Deep Learning Aided Small-Sized Portable Fluorescence Biochip Reader

Abstract

We present here the development of a novel fluorescence-based portable biochip reader system, having a novel machine learning method for fluorescence measurement of multianalytes. It is suitable for reading DNA/protein biochips in basic research and clinical diagnosis. We aim to develop a point-of-care diagnostic device in the healthcare sector to accelerate biomedical research at an affordable cost. Innovation is involved in the design of each module of the system that is compact, detachable, and reconfigurable. The system can be used with multiple fluorescent dyes depending on specific requirements. It provides high-sensitivity and high-resolution fluorescent images with lesser scan time compared to currently available state-of-the-art systems. The system is autoaligned/autofocus with a standard microscopic glass slide and custom-developed membrane-based biochip. A new deep learning network Bio-U-Net has been developed for real-time detection and quantification of microarrays. On the real test images, it achieved the precision of 0.9928 and the sensitivity of 0.9698 along with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula> -score of 0.9812. The system and deep learning network have been extensively tested and these together provide a low-cost option while maintaining the required accuracy.