Visible Particle Identification Using Raman Spectroscopy and Machine Learning

Fast Detection and Classification of Microplastics below 10 μm Using CNN with Raman Spectroscopy

In light of the growing awareness regarding the ubiquitous presence of microplastics (MPs) in our environment, recent efforts have been made to integrate Artificial Intelligence (AI) technology into MP detection. Among spectroscopic techniques, Raman spectroscopy is preferred for the detection of MP particles measuring less than 10 μm, as it overcomes the diffraction limitations encountered in Fourier transform infrared (FTIR). However, Raman spectroscopy's inherent limitation is its low scattering cross section, which often results in prolonged data collection times during practical sample measurements. In this study, we implemented a convolutional neural network (CNN) model alongside a tailored data interpolation strategy to expedite data collection for MP particles within the 1-10 μm range. Remarkably, we achieved the classification of plastic types for individual particles with a mere 0.4 s of exposure time, reaching an approximate confidence level of 85.47(±5.00)%. We postulate that the result significantly accelerates the aggregation of microplastic distribution data in diverse scenarios, contributing to the development of a comprehensive global microplastic map.

Closed, one-stop intelligent and accurate particle characterization based on micro-Raman spectroscopy and digital microfluidics

Monoclonal antibodies are prone to form protein particles through aggregation, fragmentation, and oxidation under varying stress conditions during the manufacturing, shipping, and storage of parenteral drug products. According to pharmacopeia requirements, sub-visible particle levels need to be controlled throughout the shelf life of the product. Therefore, in addition to determining particle counts, it is crucial to accurately characterize particles in drug product to understand the stress condition of exposure and to implement appropriate mitigation actions for a specific formulation. In this study, we developed a new method for intelligent characterization of protein particles using micro-Raman spectroscopy on a digital microfluidic chip (DMF). Several microliters of protein particle solutions induced by stress degradation were loaded onto a DMF chip to generate multiple droplets for Raman spectroscopy testing. By training multiple machine learning classification models on the obtained Raman spectra of protein particles, eight types of protein particles were successfully characterized and predicted with high classification accuracy (93%-100%). The advantages of the novel particle characterization method proposed in this study include a closed system to prevent particle contamination, one-stop testing of morphological and chemical structure information, low sample volume consumption, reusable particle droplets, and simplified data analysis with high classification accuracy. It provides great potential to determine the probable root cause of the particle source or stress conditions by a single testing, so that an accurate particle control strategy can be developed and ultimately extend the product shelf-life.

RamanCMP: A Raman spectral classification acceleration method based on lightweight model and model compression techniques

In recent years, Raman spectroscopy combined with deep learning techniques has been widely used in various fields such as medical, chemical, and geological. However, there is still room for optimization of deep learning techniques and model compression algorithms for processing Raman spectral data. To further optimize deep learning models applied to Raman spectroscopy, in this study time, accuracy, sensitivity, specificity and floating point operations numbers(FLOPs) are used as evaluation metrics to optimize the model, which is named RamanCompact(RamanCMP). The experimental data used in this research are selected from the RRUFF public dataset, which consists of 723 Raman spectroscopy data samples from 10 different mineral categories. In this paper, 1D-EfficientNet adapted to the spectral data as well as 1D-DRSN are proposed to improve the model classification accuracy. To achieve better classification accuracy while optimizing the time parameters, three model compression methods are designed: knowledge distillation using 1D-EfficientNet model as a teacher model to train convolutional neural networks(CNN), proposing a channel conversion method to optimize 1D-DRSN model, and using 1D-DRSN model as a feature extractor in combination with linear discriminant analysis(LDA) model for classification. Compared with the traditional LDA and CNN models, the accuracy of 1D-EfficientNet and 1D-DRSN is improved by more than 20%. The time of the distilled model is reduced by 9680.9s compared with the teacher model 1D-EfficientNet under the condition of losing 2.07% accuracy. The accuracy of the distilled model is improved by 20% compared to the CNN student model while keeping inference efficiency constant. The 1D-DRSN optimized with channel conversion method saves 60% inference time of the original 1D-DRSN model. Feature extraction reduces the inference time of 1D-DRSN model by 93% with 94.48% accuracy. This study innovatively combines lightweight models and model compression algorithms to improve the classification speed of deep learning models in the field of Raman spectroscopy, forming a complete set of analysis methods and laying the foundation for future research.

A top-down spectroscopic approach for correlating coating thickness distributions with the dissolution profiles of enterically coated pellets

Pharmaceutical dosage forms such as tablets and capsules are often coated with a functional polymer to modify the drug release. To obtain the drug release profiles, ensure quality control and predict in-vivo performance, dissolution studies are performed. However, dissolution tests are time-consuming, sample destructive and do not readily allow for at-line or in-line characterization. Rapid assessment of functional coatings is essential for products where a single capsule is comprised of hundreds of functionally-coated pellets and the collective drug release kinetics of the entire capsule depends on contributions from each pellet. Here, single Raman measurements were used to evaluate the coating thickness distributions of a dosage form comprised of small, functionally-coated pellets in capsules. First, the composition and physicochemical properties of pellets were characterized by multivariate analysis assisted Raman mapping of pellet cross-sections. Second, a method of collecting single Raman spectrum with spectral contributions from the coating and API layers was developed and optimized to estimate the thickness of coatings. The coating thicknesses obtained from single Raman measurements of pellets in each capsule revealed thickness distributions that correlated with the dissolution profiles (capsules with one distribution had single stage release and capsules with two distributions had a two-stage release). Finally, an unsupervised multivariate analysis method was demonstrated as a rapid and efficient way to correlate dissolution profiles of enterically coated pellets. In summary, this study presents a non-destructive and rapid characterization method for assessing coating thickness and has the potential to be applied in process analytical technologies to ensure coating uniformity and predict product dissolution rate performance.