Vibrational Spectroscopy最新文献_第3页

Lightweight double attention neural network based on Raman spectroscopy for diagnosis of osteoporosis 基于拉曼光谱的轻量双注意神经网络诊断骨质疏松症

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-10-18 DOI: 10.1016/j.vibspec.2025.103861

Haoshaqiang Zhang , Xuguang Zhou , Cheng Chen

Osteoporosis is a common chronic bone metabolic disease, and its early diagnosis is important for preventing fractures and delaying the disease process. Raman spectroscopy, as a non-invasive and high-throughput molecular detection method, has shown unique advantages in bone tissue composition detection. However, limited by the high dimensionality, peak redundancy and biological variability of spectral data, traditional machine learning methods have bottlenecks in feature extraction and classification accuracy. To address this problem, this paper proposes a lightweight one-dimensional Double Attention Neural Network (DAN) based on Raman spectra, combining an encoder-decoder structure with a spatial-channel double attention mechanism for efficient intelligent diagnosis of osteoporosis. The proposed double-attention module effectively enhances the model's ability to perceive spectral structures and pathological patterns by modeling the position dependence between bands and feature focusing between channels in parallel via two independent paths. In this paper, the system is validated on a real clinical Raman dataset, and the DAN achieves optimal performance in all kinds of indexes, with an accuracy of 97.50 %, which is better than the traditional machine learning model and deep learning model. At the same time, this paper explores the contribution of the attention mechanism in depth by designing ablation experiments, and the results show that the double attention mechanism is significantly better than the model that only adopts a single spatial or channel attention in terms of both accuracy and robustness. With a parameter count of only 0.11 M and an inference overhead as low as 0.01 GFlops, the model has the advantage of lightweight deployment, as well as good interpretability and medical adaptability, which provides a new deep learning path for future spectral-based assisted diagnosis of osteoporosis.

骨质疏松症是一种常见的慢性骨代谢性疾病，其早期诊断对于预防骨折和延缓疾病进程具有重要意义。拉曼光谱作为一种无创、高通量的分子检测方法，在骨组织成分检测中显示出独特的优势。然而，受光谱数据的高维性、峰冗余性和生物可变性等因素的限制，传统的机器学习方法在特征提取和分类精度方面存在瓶颈。针对这一问题，本文提出了一种基于拉曼光谱的轻量级一维双注意神经网络（DAN），将编码器-解码器结构与空间通道双注意机制相结合，实现了骨质疏松症的高效智能诊断。本文提出的双注意模块通过模拟波段之间的位置依赖和通道之间的特征聚焦，通过两条独立的路径并行，有效地增强了模型对光谱结构和病理模式的感知能力。本文在真实的临床Raman数据集上对系统进行了验证，DAN在各项指标上都达到了最优的性能，准确率达到97.50 %，优于传统的机器学习模型和深度学习模型。同时，本文通过设计烧蚀实验，深入探讨了注意机制的贡献，结果表明，双注意机制在准确性和鲁棒性方面都明显优于仅采用单一空间或通道注意的模型。该模型参数数仅为0.11 M，推理开销低至0.01 GFlops，具有轻量级部署、良好的可解释性和医学适应性等优点，为未来基于光谱的骨质疏松辅助诊断提供了新的深度学习路径。

{"title":"Lightweight double attention neural network based on Raman spectroscopy for diagnosis of osteoporosis","authors":"Haoshaqiang Zhang , Xuguang Zhou , Cheng Chen","doi":"10.1016/j.vibspec.2025.103861","DOIUrl":"10.1016/j.vibspec.2025.103861","url":null,"abstract":"<div><div>Osteoporosis is a common chronic bone metabolic disease, and its early diagnosis is important for preventing fractures and delaying the disease process. Raman spectroscopy, as a non-invasive and high-throughput molecular detection method, has shown unique advantages in bone tissue composition detection. However, limited by the high dimensionality, peak redundancy and biological variability of spectral data, traditional machine learning methods have bottlenecks in feature extraction and classification accuracy. To address this problem, this paper proposes a lightweight one-dimensional Double Attention Neural Network (DAN) based on Raman spectra, combining an encoder-decoder structure with a spatial-channel double attention mechanism for efficient intelligent diagnosis of osteoporosis. The proposed double-attention module effectively enhances the model's ability to perceive spectral structures and pathological patterns by modeling the position dependence between bands and feature focusing between channels in parallel via two independent paths. In this paper, the system is validated on a real clinical Raman dataset, and the DAN achieves optimal performance in all kinds of indexes, with an accuracy of 97.50 %, which is better than the traditional machine learning model and deep learning model. At the same time, this paper explores the contribution of the attention mechanism in depth by designing ablation experiments, and the results show that the double attention mechanism is significantly better than the model that only adopts a single spatial or channel attention in terms of both accuracy and robustness. With a parameter count of only 0.11 M and an inference overhead as low as 0.01 GFlops, the model has the advantage of lightweight deployment, as well as good interpretability and medical adaptability, which provides a new deep learning path for future spectral-based assisted diagnosis of osteoporosis.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103861"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362734","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Raman spectroscopic evaluation of silver diamine fluoride in preventing acid-induced caries progression 二胺氟化银预防酸致龋进展的拉曼光谱评价

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-09-05 DOI: 10.1016/j.vibspec.2025.103856

Agata Daktera-Micker , Maciej Rzeczkowski , Zuzanna Buchwald , Tomasz Buchwald

The antibacterial mechanisms of silver diamine fluoride, as well as its ability to prevent tooth caries, are well-established. The aim of this study was to provide a clear evaluation of whether silver diamine fluoride, in addition to its antibacterial properties, is capable of inhibiting the progression of dental caries in an acidic environment to the same extent as silver nitrate. The results will help determine which of these two compounds has greater potential for application in the prevention and treatment of early carious lesions. To achieve this goal, studies were conducted on teeth under four conditions: healthy, after first demineralization, following the application of the tested compounds, and after second demineralization. At each stage of the research, Raman spectroscopy was used to assess changes in the enamel structure. These findings showed that, although neither compound completely inhibits the progression of demineralization, both can slow it down. Furthermore, no significant differences were observed between the two compounds in their effectiveness in reducing lesion progression. A notable observation was that the silver diamine fluoride was washed away by the demineralizing solution, causing the protective barrier it provided to disappear. This finding may be attributed to the limited penetration depth of the compounds into the enamel. The compounds primarily penetrated the demineralized areas, but the depth of enamel demineralization in our study was approximately 200 µm. This limited penetration could explain the washing out of the compound and its reduced effectiveness in inhibiting demineralization.

