Chemometrics and Intelligent Laboratory Systems最新文献

{"title":"Measurement uncertainty and risk of false decisions for similarity factor (f2): Boostrapping method or spreadsheet?","authors":"Jheniffer Rabelo,&nbsp;Felipe R. Lourenço","doi":"10.1016/j.chemolab.2025.105475","DOIUrl":"10.1016/j.chemolab.2025.105475","url":null,"abstract":"<div><div>The similarity factor (<em>f</em>_<em>2</em>) is a key metric to compare generic and reference drug and to obtain biowaivers of bioequivalence studies of Active Product Ingredient (API) in dissolution testing. Yet, its statistical limitations - including low power and undefined confidence levels - restrict its reliability. Therefore, Bootstrapping is a widely used approach for establishing Confidence Interval for <em>f</em>_<em>2</em>. Nevertheless, this approach is not user-friendly for non-statisticians, and to obtain the uncertainty related to the risk assessment of the <em>f</em>_<em>2</em> in the dissolution testing requires additional steps. In this investigation, we propose the Kragten spreadsheet method as a practical alternative to Bootstrapping approach in the evaluation of the consumer's risk. Comparative analysis was performed under six scenarios of API's, involving reference/test drugs and registered/new formulations. All six groups met regulatory criteria (<em>f</em>_<em>2</em>&gt;50 %), while 95 % confidence intervals from both statistical methods showed agreement, confirming methodological reliability. Despite all groups achieved the <em>f</em>_<em>2</em>&gt;50 %, the range obtained through Bootstrapping and Kragten methods determined that three scenarios (A, C, and E) presented elevated consumer risk (&gt;5 %), highlighting limitations of <em>f</em>_<em>2</em> alone. Also, the differences between both methodologies were measured. For Bootstrapping, 50 iterations per 10.000 simulations showed no statistically significant differences from Kragten (<em>p</em> &gt; 0.05), establishing method equivalence for symmetrical dissolution data. The findings advocate combining <em>f</em>_<em>2</em> with uncertainty analysis for risk assessment in dissolution testing.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105475"},"PeriodicalIF":3.7,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144516932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"(ϵ,δ)-differentially private partial least squares regression","authors":"Ramin Nikzad-Langerodi ,&nbsp;Mohit Kumar ,&nbsp;Du Nguyen Duy ,&nbsp;Mahtab Alghasi","doi":"10.1016/j.chemolab.2025.105465","DOIUrl":"10.1016/j.chemolab.2025.105465","url":null,"abstract":"<div><div>As data-privacy requirements are becoming increasingly stringent and statistical models based on sensitive data are being deployed and used more routinely, protecting data-privacy becomes pivotal. Partial Least Squares (PLS) regression is the premier tool for building such models in analytical chemistry, yet it does not inherently provide privacy guarantees, leaving sensitive (training) data vulnerable to privacy attacks. To address this gap, we propose an <span><math><mrow><mo>(</mo><mi>ϵ</mi><mo>,</mo><mi>δ</mi><mo>)</mo></mrow></math></span>-differentially private PLS (edPLS) algorithm, which integrates well-studied and theoretically motivated Gaussian noise-adding mechanisms into the PLS algorithm to ensure the privacy of the data underlying the model. Our approach involves adding carefully calibrated Gaussian noise to the outputs of four key functions in the PLS algorithm: the weights, scores, <span><math><mi>X</mi></math></span>-loadings, and <span><math><mi>Y</mi></math></span>-loadings. The noise variance is determined based on the sensitivity of each function, ensuring that the privacy loss is controlled according to the <span><math><mrow><mo>(</mo><mi>ϵ</mi><mo>,</mo><mi>δ</mi><mo>)</mo></mrow></math></span>-differential privacy framework. Specifically, we derive the sensitivity bounds for each function and use these bounds to calibrate the noise added to the model components. Experimental results demonstrate that edPLS effectively renders privacy attacks, aimed at recovering unique sources of variability in the training data, ineffective. Application of edPLS to the NIR corn benchmark dataset shows that the root mean squared error of prediction (RMSEP) remains competitive even at strong privacy levels (i.e., <span><math><mrow><mi>ϵ</mi><mo>=</mo><mn>1</mn></mrow></math></span>), given proper pre-processing of the corresponding spectra. These findings highlight the practical utility of edPLS in creating privacy-preserving multivariate calibrations and for the analysis of their privacy-utility trade-offs.