TITLE: Practical Sampling, NIR Spectroscopy and data analysis for PAT applications

   BIO Phil Doherty

Phil Doherty is CEO of Acuitis Process Analytical Technologies and has over 30 years’ experience in the development and implementation of sampling systems used for interfacing PAT instruments into pharmaceutical processes. Phil was part of the Pfizer global PAT group in Ireland and held positions at Expo Process Analytics including Managing Director and Chief Scientific Officer.
Recently, Phil started Acuitis Process Analytical Technologies where he provides expert consulting, implementation and strategy development services related sampling, PAT and data analysis into regulated industries. Phil holds a bachelor of science with honours in chemistry and physiology and has a postgraduate diploma in business leadership and organisation.

    BIO Steve Hammond

Steve Hammond is owner of Steve Hammond Consulting and has over 40 years’ experience in the deployment of PAT in pharmaceutical manufacturing, including implementation of representative sampling for PAT systems in pharmaceutical manufacturing equipment.  Steve was the director of the Process Analytical Sciences Group (PASG) at Pfizer for 30 years and has held roles with Expo Process Analytics amongst other roles on industry advisory boards.
Steve now works as a consultant to industry helping organisations to optimise their approaches to sampling and PAT implementation in oral solid dose, biopharmaceutical and additive manufacturing operations.

   BIO Brad Swarbrick

Brad Swarbrick is Managing Director and Cofounder of KAX Group, a leading provider of multivariate data analysis and model deployment platforms to industry, research and academia. Brad was part of the Pfizer global PAT group and has held many positions in industry and academia, including the Chief Operating Officer position for a worldwide software provider.
Brad has over 25 years’ experience in the implementation of PAT into pharmaceutical and other industries and has expertise in the areas of chemometrics and vibrational spectroscopy. Brad has a PhD in Biospectroscopy from the University of Sydney and in the current pharmaceutical section editor of the Journal of Near Infrared Spectroscopy.

TITLE: #NIR4 – How to Analyse, Structure & Handle NIR Results in R Resiliently

BIO: Zoltan Kovacs is a full professor at the Institute of Food Science and Technology, Hungarian University of Agriculture and Life Sciences (MATE). His research focuses on artificial taste and odor sensing systems, as well as near-infrared spectroscopy (NIRS), where he utilizes both conventional and novel aquaphotomics data evaluation techniques. He has investigated the application of NIRS and aquaphotomics across various systems, including food matrices, biological organisms, and different types of water samples and developed measurement and data evaluation protocols for rapid, non-invasive analyses. Professor Kovacs is an active member of several scientific committees, including the Agro- and Biosystems Engineering Research Committee of the Hungarian Academy of Sciences, and serves as the European Liaison for the International Aquaphotomics Society.

ABSTRACT: Achieving reliable and adequate results from your NIR analysis efficiently begins long before the evaluation of the spectra. In addition to being an expert in various chemometric techniques, you must also know how to design experiments, structure, archive, and curate your NIR data to be effective in NIR applications. In this precourse, the major steps of NIR analysis will be covered with hands-on practical examples – from experimental design to saving and structuring the collected data as part of a larger NIR database – ensuring your data is instantly ready for effective data evaluation using the powerful R project and the aquaP2 package (Pollner & Kovacs, 2016). If you are interested in gaining practical skills in these areas, please ensure your computer is ready with R and aquaP2 installed, and join us for this exciting journey.

TITLE: Determining Detection Limits: Key Instrument Parameters and Spectral Processing

BIO: Tetsuya Inagaki is an Associate Professor at the Graduate School of Bioagricultural Sciences, Nagoya University (since 2021). Previously, he served as a Lecturer (2017–2021) and Assistant Professor (2011–2017) at the same institution. His research focuses on forest product science, with expertise in chemometrics and machine learning. He has contributed to the advancement of analytical methods for agricultural and forestry products.

ABSTRACT: Near-infrared spectroscopy (NIRS) enables non-destructive, non-contact measurement of samples due to its inherently low molar absorptivity. However, the interpretation of detection limits in NIRS becomes complex as chemometric techniques are typically employed for analysis. Despite this complexity, as a spectroscopic method, the detection limit should fundamentally be determined by the molar absorptivity and noise contributions, including detector noise, photon noise, and modulation noise. This course will explore the concept of detection limits from this perspective. Additionally, understanding detection limits requires consideration of spectrometer configurations (e.g., Fourier transform vs. dispersive), spectral resolution, and wavelength interval. These aspects will be discussed in detail.

