Hilai Karlin, Grubich Daniil, Akrawi Marcus, Zhu Hui, Zaghloul Razanne, Shi Chenjun, Do Man, Zhu Dongxiao, Zhang Jitao
Department of Biomedical Engineering, Wayne State University, Detroit, MI 48202, USA.
Department of Computer Science, Wayne State University, Detroit, MI, 48202, USA.
bioRxiv. 2024 Jul 2:2024.06.27.601015. doi: 10.1101/2024.06.27.601015.
Cellular biomechanics plays critical roles in cancer metastasis and tumor progression. Existing studies on cancer cell biomechanics are mostly conducted in flat 2D conditions, where cells' behavior can differ considerably from those in 3D physiological environments. Despite great advances in developing 3D models, probing cellular elasticity in 3D conditions remains a major challenge for existing technologies. In this work, we utilize optical Brillouin microscopy to longitudinally acquire mechanical images of growing cancerous spheroids over the period of eight days. The dense mechanical mapping from Brillouin microscopy enables us to extract spatially resolved and temporally evolving mechanical features that were previously inaccessible. Using an established machine learning algorithm, we demonstrate that incorporating these extracted mechanical features significantly improves the classification accuracy of cancer cells, from 74% to 95%. Building on this finding, we have developed a deep learning pipeline capable of accurately differentiating cancerous spheroids from normal ones solely using Brillouin images, suggesting the mechanical features of cancer cells could potentially serve as a new biomarker in cancer classification and detection.
细胞生物力学在癌症转移和肿瘤进展中起着关键作用。现有的关于癌细胞生物力学的研究大多是在平坦的二维条件下进行的，在这种条件下，细胞的行为可能与三维生理环境中的行为有很大差异。尽管在开发三维模型方面取得了巨大进展，但在三维条件下探测细胞弹性对现有技术来说仍然是一项重大挑战。在这项工作中，我们利用光学布里渊显微镜纵向获取了生长八天的癌性球体的力学图像。布里渊显微镜的密集力学映射使我们能够提取以前无法获得的空间分辨和随时间变化的力学特征。使用一种成熟的机器学习算法，我们证明，纳入这些提取的力学特征可将癌细胞的分类准确率从74%显著提高到95%。基于这一发现，我们开发了一种深度学习流程，仅使用布里渊图像就能准确区分癌性球体和正常球体，这表明癌细胞的力学特征有可能作为癌症分类和检测中的一种新生物标志物。
