A deterministic method for quantifying spindle-shaped cells in noisy bright-field microscopy.

Accurate quantification of spindle-shaped cells in bright-field microscopy remains challenging due to low contrast, noise, and highly variable cell morphology. Conventional approaches often rely on fluorescent staining or deep learning models, which may introduce phototoxic effects, require extensive training data, or offer limited interpretability. Here, we present a deterministic image analysis method for quantifying spindle-shaped cells directly from noisy bright-field microscopy images without the need for fluorescent labeling or supervised training. The proposed workflow combines contrast enhancement, adaptive thresholding, contour filtering, and shape-guided refinement to detect elongated cell bodies under challenging imaging conditions. The method is designed to be robust to irregular cell morphology and heterogeneous background commonly encountered in bright-field time-lapse microscopy. Quantitative evaluation demonstrates reliable cell counting and consistent performance across varying noise levels and imaging conditions. The approach achieves competitive accuracy while maintaining interpretability and low computational complexity, enabling straightforward integration into existing biomedical imaging workflows. By preserving cell morphology and avoiding fluorescent staining and complex model training, the method provides a practical solution for large-scale analysis of spindle-shaped cell populations in bright-field microscopy experiments.