AMAP-APP: Efficient Segmentation and Morphometry Quantification of Fluorescent Microscopy Images of Podocytes.

BACKGROUND: Automated quantification of podocyte foot process morphology is essential for unbiased, high-throughput analysis in kidney disease research. The previously established "Automatic Morphological Analysis of Podocytes" (AMAP) method allows for such analysis but is limited by high computational demands requiring high-performance computing systems, the lack of a user interface, and restriction to Linux operating systems. This study aimed to develop AMAP-APP, an optimized, cross-platform desktop application designed to overcome these hardware and usability barriers while maintaining analytical accuracy. METHODS: AMAP-APP replaces the computationally intensive instance segmentation method of the original AMAP with efficient classic image processing algorithms while retaining the original semantic segmentation model. Additionally, a new Region of Interest algorithm was introduced to improve precision. The tool was validated using 175 mouse and 189 human super-resolution stimulated emission depletion (STED) and confocal microscopy images. Performance was evaluated by benchmarking execution times on consumer hardware. Agreement with the original AMAP was assessed using Pearson correlation, Lin's concordance correlation coefficient (CCC), Bland-Altman plots, and paired Two One-Sided T-tests (TOST) for equivalence at both the image and animal level. RESULTS: AMAP-APP achieved an approximately 147-fold higher processing speed compared to the original GPU-accelerated AMAP. Morphometric quantifications (foot process area, perimeter, circularity, and slit diaphragm length density) demonstrated high agreement (Pearson r > 0.80 for all features and r > 0.90 for most, with comparably high Lin's CCC) and statistical equivalence to the original AMAP method across both mouse and human datasets (paired TOST P < 0.05 at both the image and animal level). Furthermore, the new ROI algorithm exhibited reduced deviation from manual delineations compared to the original AMAP ROI algorithm, enhancing the accuracy of slit diaphragm length density measurements. CONCLUSIONS: AMAP-APP provides substantially higher efficiency and accessibility for deep learning-based podocyte morphometry. By eliminating the need for high-performance computing clusters and providing a user-friendly interface for Windows, macOS, and Linux, it makes this analytical method broadly available for nephrology research.