Multiplexing steganography based on laser-induced breakdown spectroscopy coupled with machine learning

Optimization of machine learning classification models for tumor cells based on cell elements heterogeneity with laser-induced breakdown spectroscopy

The rapid and accurate diagnosis of cancer is an important topic in clinical medicine. In the present work, an innovative method based on laser-induced breakdown spectroscopy (LIBS) combined with machine learning was developed to distinguish and classify different tumor cell lines. The LIBS spectra of cells were first acquired. Then the spectral pre-processing was performed as well as detailed optimization to improve the classification accuracy. After that, the convolutional neural network (CNN), support vector machine (SVM), and K-nearest neighbors were further compared for the optimized classification ability of tumor cells. Both the CNN algorithm and SVM algorithm have achieved impressive discrimination performances for tumor cells distinguishing, with an accuracy of 97.72%. The results show that the heterogeneity of elements in tumor cells plays an important role in distinguishing the cells. It also means that the LIBS technique can be used as a fast classification method for classifying tumor cells.

Molecular Visual Sensing, Boolean Logic Computing, and Data Security Using a Droplet-Based Superwetting Paradigm

Inspired by information processing and logic operations of life, many artificial biochemical systems have been designed for applications in molecular information processing. However, encoding the binary synergism between matter, energy, and information in a superwetting system remains challenging. Herein, a superwetting paradigm was proposed for multifunctional applications including molecular visual sensing and data security on a superhydrophobic surface. A Triton X-100-encapsulated gelatin (TeG) hydrogel was prepared and selectively decomposed by trypsin, releasing the surfactant to decrease the surface tension of a droplet. Integrating the droplet with the superhydrophobic surface, the superwetting behavior was utilized for visual detection and information encoding. Interestingly, the proposed TeG hydrogel can function as an artificial gelneuron for molecular-level logic computing, where the combination of matters (superhydrophobic surface, trypsin, and leupeptin) acts as inputs to interact with energy (liquid surface tension and solid surface energy) and information (binary character), resulting in superwettability transitions (droplet surface tension, contact angle, rolling angle, and bounce) as outputs. Impressively, the TeG gelneuron can be further developed as molecular-level double cryptographic steganography to encode, encrypt, and hide specific information (including the maze escape route and content of the classical literature) due to its programmability, stimuli responsive ability, and droplet concealment. This study will encourage the development of advanced molecular paradigms and their applications, such as superwetting visual sensing, molecular computing, interaction, and data security.

High-Throughput Recognition of Tumor Cells Using Label-Free Elemental Characteristics Based on Interpretable Deep Learning

With cancer seriously hampering the increasing life expectancy of people, developing an instant diagnostic method has become an urgent objective. In this work, we developed a label-free laser-induced breakdown spectroscopy (LIBS) method for high-throughput recognition of tumor cells. LIBS spectra were straightly collected from cells dropped on a silicon substrate and built into a deep learning model for simultaneous classification of various cancers. To interpret the result of the deep learning algorithm, gradient-weighted class activation mapping was utilized to a one-dimensional convolution neural network (1D-CNN), and the saliency maps thus obtained amplified the differences between the spectra of cell lines. Overall results showed that the 1D-CNN algorithms achieved a mean sensitivity of 94.00%, a mean specificity of 98.47%, and a mean accuracy of 97.56%. Thus, the proposed method performed satisfactorily and is seen as an interpretable classification process for cancer cell lines.