Analysis of Raman Spectra by Using Deep Learning Methods in the Identification of Marine Pathogens

The need for efficient and accurate identification of pathogens in seafood the environment has become increasingly urgent, given current global pandemic. Traditional methods are not only time consuming but also lead to sample wastage. Here, we have proposed two new that involve Raman spectroscopy combined with a long short-term memory (LSTM) neural network compared them method using normal convolutional (CNN). We used eight strains isolated from marine organism Urechis unicinctus, including four kinds pathogens. After models were configured trained, LSTM achieved average isolation-level accuracies exceeding 94%, meeting requirement indicating faster more than CNN models. Finally, through computational approach, designed loss function explore mechanism reflected by data, finding segments most likely exhibited characteristics nucleic acids. These novel experimental results provide insights developing additional deep learning accurately analyze complex data.