OR WAIT null SECS
© 2023 MJH Life Sciences™ and Spectroscopy Online. All rights reserved.
The DeepRaman method proposed in the study is an accurate, universal, and ready-to-use method for component identification in various application scenarios.
A group of scientists from Central South University in Changsha, China, has developed a new method for component identification in Raman spectroscopy called DeepRaman (1). The DeepRaman method utilizes deep learning, combining the comparison ability of a pseudo-Siamese neural network (pSNN) with the input-shape flexibility of spatial pyramid pooling (SPP) to accurately identify components in Raman spectra, even for mixtures.
A pSNN is a type of neural network that can learn to compare similarity between two input samples. Unlike the traditional Siamese network, which has shared weights for both input branches, pSNN only shares some of the weights, allowing for more flexible input shapes. This is achieved by using spatial pyramid pooling (SPP), which divides the input into multiple sub-regions and pools features from each region. This enables pSNN to handle inputs of varying sizes, making it more suitable for real-world applications. The pSNN can be trained using various loss functions, such as contrastive loss or triplet loss, to optimize the similarity metric between input samples. By combining the comparison ability of pSNN and the input-shape flexibility of SPP, it is possible to develop an accurate and efficient method for component identification using Raman spectra.
Raman spectroscopy is commonly used to provide the structural fingerprint for molecular identification. However, the interference from coexisting components, noise, baseline, along with the systematic differences between spectrometers, make component identification challenging. DeepRaman can achieve an impressive 96.29% accuracy, 98.40% true positive rate (TPR), and 94.36% true negative rate (TNR) on the test set, making it an accurate and universal method for component identification in various application scenarios.
The research team used 41,564 augmented Raman spectra from two databases (pharmaceutical material and S.T. Japan) to train, validate, and test DeepRaman. They also evaluated the performance of the method using six additional data sets measured on different instruments. The results showed that DeepRaman significantly outperformed the hit quality index (HQI) method and other deep learning models.
Furthermore, DeepRaman performs well in cases of different spectral complexity and low-content components. Once the model is established, it can be used directly on different data sets without retraining or transfer learning. The method also obtains promising results for the analysis of surface-enhanced Raman spectroscopy (SERS) data sets and Raman imaging data sets.
The DeepRaman method proposed in the study is an accurate, universal, and ready-to-use method for component identification in various application scenarios. It provides an innovative solution to the problem of component identification in Raman spectroscopy and could have a significant impact on future research in this field.
(1) Fan, X.; Wang, Y.; Yu, C.; Lv, Y.; Zhang, H.; Yang, Q.; Wen, M.; Lu, H.; Zhang, Z.A Universal and Accurate Method for Easily Identifying Components in Raman Spectroscopy Based on Deep Learning. Anal Chem. 2023, 95 (11), 4863–4870. DOI: 10.1021/acs.analchem.2c03853