TITLE: Variables (features) selection in multivariate calibration and classification models using NIR spectroscopy

BIO: Dr Jean-Michel Roger is a rural engineer specializing in chemometrics, as part of the development of sensors based on near-infrared spectrometry. He has focused his research on the robustness of calibrations, studying methods based on orthogonal projections, variable selection and, more recently, multiblock and multiway methods. The main applications of his research are fruit maturity measurement, disease detection and waste sorting. He has studied deep learning methods, supervising Master’s-level trainees, and helped set up a LabCom project (French Agence Nationale de la Recherche) to develop sensors using computational optics in conjunction with deep learning algorithms.

ABSTRACT: This course will introduce participants to variable selection strategies in Near-Infrared (NIR) spectrometry, focusing on both the practical and interpretative motivations for reducing the number of spectral variables in multivariate calibration models. The session will begin with a discussion of the challenges inherent to NIR data — characterized by high dimensionality and strong collinearity — and will outline why variable selection can be useful.
Two major categories of variable selection objectives will be distinguished. The first concerns the optimization of calibration models when the number of available samples is limited, aiming to improve model robustness, reduce overfitting, and simplify final predictive tools. The second focuses on the identification of a small, relevant subset of variables to better understand the underlying physico-chemical or biological phenomena.
The first category will be extensively covered, presenting three main families of approaches:

• Filter methods, which select variables based on their individual statistical relationship with the target variable.
• Embedded methods, where variable selection occurs during model construction (e.g., variable importance measures in PLS).
• Wrapper methods, which iteratively assess variable subsets through predictive model performance.

The second category, less often addressed in the literature, will be illustrated through the presentation of the xx-CovSel family of methods, designed to identify a minimal set of variables explaining the covariance structure between spectral data and reference measurements. These methods are particularly well suited for exploratory analysis and the identification of spectral regions associated with specific chemical or structural characteristics.
By the end of the course, participants will have a clear understanding of the main strategies available for variable selection in NIR spectrometry, their appropriate applications, and the trade-offs involved.

Opening Lecture – Monica Casale, Ph.D.
President, Italian Society of NIR Spectroscopy (SISNIR)

BIO: Monica Casale, Ph.D., is Associate Professor of Analytical Chemistry at the University of Genoa, Department of Pharmacy, and holds the Italian National Scientific Qualification for Full Professorship. Her research focuses on chemometrics and the development of spectroscopic methods, with particular emphasis on Near Infrared (NIR) Spectroscopy applied to food authentication and quality assessment. She has authored numerous peer-reviewed publications, book chapters, and has participated in national and European research projects. Dr. Casale is actively involved in academic teaching and Ph.D. supervision.
Since 2016, she is President of the Italian Society of NIR Spectroscopy (SISNIR).

In her opening address, Prof. Monica Casale welcomes attendees to the conference on behalf of the Italian Society of NIR Spectroscopy (SISNIR), emphasizing the importance of collaboration, knowledge exchange, and interdisciplinary dialogue in the field of Near Infrared Spectroscopy.
She expresses her deepest gratitude to the members of the organizing committee, whose exceptional professionalism and meticulous planning have been instrumental to the successful realization of this event. Prof. Casale also warmly thanks SISNIR and its members for their continued commitment and active support of the Society’s initiatives, which make gatherings like this possible. Further appreciation is extended to the sponsors for their generous support, to the invited speakers for their invaluable insights, and to all presenters, whose contributions enrich the scientific program and foster meaningful dialogue throughout the conference.
In closing, with sincere appreciation to all contributors and a special acknowledgment of the beauty of Rome as the host city, Prof. Casale invites participants to engage fully in both the scientific and social dimensions of the conference.

The pillars of Near-Infrared applications

BIO: Associate Professor at the Vet School of the University of Padua, trained as dairy nutritionist (M.S. at Virginia Tesch), he has dedicated his research in the development and improvement of forage NIR analysis. He was a Fulbright scholar at Penn State/Infrasoft International with Dr. Shenk heavily involved in developing and testing WinISI 1.0 and the Local algorithm for calibration development. He is currently inventor or co-inventor of 6 patents all involved NIR from application at the farm to NIR database management. Started a university spin-off company (www.grainit.it) developing portable NIR applications for on farm feed and forage analysis.

ABSTRACT: A NIR application is a combination of multidisciplinary fields that involve knowledge on many different scientific areas. Any application stands on three fundamental “pillars”: 1)Instrument/Sensor, 2) predicting model(s), 3)Sample preparation/presentation. Scientists and vendors are focused on the first two pillars, while sampling and sample presentation are often neglected and may be considered of lower scientific value. As scientist, we are committed to advance the performance of this technology, yet we have to make sure that transfer of knowledge to the industry also includes sample presentation and preparation. The presentation highlights some aspect of instrument trends and the importance of sample presentation.

