Ultrasonic phenotyping of breast cancer cells for molecular subtyping

Abstract

Breast cancer can be divided into molecular subtypes which are defined by their genetic and protein expression profiles. Current methods aimed at testing these biochemical signatures are effective classifiers but are not easily transferable to real-time clinical applications. The rapid, cost-effective determination of molecular subtype by other means would be a significant advancement in cancer detection and treatment. Our studies suggest that high-frequency ultrasound (10–100 MHz) may be sensitive to variations among breast cancer subtypes through their cytoskeletal structures, which have distinct biomechanical signatures. To further test this hypothesis, four breast cancer cell lines of different subtypes were cultured and ultrasonically tested. Direct pulse-echo measurements were collected from the cell layers using a 50-MHz transducer immersed in the growth media of the culture plates. Cell reflections in the waveforms were isolated and spectrally analyzed using computationally modeled spectra and principal component analysis (PCA). Cell phenotypes were profiled by using heat maps to display the relative distances between the PCA scores of the experimental and model spectra. The results indicate the phenotype and thus molecular subtype of cancer cells could potentially be determined by comparing their measured spectra to model spectra using a feature classification program such as PCA.