氟化二胺银的抗菌机制，以及它预防龋齿的能力，是公认的。本研究的目的是明确评估氟化二胺银除了抗菌性能外，是否能够在酸性环境中抑制龋齿的发展，其程度与硝酸银相同。该结果将有助于确定这两种化合物中哪一种在预防和治疗早期龋齿病变方面具有更大的应用潜力。为了实现这一目标，研究人员在四种情况下对牙齿进行了研究：健康、第一次脱矿后、使用测试化合物后和第二次脱矿后。在研究的每个阶段，使用拉曼光谱来评估牙釉质结构的变化。这些发现表明，虽然这两种化合物都不能完全抑制脱矿的进程，但它们都能减缓脱矿的进程。此外，两种化合物在减少病变进展的有效性方面没有显著差异。一个值得注意的观察结果是，氟化银二胺被脱矿溶液冲走，导致它提供的保护屏障消失。这一发现可能是由于化合物对牙釉质的渗透深度有限。化合物主要渗透到脱矿区，但我们研究的牙釉质脱矿深度约为200 µm。这种有限的渗透可以解释化合物被洗出及其抑制脱矿的效果降低的原因。

{"title":"Raman spectroscopic evaluation of silver diamine fluoride in preventing acid-induced caries progression","authors":"Agata Daktera-Micker , Maciej Rzeczkowski , Zuzanna Buchwald , Tomasz Buchwald","doi":"10.1016/j.vibspec.2025.103856","DOIUrl":"10.1016/j.vibspec.2025.103856","url":null,"abstract":"<div><div>The antibacterial mechanisms of silver diamine fluoride, as well as its ability to prevent tooth caries, are well-established. The aim of this study was to provide a clear evaluation of whether silver diamine fluoride, in addition to its antibacterial properties, is capable of inhibiting the progression of dental caries in an acidic environment to the same extent as silver nitrate. The results will help determine which of these two compounds has greater potential for application in the prevention and treatment of early carious lesions. To achieve this goal, studies were conducted on teeth under four conditions: healthy, after first demineralization, following the application of the tested compounds, and after second demineralization. At each stage of the research, Raman spectroscopy was used to assess changes in the enamel structure. These findings showed that, although neither compound completely inhibits the progression of demineralization, both can slow it down. Furthermore, no significant differences were observed between the two compounds in their effectiveness in reducing lesion progression. A notable observation was that the silver diamine fluoride was washed away by the demineralizing solution, causing the protective barrier it provided to disappear. This finding may be attributed to the limited penetration depth of the compounds into the enamel. The compounds primarily penetrated the demineralized areas, but the depth of enamel demineralization in our study was approximately 200 µm. This limited penetration could explain the washing out of the compound and its reduced effectiveness in inhibiting demineralization.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103856"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145005238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Research on quantitative analysis of single-component hydrocarbon gases based on optimized SVR with double-local strategy 基于双局部优化SVR策略的单组分烃类气体定量分析研究

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-11-10 DOI: 10.1016/j.vibspec.2025.103865

Gang Liu , Peng Han , Hai Yang , Yilin Yao , Haibo Liang

Accurate measurement of gas concentrations in drilling fluids is essential for effective gas logging in drilling engineering, as both accuracy and speed significantly influence real-time monitoring and decision-making. To address the limitations of on-site infrared spectroscopy for single-component gas measurements—such as model mismatches, redundant wavenumbers across broad spectral ranges, and substantial concentration fluctuations that lead to slow processing and large errors, which fail to meet gas logging needs—this study introduces an optimized solution. The approach involves modifying the measurement model, utilizing local characteristic wavenumbers to eliminate redundant spectral information, and dividing local calibration sets to reduce the effects of concentration fluctuations. A novel modeling technique is developed by enhancing Competitive Adaptive Reweighted Sampling (CARS) with Interval Random Frog (IRF) and combining it with the Beetle Antennae Search (BAS) algorithm to optimize Support Vector Regression (SVR). Experimental results demonstrate that the proposed IRF-CARS-BAS-SVR method significantly outperforms traditional techniques in the quantitative analysis of single-component hydrocarbons, achieving an average prediction accuracy exceeding 98 %. This method improves both the speed and precision of infrared-based quantification of single-component gases and has been successfully deployed at various drilling sites, meeting operational requirements for gas logging.

在钻井工程中，准确测量钻井液中的气体浓度对有效的气体测井至关重要，因为准确性和速度对实时监测和决策具有重要影响。为了解决现场红外光谱用于单组分气体测量的局限性，例如模型不匹配，宽光谱范围内的冗余波数，以及导致处理缓慢和误差大的浓度波动，这些都不能满足气体测井的需求，本研究引入了一种优化的解决方案。该方法包括修改测量模型，利用局部特征波数消除冗余的光谱信息，划分局部校准集以减少浓度波动的影响。提出了一种基于区间随机蛙（IRF）的竞争自适应重加权抽样（CARS）算法，并结合甲虫天线搜索（BAS）算法优化支持向量回归（SVR）的建模方法。实验结果表明，所提出的IRF-CARS-BAS-SVR方法在单组分烃定量分析方面明显优于传统技术，平均预测精度超过98 %。该方法提高了基于红外的单组分气体定量的速度和精度，并已成功应用于多个钻井现场，满足了气测井的作业要求。