The history and evolution of NIR spectroscopy in Italy

BIO: Tiziana, qualified (1983) in Food Technology at the University of Milan, Italy, has been employed by the Council for Agricultural Research and Economics (CREA), since 1987 at the Research Centre in Lodi (CREA.FLC). She began research into the application of near infrared spectroscopy on dairy field 30 years ago (1996) in cooperation with instrumentation suppliers and cheese-making firms. She became Research Manager in 2003 and continues to improve the development of applied research in dairy and “fruit and vegetable” fields. Since 2010 she moved to CREA.IT in Milan. She had her PhD in 2023 at the Tuscia University, Engineering for Energy and Environment Dept. The scientific activity is supported by about 350 research papers, lectures, and posters.
Since 2010 she is member of the Aquaphotomics Community; she was the President of the Italian Society for Near InfraRed Spectroscopy from 2010 to 2016; and she is currently member of ICNIRS (International Committee of Near InfraRed Spectroscopy) Advisory Committee for the five-year period 2021-2025.

ABSTRACT: This presentation tells the story, the evolution and the growth of the NIR community in Italy. Starting from the first pioneers in the last century up to the current situation in terms of researchers, technicians, stakeholders, collaborators and disseminators.
The actions made for information; several schools dedicated to NIR technology and data processing; the involvement of instrumentation manufacturers; the cooperation of universities and research centres; the supervision of the Italian Society for NIR Spectroscopy have proved to be valid tools for a continuous growth and above all for the incessancy and increasingly widespread use of NIRS at both research and industrial level in our Country.

TITLE: NIR spectroscopy on milk: from compositional analysis with benchtop instruments to physical properties with spatially resolved techniques

BIO: Ben Aernouts obtained his PhD degree in Bioscience Engineering in 2014, under supervision of prof. Wouter Saeys at the KU Leuven in Belgium. In this period, Dr. Aernouts studied the optical behaviour of raw milk in relation to the quality properties. As a postdoctoral fellow, he used the obtained insights to improve the design of optical sensors for milk quality and cow health monitoring, and implemented and validated these technologies on the farm. Since October 2016, Ben Aernouts is a professor in “Management in Livestock Production” at the Faculty of Engineering Technology of KU Leuven Campus Geel. He teaches courses on livestock production, cattle management and precision livestock farming, and he leads a group of 15 researchers developing innovative sensor technology and data-processing algorithms to support livestock management. 4 members of his team are developing NIR-technologies to measure different aspects of milk quality, including chemical composition as well as physical structure, both under lab conditions as well as on the farm.

ABSTRACT: NIR spectroscopy on milk: from compositional analysis with benchtop instruments to physical properties with spatially resolved techniques

As milk contains valuable information on the cow’s metabolic status, regular analysis of the produced milk is an efficient way to monitor cow and udder health. Near-infrared (NIR) spectroscopy is a rapid, non-destructive and efficient technique that has proven valuable for on-line analysis of the raw milk composition. In the past 15 years, several prototypes for in-line milk analysis have been developed and validated, some of which are commercially available nowadays. Still, none of these sensors got ICAR certified so far, generally indicating that the accuracy, repeatability and reproducibility are insufficient. Based on our own 15-year experience, I will present and discuss the main challenges for in-line and on-farm milk analysers by reflecting results and sensor performances reported in the past against those obtained with our own research-prototypes for in-line milk analysis. These prototypes have been extensively tested at different dairy farms and under varying conditions over the past 8 years. Although the initial (calibration) accuracy of the prediction of milk fat, protein and lactose is largely depending on the signal-to-noise ratio of the sensor itself, the predictions afterwards are mainly subject to bias drift. Part of this comes from variation in milk temperature, which follows a seasonal pattern and can be accounted for by robust modelling. Additionally, unsupervised techniques and using bulk milk analyses results to correct for the bias can help and bring the sensor performance within the required specifications. Apart from predicting the milk composition from the NIR spectra, also the milk physical properties can be derived from the scattering present in the NIR spectra. To this end, my team develops and tests different spatially resolved NIR spectroscopy configurations to simultaneously measure the milk composition and fat globule size distribution in-line.