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105465"},"PeriodicalIF":3.7,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144502427","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A new segmentation-assisted interpolation method for creating maps in the study of artworks","authors":"Domingo Martín ,&nbsp;Germán Arroyo ,&nbsp;Juan Carlos Torres ,&nbsp;Luis López ,&nbsp;María Rosario Blanc ,&nbsp;Juan Ruiz de Miras","doi":"10.1016/j.chemolab.2025.105466","DOIUrl":"10.1016/j.chemolab.2025.105466","url":null,"abstract":"<div><div>The generation of spatial distribution maps of chemical elements and compounds has become a crucial technique in materials research, particularly in the analysis of artworks. However, data acquisition in this context is often limited by the low number of measured points relative to the visual complexity of the artwork. As a result, interpolation methods are employed to infer unmeasured data. The most widely used method, Minimum Hypercube Distance (MHD), although statistically validated, exhibits significant limitations, as demonstrated in this study. We identified errors of up to 100% in some cases, exposing the method’s vulnerability in regions lacking sufficient data. To address these challenges, we propose a novel segmentation-assisted interpolation method. By integrating semantic segmentation, this approach improves the accuracy and interpretability of the resulting maps, allowing for the precise identification of unmeasured areas and the expert-guided replication of data from similar regions. This new methodology enhances the robustness of artwork analysis, providing more reliable tools for the study and preservation of artworks and ancient monuments.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105466"},"PeriodicalIF":3.7,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Distortion prediction model considering process types in film manufacturing process and identification of critical process variables","authors":"Yuta Wakutsu ,&nbsp;Satoshi Natori ,&nbsp;Hiroki Ochiai ,&nbsp;Kazuya Suda ,&nbsp;Hiromasa Kaneko","doi":"10.1016/j.chemolab.2025.105474","DOIUrl":"10.1016/j.chemolab.2025.105474","url":null,"abstract":"<div><div>Optical films are used in flat panel displays, touch sensors and other devices. The film is wound and sent to the next process, but defects are generated in the winding, indicating an issue. Defects are often identified by customers after shipment. It is therefore anticipated that the identification of characteristics associated with defects at the end of the process will facilitate the detection of defective products in advance, or alternatively, result in a reduction of the defects themselves. One of the factors that contribute to the occurrence of defects is the distortion of the film on the roll surface. The degree of distortion is determined by calculating the difference between the instantaneous and average distance between the film surface and the sensor, as measured by the displacement meter installed before the winding. The objective of this study was to identify the process conditions that cause the distortion of the film as measured by the displacement meter through the application of machine learning techniques. A model was constructed between the sensor data of the process conditions and the distortion index, the relationship between them was identified, and the process causing the distortion was estimated by analyzing the model. The results of this study successfully narrowed down the process variables that are common causes of distortion among three displacement meters with different measurement positions.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105474"},"PeriodicalIF":3.7,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491880","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A likelihood ratio model for three-way data coupled with a Tucker3 model","authors":"Agnieszka Martyna ,&nbsp;Eugenio Alladio ,&nbsp;Monica Romagnoli ,&nbsp;Fabrizio Malaspina ,&nbsp;Marco Pazzi","doi":"10.1016/j.chemolab.2025.105464","DOIUrl":"10.1016/j.chemolab.2025.105464","url":null,"abstract":"<div><div>In forensic science, the analysis of diesel fuel is particularly important in fire investigations. The task is usually to compare the original accelerant found among the suspect’s belongings with the fire debris to find out if the fire could have been caused by the use of that particular diesel fuel (called source). The major problem when comparing the original accelerant with the fire debris is the weathering process of the accelerant taking place during the fire. The weathering process makes the composition of the accelerant change with the weathering state and may differ from the composition of the original accelerant. In this context the question arises if samples of the fire debris containing the accelerant weathered to different degrees are still so similar to the original accelerant that they can be regarded as coming from the same source (this particular accelerant) and whether samples of fire debris with accelerants from different sources are easily identified as such regardless of their weathering state. The hybrid likelihood ratio (LR) model which takes into account the information about the similarity and the frequency of observing the compared features in the samples was used for answering the above issues. Hybrid LR models use the new set of a limited number of variables that is generated using a variety of chemometric tools to summarise the data as well as possible and highlight the features that make each source of samples uniquely defined. The model was built for three-way GC–MS data of diesel fuel samples. Tucker3 model decomposed the three-dimensional array of the database into three matrices referring to GC, MS and samples (concentration) modes. The scores on the linear discriminant functions for the concentration mode served as an input for LR models. True origins for the majority of samples were indicated despite different weathering.