: Comparison, Microscopy, Sampling, Software