TITLE: Unlocking the Potential of Spectral Imaging Through Deep Learning

BIO: Dr. Junli Xu is an Assistant Professor at University College Dublin, Ireland, where she leads a dynamic research team comprising four PhD candidates, two postdoctoral researchers, and one research assistant. She has successfully secured prestigious funding awards totalling approximately €2.7 million as a PI, including the prestigious ERC Starting Grant and the SFI-IRC Pathway Programme. Dr. Xu’s multidisciplinary research focuses on leveraging spectral imaging and advanced data analytics, such as machine learning and deep learning, to address complex challenges across diverse fields, including food, environment, geology, and human health. She serves on several editorial boards, including as an Associate Editor for Frontiers in Nutrition and an Honorary Associate Editor for the International Journal of Food Science & Technology. Dr. Xu has authored four book chapters and 58 peer-reviewed articles. She was recognized among the Top 2% of Most-Cited Scientists in 2024, reflecting the global impact of her research contributions.

ABSTRACT: This talk aims to showcase several case studies demonstrating how deep learning can enhance spectral imaging analysis. In the first case study, we present an innovative integration of deep learning and chemometrics for processing spectral images of fruits. The image processing tasks, such as object detection and recognition, are handled using deep learning, while the prediction of chemical properties is approached through chemometric modelling based on latent space analysis. The second case study focuses on leveraging deep learning to automate hyperspectral core scanning for geological applications. This approach integrates hyperspectral imaging with deep learning by introducing an automated analysis framework based on YOLOv8 (You Only Look Once), Segment Anything Model (SAM2), and Multi-Layer Perceptron (MLP). The framework is designed to identify and localize wooden boxes used for storing drill cores during hyperspectral image scanning, streamlining the analysis process. The third case study explores the application of deep learning for particle analysis in chemical imaging, specifically for microplastic detection. Deep learning models are employed to localize particles, automatically compute their morphological properties, and simultaneously identify polymer types, enhancing the accuracy and efficiency of microplastic analysis. In summary, spectral imaging is inherently data-rich, and its combination with deep learning significantly enhances its potential, making it a transformative tool across various applications.

TITLE: When imperfection is an advantage

BIO: Oxana Rodionova is a principal researcher at the Semenov Federal Research Center for Chemical Physics, Russian Academy of Sciences in Moscow. She is a founding member of the Russian Chemometrics Society, and a member of the Chemometric study group of EUCheMS. Her research activities are focused on various aspects of chemometrics, development of novel methods (SIC method, DD SIMCA) and software tools including DD-SIMCA and PLS-DA tool boxes, and free web-applications (Web-SIMCA). Special activities are directed on application of existing methods to the real world problems such as counterfeit drug detection, field-deployment methods for food authentication, application of multivariate methods in personalized medicine.

ABSTRACT: In mechanics, there is an interesting phenomenon called induced stability [1]. In short, it is a way of making an unstable system (such as an inverted pendulum) stable. To achieve this, small vibrations (actually noise) are added to the system. This little-known fact, however, is widely used in practice – for example, thanks to it we can ride a bicycle.  In this talk, we will discuss how to use this principle for the practical implementation of the NIR spectroscopy.
There are many papers showing that NIR spectroscopy in combination with chemometrics is very promising for solving various practical problems. However, when it comes to the practical application for everyday use on multiple devices in different locations, numerous obstacles arise that hinder its implementation. It is generally accepted that various fluctuations are a source of disruption of stability and resilience, and therefore they must be somehow eliminated in order to make the system more perfect. Calibration transfer is a good example. On the contrary, based on the principle of induced stability, it is necessary to introduce into the system some controlled noise, the magnitude of which varies depending on the chosen goal.
The examples are taken from a real-life large-scale screening system, which is designed to detect counterfeit and substandard tablets and capsules using non-invasive NIR and chemometrics. In this framework, several levels of generalization can be considered depending on the ultimate goal. The most general task is to distinguish some specific remedy, say Amlodipine, from any other remedy that is not Amlodipine. In this case, we need to collect samples from as many Amlodipine manufactures as possible, ensuring the widest variability. At the second level, when we narrow the task and seek to distinguish Amlodipine from a particular manufacture from other manufacturers of the same remedy, we should collect samples from different batches of this manufacturer. At the third level, if the goal is to distinguish tablets from one particular batch from tablets from tablets from other batches, the training set should consists only of tablets from that specific batch.
It is obvious that at different levels, it is necessary to adjust the degree of variability of objects in the training sets. Less obviously, the NIR instrument and chemometric method should also be selected according to the level, as excessive accuracy may lead to less than ideal results.
Thus, in order to induce a stable, fit-for-purpose NIR system, it is necessary to adjust its degree of imperfection accordingly.

