Intelligent Software System for Low-Cost, Brightfield Segmentation: Algorithmic Implementation for Cytometric Auto-Analysis

TL;DR

The paper tackles the barrier of access to accurate, label-free cytometric analysis in low-resource settings by delivering a CPU-only, training-free framework built on the Homogeneous Image Plane (HIP) model. It combines semantic and instance segmentation, post-processing, morphometric analysis, and automated reporting within a user-friendly GUI, enabling end-to-end cytometric auto-analysis without GPUs or labeled data. Key contributions include the integration of HIP-based segmentation with novel instance analysis and automated reports, plus a modular, reproducible workflow with batch processing and profile management. The framework demonstrates competitive segmentation performance on public brightfield datasets, with strong interpretability and practical relevance for regenerative medicine, transplantation, and cellular morphology studies, while outlining clear avenues for future enhancements.

Abstract

Bright-field microscopy, a cost-effective solution for live-cell culture, is often the only resource available, along with standard CPUs, for many low-budget labs. The inherent challenges of bright-field images -- their noisiness, low contrast, and dynamic morphology -- coupled with a lack of GPU resources and complex software interfaces, hinder the desired research output. This article presents a novel microscopy image analysis framework designed for low-budget labs equipped with a standard CPU desktop. The Python-based program enables cytometric analysis of live, unstained cells in culture through an advanced computer vision and machine learning pipeline. Crucially, the framework operates on label-free data, requiring no manually annotated training data or training phase. It is accessible via a user-friendly, cross-platform GUI that requires no programming skills, while also providing a scripting interface for programmatic control and integration by developers. The end-to-end workflow performs semantic and instance segmentation, feature extraction, analysis, evaluation, and automated report generation. Its modular architecture supports easy maintenance and flexible integration while supporting both single-image and batch processing. Validated on several unstained cell types from the public dataset of livecells, the framework demonstrates superior accuracy and reproducibility compared to contemporary tools like Cellpose and StarDist. Its competitive segmentation speed on a CPU-based platform highlights its significant potential for basic research and clinical applications -- particularly in cell transplantation for personalised medicine and muscle regeneration therapies. The access to the application is available for reproducibility

Figures (8)

• Figure 1: (Left) Profile Manager for batch image segmentation, displaying stored parameter configurations for reproducible analysis across sessions or datasets. (Right) Schematic diagram showing the connected modules of the microscopy segmentation application
• Figure 2: Graphical interface for selecting background pixels within a configurable neighborhood (e.g., 5×5). The window allows users to sample pixel intensities interactively, enabling adaptive thresholding and homogeneity-based segmentation calibration $5 \times 5$
• Figure 3: (Left) Calibration module for adjusting brightness, masking, and randomization thresholds prior to segmentation. (Right) Denoising module applying morphological filters (erosion, dilation, median blur) to enhance contour definition and reduce noise artifacts in unstained cell images.
• Figure 4: (b), (c) and (d) are the results obtained from our segmentation and instance analysis modules applied to the LIVECell dataset (sample image: A172_Phase_A7_1_00d00h00m_4.PNG). (e), (f) are segmentation by cellpose SAM and Stardist.
• Figure 5: (b), (c) and (d) are the results obtained from our segmentation and instance analysis modules applied to the LIVECell dataset (sample image: A172_Phase_A7_1_00d00h00m_3.PNG). (e), (f) are segmentation by cellpose SAM and Stardist.