Get started
Work

Investigation of the physical properties of breast cancer cells via statistical thermodynamics, quantitative phase microscopy, and machine learning algorithms

Year: 2026
Type: article
Abstract: The study of cellular physical properties is crucial for understanding the mecha-nisms of carcinogenesis and metastasis. This paper presents a comprehensive ap-proach for assessing the oncological status and metastatic potential of human breast cells by integrating quantitative phase microscopy, statistical thermodynamics, and ma-chine learning. The study used non-cancerous MCF-10A, low-metastatic MCF-7, and highly metastatic MDA-MB-231 cell lines as models. Phase images and track diagrams were obtained for each phenotype, followed by morphometric, multifractal, and mechani-cal analysis. The results revealed statistically significant differences between cell groups across a range of parameters. Dimensionless characteristics, such as the contour fractal dimension and the skewness of the phase image, were identified as effective diagnostic markers. Multifractal analysis of dynamic signals allowed for reliable identification of non-aggressive MCF-7 cancer cells. A statistical-thermodynamic model was applied to analyze the mechanical response, identifying parameters that characterize cell defor-mation under shear stress. A “stiffness paradox” was discovered, showing a non-monotonic dependence of the shear modulus on metastatic potential. To resolve this paradox, an integral Aggressiveness Index was proposed, which demonstrated a mono-tonic change from normal to highly aggressive cells. Furthermore, the high efficiency of a MobileNetV2 convolutional neural network in classifying phase images was demon-strated, with an accuracy exceeding 90 %. These findings confirm the potential of the proposed approach for label-free cancer diagnostics.
DOI: 10.15593/rjbiomech/2026.2.06
Language: English
Open Access status: diamond
OpenAlex ID: W7167831030
FWCI:
Cites:
Cited by:
Related to: 10