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105464"},"PeriodicalIF":3.7,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144514020","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing machine learning models for near-infrared regression by measuring stability towards diffeomorphisms","authors":"Mark Wohlers ,&nbsp;V.A. McGlone ,&nbsp;Eibe Frank ,&nbsp;Geoffrey Holmes","doi":"10.1016/j.chemolab.2025.105449","DOIUrl":"10.1016/j.chemolab.2025.105449","url":null,"abstract":"<div><div>Near infrared (NIR) spectroscopy is widely used as a tool for non-destructive assessment of fruit quality by applying measured spectra to predict quality parameters such as dry matter and soluble solids content using a suitable regression method. With continued advancements in deep learning, there is potential for improved predictive performance when neural network models are applied instead of partial least-squares regression, but choosing a model remains challenging as performance is sensitive to the model’s architecture. Taking inspiration from work done in image classification, we propose model selection by assessing relative stability to diffeomorphic transformations, providing a complementary approach to standard validation methods. This is particularly useful when labelled validation data is limited. Our empirical results on several NIR regression problems indicate that the proposed approach is comparable to the use of independent validation sets. In addition to the choice of deep learning architecture, we also consider the selection of the number of components in partial least-squares regression to demonstrate the method’s generality.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105449"},"PeriodicalIF":3.7,"publicationDate":"2025-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144517838","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Selection of important samples for calibration model updating based on kernel coefficients","authors":"Zhongjiang He ,&nbsp;Zhonghai He ,&nbsp;Xiaofang Zhang","doi":"10.1016/j.chemolab.2025.105472","DOIUrl":"10.1016/j.chemolab.2025.105472","url":null,"abstract":"<div><div>The predictive performance of PLS models often diminishes for new samples outside the application domain and requires updating. The conventional approach involves augmenting the calibration set with new samples to accommodate changes. However, the efficiency of updating is hampered by the large number of samples in the old calibration set. To expedite the updating process, it is crucial to delete some samples, maintaining a balanced class ratio between old and new samples. Surprisingly, there has been a lack of discussion on how to select old samples thus far. This paper introduces a method known as Kernel Coefficient Selection (KCS), designed to obtain coefficients for each sample in the calibration model and subsequently identify crucial old samples. A large coefficient suggests the sample's significance and the need for retention. The rationale behind this method is grounded in the theory of the duality of two types of models (similarity-based and non-similarity-based). Removing samples with small coefficients from the calibration set still preserves the integrity of the major regression model. Comparative experiments were conducted between models that retained partial important samples and those involving all samples, using both simulated and real soybean meal datasets. The experimental results demonstrate that the KCS method mitigates the issue of number imbalance between old and new samples, thereby enhancing the efficiency of model updating.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105472"},"PeriodicalIF":3.7,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144469979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Target Regression via Target Combinations using PCA with application to quality prediction in OLED manufacturing process","authors":"Takafumi Yamaguchi ,&nbsp;Sou Nakamura ,&nbsp;Yuta Yamaguchi ,&nbsp;Yoshiyuki Yamashita","doi":"10.1016/j.chemolab.2025.105462","DOIUrl":"10.1016/j.chemolab.2025.105462","url":null,"abstract":"<div><div>This study focuses on improving quality prediction in Organic Light Emitting Diode (OLED) manufacturing using multi-target regression (MTR), which is called the MTR via Target Combinations Using Principal Component Analysis method (PCA-C). OLED is a complex device, fabricated by stacking multiple layers of light-emitting materials, but the production quality is difficult to control. The PCA-C method is designed to enhance prediction accuracy by considering relationships among multiple quality inspection variables, such as CIE-x, CIE-y, luminance, Ra, and R9, which define color and luminance attributes of OLED panels. This method also utilizes a bagging technique. Our experiments compare PCA-C with conventional single-target regressors (STR), such as gradient boosting (GB) and random forest (RF), and PLS2, a conventional MTR method, revealing improvements in accuracy across a variety of conditions, from small to large training data. These improvements were observed across all target variables. PCA-C based on GB often showed the highest accuracy. Compared to the conventional GB, the accuracy has been improved by 10 %–22 %. PCA-C based on RF has also improved accuracy by 7 %–112 % compared to conventional RF. Moreover, comparative experiments between PCA-C and its base model confirmed that the improvement in the method's accuracy resulted from the effects of the bagging and relationship elements.