{"title":"Research on quantitative analysis of single-component hydrocarbon gases based on optimized SVR with double-local strategy","authors":"Gang Liu , Peng Han , Hai Yang , Yilin Yao , Haibo Liang","doi":"10.1016/j.vibspec.2025.103865","DOIUrl":"10.1016/j.vibspec.2025.103865","url":null,"abstract":"<div><div>Accurate measurement of gas concentrations in drilling fluids is essential for effective gas logging in drilling engineering, as both accuracy and speed significantly influence real-time monitoring and decision-making. To address the limitations of on-site infrared spectroscopy for single-component gas measurements—such as model mismatches, redundant wavenumbers across broad spectral ranges, and substantial concentration fluctuations that lead to slow processing and large errors, which fail to meet gas logging needs—this study introduces an optimized solution. The approach involves modifying the measurement model, utilizing local characteristic wavenumbers to eliminate redundant spectral information, and dividing local calibration sets to reduce the effects of concentration fluctuations. A novel modeling technique is developed by enhancing Competitive Adaptive Reweighted Sampling (CARS) with Interval Random Frog (IRF) and combining it with the Beetle Antennae Search (BAS) algorithm to optimize Support Vector Regression (SVR). Experimental results demonstrate that the proposed IRF-CARS-BAS-SVR method significantly outperforms traditional techniques in the quantitative analysis of single-component hydrocarbons, achieving an average prediction accuracy exceeding 98 %. This method improves both the speed and precision of infrared-based quantification of single-component gases and has been successfully deployed at various drilling sites, meeting operational requirements for gas logging.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103865"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145525771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A PCA-UVE based feature selection strategy for nutritional component quantification in milk powder using Raman spectroscopy 基于PCA-UVE的拉曼光谱定量奶粉营养成分特征选择策略

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-11-22 DOI: 10.1016/j.vibspec.2025.103867

Xiangchu Li , Yu Ding , Jianan Xu , Yihua He , Qiang Tan , Maoyuan Pang , Weiye Yu , Jinyi Li , Guang Yang , Xinxin Liu

Ensuring the rapid and accurate assessment of nutritional components in milk powder is essential for quality control in the dairy industry. In this study, a Raman spectroscopy-based analytical framework is proposed, integrating principal component analysis (PCA) and uninformative variable elimination (UVE) for feature selection, to optimize spectral modeling and enhance detection performance. A portable Raman system was employed to acquire spectra from 40 commercial milk powder samples, including skimmed, low-fat, and full-fat variants. PCA was initially used to reduce dimensionality and identify informative spectral regions, which were further refined using UVE to eliminate redundant features. The optimized spectral subset was utilized to construct partial least squares regression (PLSR) models for fat and protein prediction, achieving R² values of 0.9865 and 0.9751, respectively, with substantial reductions in RMSE_P and computational cost. Compared to full-spectrum models, the proposed approach reduced processing time by over 70 %, while maintaining high prediction accuracy. This study demonstrates the potential of integrating advanced chemometric methods with Raman spectroscopy for efficient, real-time nutritional analysis in milk powder quality monitoring.

确保奶粉营养成分的快速准确评估是乳品行业质量控制的关键。本文提出了一种基于拉曼光谱的分析框架，结合主成分分析（PCA）和无信息变量消除（UVE）进行特征选择，以优化光谱建模，提高检测性能。采用便携式拉曼系统采集40种商业奶粉样品的光谱，包括脱脂、低脂和全脂变体。该方法首先利用主成分分析（PCA）降维并识别信息丰富的光谱区域，然后利用UVE对光谱区域进行进一步细化以消除冗余特征。利用优化后的谱子集构建脂肪和蛋白质预测的偏最小二乘回归（PLSR）模型，R²值分别为0.9865和0.9751，RMSEP和计算成本大幅降低。与全谱模型相比，该方法在保持较高预测精度的同时，减少了70%以上 %的处理时间。本研究展示了将先进的化学计量方法与拉曼光谱相结合，在奶粉质量监测中进行高效、实时营养分析的潜力。

{"title":"A PCA-UVE based feature selection strategy for nutritional component quantification in milk powder using Raman spectroscopy","authors":"Xiangchu Li , Yu Ding , Jianan Xu , Yihua He , Qiang Tan , Maoyuan Pang , Weiye Yu , Jinyi Li , Guang Yang , Xinxin Liu","doi":"10.1016/j.vibspec.2025.103867","DOIUrl":"10.1016/j.vibspec.2025.103867","url":null,"abstract":"<div><div>Ensuring the rapid and accurate assessment of nutritional components in milk powder is essential for quality control in the dairy industry. In this study, a Raman spectroscopy-based analytical framework is proposed, integrating principal component analysis (PCA) and uninformative variable elimination (UVE) for feature selection, to optimize spectral modeling and enhance detection performance. A portable Raman system was employed to acquire spectra from 40 commercial milk powder samples, including skimmed, low-fat, and full-fat variants. PCA was initially used to reduce dimensionality and identify informative spectral regions, which were further refined using UVE to eliminate redundant features. The optimized spectral subset was utilized to construct partial least squares regression (PLSR) models for fat and protein prediction, achieving R² values of 0.9865 and 0.9751, respectively, with substantial reductions in RMSE<sub>P</sub> and computational cost. Compared to full-spectrum models, the proposed approach reduced processing time by over 70 %, while maintaining high prediction accuracy. This study demonstrates the potential of integrating advanced chemometric methods with Raman spectroscopy for efficient, real-time nutritional analysis in milk powder quality monitoring.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103867"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145624133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Throughput-enhanced grating spectrometers for fiber-optic Raman detection with deep learning-based spectral recovery 基于深度学习光谱恢复的光纤拉曼检测吞吐量增强光栅光谱仪

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-10-20 DOI: 10.1016/j.vibspec.2025.103860

Huijie Wang, Xu Liu, Zichun Yang, Lang Huang, Xinhang Lou, Jianbo Zhu, Linwei Shang, Jianhua Yin

