Computational Analysis and Evaluation of Fractal Geometric Descriptors and Spectral Robustness in Cellular Morphologies Under Imaging Perturbations

Abstract In order to improve our comprehension of biological structure-function interactions and disease phenotyping, quantitative measurement of cellular morphology is essential. A promising method for capturing the intricate, self-similar patterns present in cell forms is fractal geometry. Nevertheless, nothing is known about the durability and dependability of fractal-based descriptors in the presence of genuine imaging disruptions. In this work, we introduce a computational approach that uses both synthetic and experimental mask datasets to assess fractal dimensions and fractal spectral robustness across several cellular morphology classes. In order to simulate frequent imaging aberrations and noise, we systematically apply perturbations to evaluate fractal spectral similarity, compute box-counting fractal dimensions, and examine perimeter-area scaling connections.Our studies verify the potential value of fractal dimension measurements in morphological categorization by revealing statistically significant differences in these metrics across different morphological categories. Furthermore, we find that fractal spectral descriptor robustness varies with perturbation, with some perturbations leading to significant spectral similarity reduction. We highlight the need of taking into account disturbances and intensity levels that significantly affect fractal descriptor stability in morphological analysis by using thorough robustness fingerprinting. Overall, our results show that although fractal geometric descriptors offer valuable information about cellular morphology, their use necessitates a thorough evaluation of perturbation effects to guarantee accurate interpretation. This study provides useful directions for future research using fractal metrics in biological imaging contexts by establishing a rigorous computational approach for fractal-based morphological quantification.