This paper presents a novel microscopy-based framework for rapid pathogen identification using three-dimensional quantitative phase imaging (3D QPI) and an artificial neural network (ANN). The framework identifies bacterial species from a small number of cells, addressing the limitations of current methods with long turnaround times. Using 3D QPI to capture detailed morphological information, the ANN achieved an 82.5% accuracy in identifying 79 bacterial species causing bloodstream infections from a single cell or less. Accuracy increased to 99.9% with seven measurements, surpassing the performance of MALDI-TOF MS and providing a potential advisory tool for clinicians in early infection treatment.

Geon Kim, Daewoong Ahn, Minhee Kang, Jinho Park, DongHun Ryu, YoungJu Jo, Jinyeop Song, Jea Sung Ryu, Gunho Choi, Hyun Jung Chung, Kyuseok Kim, Doo Ryeon Chung, In Young Yoo, Hee Jae Huh, Hyun-seok Min, Nam Yong Lee, YongKeun Park