Grating spectrometers are widely utilized to perform the single-shot spectral measurement with relatively high performance, while inherently suffering from a conflict between optical throughput and spectral resolution. As a key component, the entrance slit is required to be narrow for high spectral resolution, inevitably limiting the optical throughput and typically hindering the ultra-weak Raman detection. In this work, the deep learning-based spectral recovery has been proposed and preliminarily explored to numerically eliminate the spectral broadening along with a wider slit in a rapid and accurate way, without demand for complicated physical modification and time-consuming iterative calculation. Given the possible influence of spectral complexity on the spectral recovery, spectral reconstruction is performed in parallel with a set of generative adversarial networks (GANs), following the spectral segmentation and classification in terms of peak number with a convolutional neural network (CNN). For instance, the spectral recovery has been performed on the fiber-optic Raman detection of common drugs, where all the core diameters of excitation and collection fibers are

200 μ m

and far larger than the detector pixel width of

15 μ m

. Through the combination of CNN classification and GAN reconstruction, low-resolution spectra of the 200-

μ

m-width slit with

\sim

7 times higher throughput can be recovered to coincide well with those of the 15-

μ

m-width slit, achieving the optimal spectral resolution. Moreover, the signal-to-noise ratio can be improved by

\sim

3 times on average, promoting more efficient weak-light detection with the flexible fiber-optic probes.

光栅光谱仪被广泛应用于单次光谱测量，具有较高的性能，但其固有的缺点是光学吞吐量和光谱分辨率之间存在冲突。入口狭缝作为关键器件，为了实现高光谱分辨率，需要窄的狭缝，这不可避免地限制了光吞吐量，通常会阻碍超弱拉曼检测。本文提出并初步探索了基于深度学习的光谱恢复方法，在数值上快速准确地消除随狭缝变宽而产生的光谱展宽，不需要进行复杂的物理修正和耗时的迭代计算。考虑到光谱复杂性对光谱恢复的可能影响，在使用卷积神经网络（CNN）对光谱进行峰数分割和分类之后，使用一组生成式对抗网络（gan）并行进行光谱重建。例如，对常见药物的光纤拉曼检测进行了光谱恢复，其中激发和收集光纤的芯直径均为200μm，远大于检测器的像元宽度15μm。通过CNN分类和GAN重构相结合的方法，恢复了吞吐量提高约7倍的200 μm宽狭缝的低分辨率光谱，使其与15 μm宽狭缝的低分辨率光谱吻合良好，达到了最佳的光谱分辨率。此外，信噪比可以平均提高约3倍，从而促进柔性光纤探头更有效地检测弱光。

{"title":"Throughput-enhanced grating spectrometers for fiber-optic Raman detection with deep learning-based spectral recovery","authors":"Huijie Wang, Xu Liu, Zichun Yang, Lang Huang, Xinhang Lou, Jianbo Zhu, Linwei Shang, Jianhua Yin","doi":"10.1016/j.vibspec.2025.103860","DOIUrl":"10.1016/j.vibspec.2025.103860","url":null,"abstract":"<div><div>Grating spectrometers are widely utilized to perform the single-shot spectral measurement with relatively high performance, while inherently suffering from a conflict between optical throughput and spectral resolution. As a key component, the entrance slit is required to be narrow for high spectral resolution, inevitably limiting the optical throughput and typically hindering the ultra-weak Raman detection. In this work, the deep learning-based spectral recovery has been proposed and preliminarily explored to numerically eliminate the spectral broadening along with a wider slit in a rapid and accurate way, without demand for complicated physical modification and time-consuming iterative calculation. Given the possible influence of spectral complexity on the spectral recovery, spectral reconstruction is performed in parallel with a set of generative adversarial networks (GANs), following the spectral segmentation and classification in terms of peak number with a convolutional neural network (CNN). For instance, the spectral recovery has been performed on the fiber-optic Raman detection of common drugs, where all the core diameters of excitation and collection fibers are <span><math><mrow><mn>200</mn><mspace></mspace><mi>μ</mi><mi>m</mi></mrow></math></span> and far larger than the detector pixel width of <span><math><mrow><mn>15</mn><mspace></mspace><mi>μ</mi><mi>m</mi></mrow></math></span>. Through the combination of CNN classification and GAN reconstruction, low-resolution spectra of the 200-<span><math><mi>μ</mi></math></span>m-width slit with <span><math><mo>∼</mo></math></span>7 times higher throughput can be recovered to coincide well with those of the 15-<span><math><mi>μ</mi></math></span>m-width slit, achieving the optimal spectral resolution. Moreover, the signal-to-noise ratio can be improved by <span><math><mo>∼</mo></math></span>3 times on average, promoting more efficient weak-light detection with the flexible fiber-optic probes.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103860"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The usage of ATR-FTIR spectroscopy combined with chemometrics in postmortem diagnosis of acute myocardial infarction ATR-FTIR光谱联合化学计量学在急性心肌梗死死后诊断中的应用

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-09-04 DOI: 10.1016/j.vibspec.2025.103857

Mustafa Talip Sener , Nihal Simsek Ozek , Ferhunde Aysin , Sezen Yildiz , Ozkan Aksakal

Myocardial infarction (MI) represents a leading cause of sudden death; however, the identification of acute MI in postmortem examinations remains challenging due to the lack of specific early morphological markers. This study explores the utility of attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy, in combination with chemometric analysis, for the postmortem diagnosis of MI. A rat model of MI was employed, and the expression levels and protein concentrations of JUNB and myoglobin (MYO) were quantified to corroborate the spectral findings. ATR-FTIR analysis revealed significant alterations in collagen and nucleic acid content between the control and MI groups, particularly during the early postmortem intervals. Principal component analysis (PCA) and hierarchical cluster analysis (HCA) successfully differentiated between groups, achieving 100 % sensitivity and specificity, and 97 % classification accuracy through PCA-based linear discriminant analysis (LDA). Support vector machine (SVM) classification further confirmed these findings. Notably, spectral changes at 3285, 3016, 2920, 1338, 1236, and 974 cm⁻¹ exhibited strong correlations with JUNB and MYO gene and protein levels. This integrative approach demonstrates that ATR-FTIR spectroscopy, combined with multivariate analysis and molecular markers, offers a rapid, non-destructive, and highly accurate method for the early postmortem diagnosis of acute MI, highlighting its potential for forensic and clinical applications.

