Automation of morphometric analyses in small intestine images

This study presents an automated method to accelerate morphometric analyses of biomedical images, specifically targeting the segmentation of regions in the myenteric plexus of rat small intestine samples. Employing the deep learning models U-Net, Pre-trained U-Net and Mask R-CNN, the research aimed to streamline the segmentation process, facilitating precise extraction of morphometric measurements. Comparative analyses revealed that both models performed similarly and effectively segmented regions of interest. Results demonstrated statistically significant differences between morphometric analyses conducted on segmented regions compared to analyses applied to entire images, highlighting the importance of accurate segmentation. Additionally, statistical validation using the Wilcoxon Signed-Rank test confirmed the equivalence of automated segmentation results to manually annotated ground truth data. This confirms that deep learning-based segmentation models are reliable tools for biomedical research, significantly reducing manual tasks and enhancing overall efficiency.