Deep learning-assisted Fe3O4 NPs for in-situ visual recognition of multi-dimensional information on latent fingerprints

Latent fingerprints (LFPs) are important trace evidences found at crime scenes that assist forensic experts in human identification, and a growing number of studies show that LFPs have some medical value. Herein, as one of the common nanoenzymes, Fe3O4 nanoparticles (Fe3O4 NPs) was used for LFPs imaging due to its excellent superparamagnetic properties and for recognition of glucose/H2O2 on LFPs because of its horseradish peroxidase-like activity, which makes it possible to diagnose diabetes non-invasively through LFPs. Furthermore, the intelligent qualitative and semi-quantitative analysis of glucose/H2O2 on LFPs was achieved by designing the classifier based on the multi-channel convolutional neural network (MC-CNN) technology. The qualitative accuracy rate is over 97% and the semi-quantitative accuracy rate reaches 90%. This work offers the promise to facilitate cross-disciplinary studies between artificial intelligence methods and fingerprint development technique.