心肌梗死（MI）是猝死的主要原因；然而，由于缺乏特定的早期形态学标记，在死后检查中鉴定急性心肌梗死仍然具有挑战性。本研究探讨了衰减全反射-傅里叶变换红外光谱（ATR-FTIR）结合化学计量学分析在心肌梗死死后诊断中的应用。采用心肌梗死大鼠模型，定量分析了JUNB和肌红蛋白（MYO）的表达水平和蛋白浓度，以证实光谱结果。ATR-FTIR分析显示，对照组和心肌梗死组之间的胶原蛋白和核酸含量发生了显著变化，尤其是在死后早期。主成分分析（PCA）和层次聚类分析（HCA）通过基于PCA的线性判别分析（LDA）实现了组间的成功区分，灵敏度和特异度分别达到100 %和97 %。支持向量机（SVM）分类进一步证实了这些发现。值得注意的是，3285、3016、2920、1338、1236和974 cm⁻¹ 的光谱变化与JUNB和MYO基因和蛋白质水平有很强的相关性。这种综合方法表明，ATR-FTIR光谱与多变量分析和分子标记相结合，为急性心肌梗死的早期死后诊断提供了一种快速、非破坏性和高度准确的方法，突出了其在法医和临床应用中的潜力。

{"title":"The usage of ATR-FTIR spectroscopy combined with chemometrics in postmortem diagnosis of acute myocardial infarction","authors":"Mustafa Talip Sener , Nihal Simsek Ozek , Ferhunde Aysin , Sezen Yildiz , Ozkan Aksakal","doi":"10.1016/j.vibspec.2025.103857","DOIUrl":"10.1016/j.vibspec.2025.103857","url":null,"abstract":"<div><div>Myocardial infarction (MI) represents a leading cause of sudden death; however, the identification of acute MI in postmortem examinations remains challenging due to the lack of specific early morphological markers. This study explores the utility of attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy, in combination with chemometric analysis, for the postmortem diagnosis of MI. A rat model of MI was employed, and the expression levels and protein concentrations of JUNB and myoglobin (MYO) were quantified to corroborate the spectral findings. ATR-FTIR analysis revealed significant alterations in collagen and nucleic acid content between the control and MI groups, particularly during the early postmortem intervals. Principal component analysis (PCA) and hierarchical cluster analysis (HCA) successfully differentiated between groups, achieving 100 % sensitivity and specificity, and 97 % classification accuracy through PCA-based linear discriminant analysis (LDA). Support vector machine (SVM) classification further confirmed these findings. Notably, spectral changes at 3285, 3016, 2920, 1338, 1236, and 974 cm⁻¹ exhibited strong correlations with JUNB and MYO gene and protein levels. This integrative approach demonstrates that ATR-FTIR spectroscopy, combined with multivariate analysis and molecular markers, offers a rapid, non-destructive, and highly accurate method for the early postmortem diagnosis of acute MI, highlighting its potential for forensic and clinical applications.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103857"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145005239","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

In-situ SERS monitoring of light-gated reaction switching on magnetic-plasmonic CoFe2O4@TiO2@Ag nanorods 磁等离子体CoFe2O4@TiO2@Ag纳米棒上光门控反应开关的原位SERS监测

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-10-26 DOI: 10.1016/j.vibspec.2025.103864

Shuang Wang , Zian Li , Wei Lv , Guolin Zhang , Xiaoqi Fu , Juan Yang

Here, we fabricate magnetically recyclable CoFe₂O₄@TiO₂@Ag nanorods (NRs) for in-situ surface-enhanced Raman scattering (SERS) monitoring of light-gated catalytic pathways. The hierarchical structure, comprising a magnetic CoFe₂O₄ NR core, a TiO₂ interlayer, and Ag nanoparticles (NPs) shell, enables the SERS detection of rhodamine 6 G (R6G) at a low concentration of 10⁻⁸ mol L⁻¹ with excellent signal reproducibility. In-situ SERS reveals distinct reduction mechanisms for 4-nitrothiophenol (4-NTP) governed by the excitation wavelength. Under 532 nm visible laser irradiation, plasmon-induced hot electrons from Ag NPs drive the selective conversion of 4-NTP to trans-dimercaptoazobenzene (DMAB) with a rate constant of 0.118 min⁻¹. In contrast, under 365 nm UV light, TiO₂-mediated electron transfer promotes the formation of cis-DMAB at a rate constant of 0.034 min⁻¹. Reversible cis-trans isomerization is achieved by alternating the light sources. This work establishes CoFe₂O₄/TiO₂/Ag NRs as a versatile platform for monitoring light-gated catalysis in real time, with promising applications in energy conversion, environmental remediation, and selective chemical transformations.

在这里，我们制造了磁性可回收的CoFe2O4@TiO2@Ag纳米棒（nr），用于光门控催化途径的原位表面增强拉曼散射（SERS）监测。由磁性CoFe2O4 NR核、TiO2中间层和Ag纳米颗粒（NPs）壳组成的分层结构，使罗丹明6 G （R6G）在低浓度（10−8 mol L−1）下的SERS检测具有良好的信号再现性。原位SERS揭示了4-硝基苯酚（4-NTP）在不同激发波长下的不同还原机制。在532 nm可见激光照射下，Ag纳米粒子的等离子体诱导热电子驱动4-NTP选择性转化为反式二巯基偶氮苯（DMAB），速率常数为0.118 min−1。相反，在365 nm紫外光下，tio2介导的电子转移促进顺式dmab的形成，速率常数为0.034 min−1。可逆顺反异构化是通过交替光源实现的。这项工作建立了CoFe2O4/TiO2/Ag NRs作为实时监测光门控催化的多功能平台，在能量转换，环境修复和选择性化学转化方面具有广阔的应用前景。

