B110-06 OpenMLI: An Open-Source Platform for Automated Mean Linear Intercept Analysis in Lung Histomorphometry

Rationale Mean Linear Intercept (MLI) quantification remains the gold standard for assessing alveolar dimensions and emphysema severity in lung histology. However, traditional manual MLI measurement is labor-intensive, time-consuming, and prone to inter-observer variability, limiting reproducibility across studies and institutions. Despite decades of use, no standardized open-source software exists for automated MLI analysis with community validation. We developed OpenMLI, a freely available Python-based platform combining automated morphometry with interactive manual correction capabilities and validated its performance against manual measurements. Methods OpenMLI implements traditional MLI methodology through a modern computational framework. The core algorithm performs automated airspace segmentation using deep learning-based on Segment-Anything-Model(SAM). A synthetic grid system with customizable spacing and orientation (horizontal, vertical, or both) is overlaid on segmented images. The software employs two intercept counting modes: “wall” mode counting contiguous tissue segments (recommended), and “boundary” mode detecting binary transitions. The PyQt5-based graphical interface enables real-time visualization of detected intercepts, interactive point-and-click editing for manual correction, mouse-wheel zoom and pan navigation, batch processing with automated result export, and generation of annotated overlay images with per-line intercept counts. We validated OpenMLI by analyzing fifteen lung histology images from a PPE-hypoxia mouse model, comparing automated measurements against expert manual human measurements performed using established protocols. Results OpenMLI successfully processed all images with consistent performance across the validation dataset. OpenMLI measurements ranged from 49.92 to 70.04 μm with a mean of 57.61 ± 6.48 μm, while manual human measurements ranged from 46.11 to 61.41 μm with a mean of 52.47 ± 5.12 μm. The automated method consistently detected less tissue-airspace intercepts, yielding MLI values approximately 9.9% higher than manual measurements (mean difference: 5.19 μm, SD: 4.01 μm). This systematic difference suggests the automated method is less prone to over-counting ambiguous intercepts, applying a uniform set of rules that results in a more reproducible measurement compared to the inherent subjectivity of manual assessment. Strong inter-method correlation (Pearson r = 0.84, p < 0.001) validates the automated approach. Individual image analysis completion time averaged under a second-per-image compared to 15-20 minutes for traditional manual counting, representing a greater than 1000-fold improvement in efficiency. OpenMLI is accessible at https://gitlab.com/dig5485742/OpenMLI.git Conclusions OpenMLI provides the research community with a validated, freely available tool for standardized lung morphometry that significantly reduces analysis time while maintaining measurement accuracy. The platform addresses critical needs for reproducibility and throughput, released under MIT license enables community-driven refinement and adaptation, promoting standardization of MLI methodology. This abstract is funded by: NIH/NHLBI K01HL163249