Effect of non-resonant background on the extraction of Raman signals from CARS spectra using deep neural networks

Abstract

Typical schematic of the CNN model architecture trained with CARS data generated from different non resonant backgrounds. The input is a CARS spectrum and the output is a Raman spectrum.

Evaluating different deep learning models for efficient extraction of Raman signals from CARS spectra

“…The CARS spectrum simulation procedure can be found in our previous work. 35 The simulation parameter details are presented in Table 1.…”

“…42 So, we have directly retrained it by utilizing their code without modifying any of its model parameters. In the case of the CNN, the trained model weights are directly taken from our previous work 35 and can be accessed from here. 43 Houhou et al have reported the LSTM model, but the trained weights and the respective codes are not openly available.…”

“…The CNN architecture used here is SpecNet. 35 The typical schematic of the CNN model's architecture is presented in Fig. 1(a).…”

“…These three works have shown superior performance compared to the SpecNet, where spectral lines with minimal intensities are also predicted. 33–35 Even though the CNN model trained with polynomial NRB has predicted all the spectral lines of the experimental data, the intensity of a few lines deviated from the true one. Also, similar results were obtained with the LSTM 30 and VECTOR models, 32 where the performance was found to be sensitive when evaluating the experimental CARS data.…”

“…Also, the NRB is assumed to be a fourth-order polynomial function while producing the CARS training data, which has already shown optimum results. 35 Furthermore, a comprehensive study is performed by comparing the performance of four DL models, namely, (1) VECTOR, (2) CNN, (3) LSTM, and (4) Bi-LSTM. This comparative study has been done for the first time to the best of our knowledge, and critically evaluates the trained model's efficiency in retrieving the Raman signal from the CARS data.…”