Single-cell laser analysis of cancer model using machine learning

Biolaser is an emerging technology offering distinct advantages over fluorescence-based detection in terms of narrow linewidth, distinct signals, and strong intensity. Thanks to the strong optical feedback provided by optical cavity, a small change in the gain induced by the underlying biological processes is significantly amplified, leading to a drastic change in the laser output characteristics. As such, laser modes possess the huge potential to identify cellular changes in abnormal (cancer) cells by imaging single cellular lasing modes. However, the high chaotic and complex variation of laser modes makes it challenging for practical applications. By integrating with machine learning, laser modes could be trained and classified more precisely for cancer analysis. In this work, laser modes generated from Fabry sizes of dye-doped artificial nucleus and nucleus-cell ratio were fabricated to serve as the cancer cell model. By sandwiching the artificial cell modes in FP cavities with fixed cavity lengths, we found that the output laser mode pattern and orders change remarkably with the cell dimension. Furthermore, the experimental data were trained in a convolutional neural network (CNN)-based machine learning system, then the prediction of cell modes from the output laser modes were implemented. The results demonstrated a prediction with high accuracy, showing the potential of this novel method to be applied for biophysical studies and cellular mechanics.Bachelor of Engineering (Electrical and Electronic Engineering