{"title":"In-situ SERS monitoring of light-gated reaction switching on magnetic-plasmonic CoFe2O4@TiO2@Ag nanorods","authors":"Shuang Wang , Zian Li , Wei Lv , Guolin Zhang , Xiaoqi Fu , Juan Yang","doi":"10.1016/j.vibspec.2025.103864","DOIUrl":"10.1016/j.vibspec.2025.103864","url":null,"abstract":"<div><div>Here, we fabricate magnetically recyclable CoFe<sub>2</sub>O<sub>4</sub>@TiO<sub>2</sub>@Ag nanorods (NRs) for <em>in-situ</em> surface-enhanced Raman scattering (SERS) monitoring of light-gated catalytic pathways. The hierarchical structure, comprising a magnetic CoFe<sub>2</sub>O<sub>4</sub> NR core, a TiO<sub>2</sub> interlayer, and Ag nanoparticles (NPs) shell, enables the SERS detection of rhodamine 6 G (R6G) at a low concentration of 10<sup>−8</sup> mol L<sup>−1</sup> with excellent signal reproducibility. <em>In-situ</em> SERS reveals distinct reduction mechanisms for 4-nitrothiophenol (4-NTP) governed by the excitation wavelength. Under 532 nm visible laser irradiation, plasmon-induced hot electrons from Ag NPs drive the selective conversion of 4-NTP to <em>trans</em>-dimercaptoazobenzene (DMAB) with a rate constant of 0.118 min<sup>−1</sup>. In contrast, under 365 nm UV light, TiO<sub>2</sub>-mediated electron transfer promotes the formation of <em>cis</em>-DMAB at a rate constant of 0.034 min<sup>−1</sup>. Reversible <em>cis</em>-<em>trans</em> isomerization is achieved by alternating the light sources. This work establishes CoFe<sub>2</sub>O<sub>4</sub>/TiO<sub>2</sub>/Ag NRs as a versatile platform for monitoring light-gated catalysis in real time, with promising applications in energy conversion, environmental remediation, and selective chemical transformations.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103864"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145416972","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Comparative assessment of serum dilution and pure samples for Raman-based oral cancer detection: Evaluating dilution rate performance 血清稀释和纯样品用于拉曼检测口腔癌的比较评估：评估稀释率性能

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-11-01 Epub Date: 2025-10-17 DOI: 10.1016/j.vibspec.2025.103858

Mukta Sharma , Ajay Kumar , Chia-Lung Tsai , Shiang-Fu Huang , Yu-Li Hsieh , Liann-Be Chang

Raman spectroscopy is a powerful non-invasive tool for biomolecular analysis, offering potential for early oral cancer detection. However, the impact of serum dilution rates on classification performance remains underexplored. This study evaluates the diagnostic efficacy of pure, 1:1, and 1:2 diluted serum samples for classifying malignant, premalignant, and normal groups using machine learning techniques. Serum samples from 80 subjects (33 malignant, 27 premalignant, and 20 normal) were analyzed via Raman spectroscopy. Spectral data were preprocessed and reduced using Principal Component Analysis (PCA), followed by classification using Support Vector Machines (SVM) and Random Forest (RF). Notably, the 1:2 dilution rate (DR) demonstrated comparable diagnostic performance to pure serum while preserving critical spectral features. For PCA-SVM under 1:2 DR dilution (with 90.5% accuracy) and pure serum conditions (with 92.5% accuracy), the sensitivity and specificity for malignant samples were 93.9% and 87.2%, & 91% and 94%, respectively. For normal samples, these values were 95% and 100% for the 1:2 dilution, and 95% and 98.3% for pure serum. Although both classifiers differentiated successfully, PCA-SVM demonstrated slightly better performance compared to PCA-RF across dilution rates. Bar charts confirmed consistent spectral trends across DRs. The study demonstrates the potential of Raman spectroscopy combined with PCA-SVM for accurate oral cancer detection. The selected 1:2 DR balances sample conservation with robust diagnostic performance, offering a practical and efficient approach for clinical applications. This work emphasizes the feasibility of using diluted biosamples for machine learning-based cancer diagnostics, achieving performance comparable to pure samples.

拉曼光谱是一种强大的非侵入性生物分子分析工具，为早期口腔癌检测提供了潜力。然而，血清稀释率对分类性能的影响仍未得到充分探讨。本研究评估了使用机器学习技术对纯、1:1和1:2稀释血清样本进行恶性、癌前和正常组分类的诊断效果。采用拉曼光谱对80例受试者（33例恶性，27例癌前病变，20例正常）的血清样本进行分析。利用主成分分析（PCA）对光谱数据进行预处理和约简，然后利用支持向量机（SVM）和随机森林（RF）进行分类。值得注意的是，1:2稀释率（DR）与纯血清的诊断性能相当，同时保留了关键的光谱特征。在1:2 DR稀释（准确率为90.5%）和纯血清条件下（准确率为92.5%），PCA-SVM对恶性样本的敏感性和特异性分别为93.9%和87.2%，91%和94%。对于正常样品，1:2稀释时，这些值分别为95%和100%，纯血清为95%和98.3%。尽管两种分类器都能成功区分，但PCA-SVM在稀释率上比PCA-RF表现出稍好的性能。柱状图证实了各dr间一致的光谱趋势。本研究证明了拉曼光谱结合PCA-SVM对口腔癌进行准确检测的潜力。所选择的1:2 DR平衡了样本保存与强大的诊断性能，为临床应用提供了实用而有效的方法。这项工作强调了使用稀释生物样本进行基于机器学习的癌症诊断的可行性，实现了与纯样本相当的性能。