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105462"},"PeriodicalIF":3.7,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144313802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Use of measurement uncertainty in conformity assessment and process capability indices in quality evaluation of pharmaceutical manufacturing processes","authors":"Gynna Gómez Barrios ,&nbsp;Felipe Rebello Lourenço ,&nbsp;Elcio Cruz de Oliveira","doi":"10.1016/j.chemolab.2025.105470","DOIUrl":"10.1016/j.chemolab.2025.105470","url":null,"abstract":"<div><div>The pharmaceutical industry operates under strict regulatory standards to ensure high-quality, safe, and effective products. Several key indices and metrics are employed to ensure that the pharmaceutical industry operates according to these standards, with process capability indices (<span><math><mrow><mi>C</mi><mi>p</mi></mrow></math></span> and <span><math><mrow><mi>C</mi><mi>p</mi><mi>k</mi></mrow></math></span>) being the most widely utilized. Regular monitoring and analysis of this metric helps organizations maintain compliance, improve processes, and ultimately safeguard public health. Since the traditional conformity assessment approach (i.e., the simple acceptance rule) does not explicitly consider measurement uncertainty, this study proposes measurement capability indices, <span><math><mrow><mi>C</mi><mi>m</mi></mrow></math></span> and <span><math><mrow><mi>C</mi><mi>m</mi><mi>k</mi></mrow></math></span> analogous to the process capability indices <span><math><mrow><mi>C</mi><mi>p</mi></mrow></math></span> and <span><math><mrow><mi>C</mi><mi>p</mi><mi>k</mi></mrow></math></span>, but based on the measurement uncertainty. The <span><math><mrow><mi>C</mi><mi>m</mi><mi>k</mi></mrow></math></span> index was applied to evaluate the conformity of 750 mg paracetamol tablets. The results indicated that while 99.9 % of the reference drug product lots were compliant, only 68.02 % of the generic drug product lots met these specifications. Furthermore, Monte Carlo simulations estimated a consumer risk of 4.76 % for accepting non-compliant lots and a producer risk of 6.82 % for incorrectly rejecting compliant lots. By combining <span><math><mrow><mi>C</mi><mi>m</mi><mi>k</mi></mrow></math></span> and <span><math><mrow><mi>C</mi><mi>p</mi><mi>k</mi></mrow></math></span> strengthens quality control by enabling more conservative acceptance limits and providing a robust statistical basis for conformity decisions, thereby minimizing the risk of erroneous lot acceptance or rejection. This approach reduces risks for consumers and producers, enhances reliability in pharmaceutical production, and provides a precise tool to ensure product quality and safety.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105470"},"PeriodicalIF":3.7,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298844","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantification of metformin in pharmaceutical formulations using Vis-SWNIR hyperspectral imaging combined with multivariate curve resolution","authors":"Zahra Bolhassani ,&nbsp;Ali Aghaei ,&nbsp;Hadi Parastar","doi":"10.1016/j.chemolab.2025.105469","DOIUrl":"10.1016/j.chemolab.2025.105469","url":null,"abstract":"<div><div>Hyperspectral imaging (HSI), integrated with chemometrics, is a powerful tool in process analytical technology (PAT) for accurately determining the concentration of metformin hydrochloride, the active pharmaceutical ingredient (API), during the manufacturing of extended-release tablets. This study evaluated the effectiveness of visible-short wavelength near infrared (Vis-SWNIR) HSI (400–950 nm) combined with multivariate curve resolution-alternating least squares (MCR-ALS) for determining the API in a six-component mixture, which included excipients such as hydroxypropyl methylcellulose (HPMC), polyvinylpyrrolidone (PVP), and microcrystalline cellulose. Both linear and nonlinear calibration models were assessed. Partial least squares regression (PLSR) with the API concentration profile from MCR-ALS yielded promising calibration metrics, with a root mean square error of prediction (RMSEP) of 6.1 % w/w and a coefficient of determination (R²p) of 0.94, based on a set covering an API range of 0.0–70.6 % w/w. The method also demonstrated promising figures of merit (FOMs), including a limit of detection (LOD) of 4.7 % w/w and a limit of quantification (LOQ) of 14.2 % w/w, indicating its effectiveness in detecting API levels below standard thresholds. Further improvement was achieved using support vector machine (SVM) with radial basis function (RBF), enhancing RMSEP to 5.6 % w/w and R²p to 0.98, aiming to evaluate a non-linear method as a proof of concept. The study concluded that Vis-SWNIR HSI combined with chemometrics, provides an effective and non-destructive method for determining the correct API concentration in powder blends during blending, without the need for sample preparation.</div></div>","PeriodicalId":9774,"journal":{"name":"Chemometrics and Intelligent Laboratory Systems","volume":"264 ","pages":"Article 105469"},"PeriodicalIF":3.7,"publicationDate":"2025-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144289011","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}