Advanced Nosema bombycis Spore Identification: Single-Cell\nRaman Spectroscopy Combined with\nSelf-Attention Mechanism-Guided Deep Learning

Nosema bombycis (Nb) has\nbeen considered\na dangerous pathogen, which can spread rapidly through free spores.\nNowadays, pebrine disease caused by Nb spores is a serious threat\nto silkworms, causing huge economic losses in both the silk industry\nand agriculture every year. Thus, how to accurately identify living\nNb spores at a single-cell level is greatly demanded. In this work,\nwe proposed a novel approach to accurately and conveniently identify\nNb spores using single-cell Raman spectroscopy and a self-attention\nmechanism (SAM)-guided convolutional neural network (CNN) framework.\nWith the assistance of SAM and data augmentation methods, an optimal\nCNN model can not only efficiently extract spectral feature information\nbut also construct potential relationships of global spectral features.\nCompared with the case without both SAM and data augmentation, the\naverage prediction accuracy of Nb spores from nine different Bombyx mori larvae can be significantly developed\nby almost 18%, from original 83.93 ± 4.88% to 99.27 ± 0.25%.\nTo visualize the individual classification weight, a local feature\nextraction strategy named blocking individual Raman bands was proposed.\nAccording to the relative weight, these four Raman bands located at\n1658, 1458, 1127, and 849 cm<sup>–1</sup>, mainly contribute\nto the high prediction accuracy of 99.27 ± 0.25%. It is worth\nnoting that these Raman bands were also highlighted by the weight\ncurve of SAM, indicating that the four Raman bands proposed by our\noptimal CNN model are reliable. Our findings clearly show that single-cell\nRaman spectroscopy combined with SAM-mediated CNN configuration has\ngreat potential in performing early diagnosis of Nb spores and monitoring\npebrine disease.