{"title":"Comparative assessment of serum dilution and pure samples for Raman-based oral cancer detection: Evaluating dilution rate performance","authors":"Mukta Sharma , Ajay Kumar , Chia-Lung Tsai , Shiang-Fu Huang , Yu-Li Hsieh , Liann-Be Chang","doi":"10.1016/j.vibspec.2025.103858","DOIUrl":"10.1016/j.vibspec.2025.103858","url":null,"abstract":"<div><div>Raman spectroscopy is a powerful non-invasive tool for biomolecular analysis, offering potential for early oral cancer detection. However, the impact of serum dilution rates on classification performance remains underexplored. This study evaluates the diagnostic efficacy of pure, 1:1, and 1:2 diluted serum samples for classifying malignant, premalignant, and normal groups using machine learning techniques. Serum samples from 80 subjects (33 malignant, 27 premalignant, and 20 normal) were analyzed via Raman spectroscopy. Spectral data were preprocessed and reduced using Principal Component Analysis (PCA), followed by classification using Support Vector Machines (SVM) and Random Forest (RF). Notably, the 1:2 dilution rate (DR) demonstrated comparable diagnostic performance to pure serum while preserving critical spectral features. For PCA-SVM under 1:2 DR dilution (with 90.5% accuracy) and pure serum conditions (with 92.5% accuracy), the sensitivity and specificity for malignant samples were 93.9% and 87.2%, & 91% and 94%, respectively. For normal samples, these values were 95% and 100% for the 1:2 dilution, and 95% and 98.3% for pure serum. Although both classifiers differentiated successfully, PCA-SVM demonstrated slightly better performance compared to PCA-RF across dilution rates. Bar charts confirmed consistent spectral trends across DRs. The study demonstrates the potential of Raman spectroscopy combined with PCA-SVM for accurate oral cancer detection. The selected 1:2 DR balances sample conservation with robust diagnostic performance, offering a practical and efficient approach for clinical applications. This work emphasizes the feasibility of using diluted biosamples for machine learning-based cancer diagnostics, achieving performance comparable to pure samples.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"141 ","pages":"Article 103858"},"PeriodicalIF":3.1,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145362735","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Fast ATR-FTIR method for quantifying silicates presence on PE plastic fragments from soil 快速ATR-FTIR法定量土壤PE塑料碎片上硅酸盐的存在

IF 2.7 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-09-01 Epub Date: 2025-07-09 DOI: 10.1016/j.vibspec.2025.103833

David Picón-Borregales , Leticia Pastormerlo , Eduardo Reciulschi , Javier M. Montserrat

The interaction of plastic debris with the soil environment remains insufficiently studied. In particular, we have recently reported the incorporation of a mechanically stable clay phase—mainly composed of silicates—onto polyethylene (PE) macro-, meso-, and microplastic surfaces. This incorporation transforms plastic fragments into a composite material, potentially leading to significant changes in properties such as density, hydrophobicity, and contaminant sorption capacity. Therefore, quantifying the siliceous fraction is essential to better understand plastic–environment interactions. Determination of silicon by EDX is a conventional method, but is time-consuming, technically demanding, and not widely accessible. Moreover, the presence of clay onto the PE matrix complicates the identification of oxygen-containing functional groups due to spectral overlap between C–O and Si–O stretching vibrations in the sample's FTIR spectra. In this study, a rapid and straightforward ATR-FTIR-based methodology for the quantitative determination of silicon on weathered PE mulch fragments was developed. Furthermore, a reliable approach for the identification of Si–O and C–O functional groups in PE samples with high silicon content was established. The peak area of the Si–O stretching band showed a strong linear correlation with silicon concentration in PE–sand standards (R²=0.9878). The proposed method was validated against EDX measurements of PE samples extracted from agricultural soils, showing good agreement. Additionally, sodium citrate treatment effectively removed the siliceous fraction without the use of hazardous hydrofluoric acid, allowing for accurate determination of oxidation indices. The developed method is simple, rapid, and requires minimal sample preparation, offering a practical alternative for laboratories lacking access to advanced analytical techniques.

塑料垃圾与土壤环境的相互作用研究尚不充分。特别是，我们最近报道了在聚乙烯（PE）宏观、中观和微塑性表面上掺入一种主要由硅酸盐组成的机械稳定粘土相。这种结合将塑料碎片转化为复合材料，可能导致密度、疏水性和污染物吸附能力等性能的显著变化。因此，量化硅质组分对于更好地理解塑料与环境的相互作用至关重要。EDX法测定硅是一种传统的方法，但耗时长，技术要求高，而且应用范围不广。此外，由于样品FTIR光谱中C-O和Si-O拉伸振动之间的光谱重叠，PE基体上粘土的存在使含氧官能团的识别变得复杂。在这项研究中，开发了一种快速、直接的基于atr - ftir的方法，用于定量测定风化PE覆盖物碎片上的硅。此外，建立了一种鉴定高硅PE样品中Si-O和C-O官能团的可靠方法。Si-O拉伸带的峰面积与pe砂标准硅浓度呈较强的线性相关（R²=0.9878）。该方法与农业土壤中PE样品的EDX测量结果进行了验证，结果吻合良好。此外，柠檬酸钠处理有效地去除了硅质部分，而无需使用有害的氢氟酸，从而可以准确测定氧化指数。所开发的方法简单，快速，并且需要最少的样品制备，为缺乏先进分析技术的实验室提供了一种实用的替代方法。