Reference
[1] Stephenson, A. (1908). XX. On induced stability    . The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 15(86), 233–236. https://doi.org/10.1080/14786440809463763

TITLE: Evaluation of Agricultural Products Using HSI + ‘X’

BIO: Tetsuya Inagaki is an Associate Professor at the Graduate School of Bioagricultural Sciences, Nagoya University (since 2021). Previously, he served as a Lecturer (2017–2021) and Assistant Professor (2011–2017) at the same institution. His research focuses on forest product science, with expertise in chemometrics and machine learning. He has contributed to the advancement of analytical methods for agricultural and forestry products.

ABSTRACT: NIR hyperspectral imaging (NIR-HSI) allows for the visualization of spatial distributions of chemical components in agricultural products. However, mere visualization does not fully utilize the potential of HSI. By integrating HSI with various analytical approaches, its effectiveness can be significantly enhanced. This presentation will introduce advanced methodologies that extend HSI applications, including reaction kinetics analysis with HSI, the combination of HSI with X-ray CT and finite element analysis, and the integration of HSI with convolutional neural networks (CNNs). These synergistic approaches will be demonstrated to highlight their potential in agricultural product evaluation.

The pride and the passion by nirs”: 36 years of dedication to learning and teaching in near infrared spectroscopy

BIO: Prof. Dr. Ana Garrido-Varo received her PhD in Agricultural Engineering from the University of Córdoba (UCO), Spain, in 1987. Since 1980, she has held various teaching and academic positions at the Faculty of Agricultural and Forestry Engineering (ETSIAM), where she served until her retirement in 2023. She was appointed Honorary Professor at ETSIAM for the period 2023–2025. An active member of the ICNIRS community since 1989. She served as Convenor of NIR2003 and since then, she held several key positions within the ICNIRS Executive, Advisory, and Educational Committees before becoming President of ICNIRS (2013–2017). Her contributions have been recognized with prestigious honors, including the Tomas Hirschfeld Award (2005) and the Medal of the Spanish Federation of Compound Feed Manufacturers, acknowledging her pivotal role in promoting and implementing NIRS technology in the feed industry. She supervised over 45 BSc and MSc theses and 16 PhD dissertations. Her research output includes 126 peer-reviewed NIRS-related publications indexed in SCI, five book chapters, and more than 200 non-indexed contributions in national and international publications.

ABSTRACT: Faculty of Agricultural and Forestry Engineering, University of Córdoba. ETSIAM, Campus Rabanales, 14071 Córdoba, Spain. pa1gavaa@uco.es
This film, directed by the awardee, summarizes 36 years of dedication to learning, teaching, research, and technology transfer in Near Infrared Spectroscopy (NIRS). One of its main objectives is to share valuable insights and experiences from her R&D Group that are not typically published in indexed scientific journals but have had a significant impact on the advancement and funding of NIRS-related activities. The film aims to support industry professionals and laboratory practitioners in the challenging task of implementing NIRS successfully in real-world applications. Another key goal is to inspire young NIRS scientists and professionals to engage with that unique optical sensor- in combination with Machine Learning methods and Artificial Intelligence approaches- as pivotal in broader digital strategies to strengthen adaptive and resilient global food systems.

Keywords: NIR sensors, education, research, technology transfer, Machine Learning Artificial Intelligence, adaptive and resilient global food systems.

Unmasking the Phantom: From Mystery to Meaning

Near-Infrared (NIR) spectroscopy was introduced to the wheat industry in South Africa in the late 1970s to early 1980s, following global trends in grain quality analysis. The Technicon InfraAlyzer was employed for rapid moisture and protein testing and particularly for quality grading and wheat cultivar evaluation. In the mid-1980’s, as a first time NIR instrument user I was confronted by the ‘Phantom in the Lab’ a mysterious ‘black box’ where data went in and predictions came out. Soon, research began to unmask this Phantom and with the advent of chemometrics, the black box began to clear – first just a faint outline, then growing in transparency until it stood as a glass box. Understanding the ‘why’ behind the predictions – equally as important as the predictions themselves – became possible. Join me on a journey along the path of light and learning from those early days to where we stand today: with transparent, interpretable, integrated NIR science.

BIO: Prof Marena Manley of Stellenbosch University has significantly contributed to the advancement of NIR spectroscopy and hyperspectral imaging in South Africa and beyond, particularly in the characterisation of cereal grains. She supervised over 75 postgraduate students and published 135 peer-reviewed papers. She played a key role in establishing Africa’s only Vibrational Spectroscopy Unit with hyperspectral imaging capabilities, positioning SU as a leading NIR spectroscopy research hub on the continent. In recognition of her excellence in education and teaching, she received the SU’s Chancellor’s Award in 2022. Marena chaired the 15th ICNIRS Conference in Cape Town and will chair the 22nd International Diffuse Reflectance Conference (IDRC) in Knoxville, USA in 2026.