{"title":"Fast ATR-FTIR method for quantifying silicates presence on PE plastic fragments from soil","authors":"David Picón-Borregales , Leticia Pastormerlo , Eduardo Reciulschi , Javier M. Montserrat","doi":"10.1016/j.vibspec.2025.103833","DOIUrl":"10.1016/j.vibspec.2025.103833","url":null,"abstract":"<div><div>The interaction of plastic debris with the soil environment remains insufficiently studied. In particular, we have recently reported the incorporation of a mechanically stable clay phase—mainly composed of silicates—onto polyethylene (PE) macro-, meso-, and microplastic surfaces. This incorporation transforms plastic fragments into a composite material, potentially leading to significant changes in properties such as density, hydrophobicity, and contaminant sorption capacity. Therefore, quantifying the siliceous fraction is essential to better understand plastic–environment interactions. Determination of silicon by EDX is a conventional method, but is time-consuming, technically demanding, and not widely accessible. Moreover, the presence of clay onto the PE matrix complicates the identification of oxygen-containing functional groups due to spectral overlap between C–O and Si–O stretching vibrations in the sample's FTIR spectra. In this study, a rapid and straightforward ATR-FTIR-based methodology for the quantitative determination of silicon on weathered PE mulch fragments was developed. Furthermore, a reliable approach for the identification of Si–O and C–O functional groups in PE samples with high silicon content was established. The peak area of the Si–O stretching band showed a strong linear correlation with silicon concentration in PE–sand standards (R²=0.9878). The proposed method was validated against EDX measurements of PE samples extracted from agricultural soils, showing good agreement. Additionally, sodium citrate treatment effectively removed the siliceous fraction without the use of hazardous hydrofluoric acid, allowing for accurate determination of oxidation indices. The developed method is simple, rapid, and requires minimal sample preparation, offering a practical alternative for laboratories lacking access to advanced analytical techniques.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"140 ","pages":"Article 103833"},"PeriodicalIF":2.7,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656033","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Identification of environmental microplastics using large language models: DeepSeek-R1-Distill-Llama-8B, GPT-4o, and GPT-4o-mini 使用大型语言模型识别环境微塑料：deepseek - r1 -蒸馏- llama - 8b， gpt - 40和gpt - 40 -mini

IF 3.1 3区化学 Q2 CHEMISTRY, ANALYTICAL

Vibrational Spectroscopy

Pub Date : 2025-09-01 Epub Date: 2025-07-23 DOI: 10.1016/j.vibspec.2025.103842

Zijiang Yang , Hisayuki Arakawa

Microplastic pollution in the environment poses increasing risks to both ecological and human health. Identifying microplastics in environmental samples is important for monitoring and mitigation. However, current methods rely on manual interpretation of infrared (IR) spectra, which is time-consuming and labor-intensive. Thus, this study investigates the potential of large language models (LLMs) for identifying microplastics using IR spectra from environmental samples. Three models, DeepSeek-R1-Distill-Llama-8B, GPT-4o-2024–08–06 (GPT-4o), and GPT-4o-mini-2024–07–18 (GPT-4o-mini), were evaluated within a structured workflow that integrates spectral processing and model implementation. A performance evaluation framework was developed to measure identification accuracy. Results indicate that DeepSeek-R1-Distill-Llama-8B outperformed others, achieving an accuracy exceeding 0.93 across all tested polymer types, making it the preferred choice. GPT-4o proved a strong alternative, particularly when local execution is impractical, with accuracy above 0.86. GPT-4o-mini underperformed and is not recommended. Despite these promising outcomes, challenges persist, including the need to optimize spectral processing parameters and refine prompt design. As the first study to apply LLMs to microplastic identification, this work offers a foundational reference for leveraging LLM-driven spectral analysis in environmental monitoring.

环境中的微塑料污染对生态和人类健康构成越来越大的风险。鉴定环境样品中的微塑料对于监测和缓解至关重要。然而，目前的方法依赖于人工解释红外光谱，这是费时费力的。因此，本研究探讨了利用环境样品的红外光谱识别微塑料的大语言模型（LLMs）的潜力。DeepSeek-R1-Distill-Llama-8B、gpt - 40 -2024 - 08 - 06 （gpt - 40）和gpt - 40 -mini-2024 - 07 - 18 （gpt - 40 -mini）三种模型在集成了光谱处理和模型实现的结构化工作流程中进行了评估。开发了一个性能评估框架来衡量识别的准确性。结果表明，DeepSeek-R1-Distill-Llama-8B优于其他方法，在所有测试的聚合物类型中实现了超过0.93的精度，使其成为首选。gpt - 40被证明是一个强大的替代方案，特别是当本地执行不切实际时，其精度高于0.86。gpt - 40 -mini表现不佳，不推荐使用。尽管取得了这些有希望的成果，但挑战依然存在，包括需要优化光谱处理参数和改进提示设计。作为第一个将llm应用于微塑料识别的研究，本工作为利用llm驱动的光谱分析在环境监测中提供了基础参考。

{"title":"Identification of environmental microplastics using large language models: DeepSeek-R1-Distill-Llama-8B, GPT-4o, and GPT-4o-mini","authors":"Zijiang Yang , Hisayuki Arakawa","doi":"10.1016/j.vibspec.2025.103842","DOIUrl":"10.1016/j.vibspec.2025.103842","url":null,"abstract":"<div><div>Microplastic pollution in the environment poses increasing risks to both ecological and human health. Identifying microplastics in environmental samples is important for monitoring and mitigation. However, current methods rely on manual interpretation of infrared (IR) spectra, which is time-consuming and labor-intensive. Thus, this study investigates the potential of large language models (LLMs) for identifying microplastics using IR spectra from environmental samples. Three models, DeepSeek-R1-Distill-Llama-8B, GPT-4o-2024–08–06 (GPT-4o), and GPT-4o-mini-2024–07–18 (GPT-4o-mini), were evaluated within a structured workflow that integrates spectral processing and model implementation. A performance evaluation framework was developed to measure identification accuracy. Results indicate that DeepSeek-R1-Distill-Llama-8B outperformed others, achieving an accuracy exceeding 0.93 across all tested polymer types, making it the preferred choice. GPT-4o proved a strong alternative, particularly when local execution is impractical, with accuracy above 0.86. GPT-4o-mini underperformed and is not recommended. Despite these promising outcomes, challenges persist, including the need to optimize spectral processing parameters and refine prompt design. As the first study to apply LLMs to microplastic identification, this work offers a foundational reference for leveraging LLM-driven spectral analysis in environmental monitoring.</div></div>","PeriodicalId":23656,"journal":{"name":"Vibrational Spectroscopy","volume":"140 ","pages":"Article 103842"},"PeriodicalIF":3.1,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144739301","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0