High-efficacy and affordable hyperspectral pancreatic tissue image analysis using near-infrared spectroscopy

Sep 22, 2026

Pancreatic cancer remains one of the most lethal malignancies with less than 10% five-year survival rates, primarily due to late-stage diagnosis and limited early detection capabilities. Current diagnostic methods are expensive, time-intensive, and often inadequate for widespread screening applications. The study develops a cost-effective NIR hyperspectral imaging approach with advanced machine learning for automated pancreatic tissue classification. A pipeline using autoencoder-based spatial feature extraction, consensus outlier detection, and optimized neural network classifiers is evaluated on tissue microarrays and achieves 84% balanced accuracy with improved performance and interpretable attention mechanisms.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

Clinical challenge and survival rates

Need for rapid, cost-effective screening

Limitations of traditional histopathology

Promise of digital pathology and HSI

Journal of Pathology Informatics 21 (2026) 100651 Contents lists available at ScienceDirect Journal of Pathology Informatics journal [homepage:](homepage: www. elsevier. com/locate/jpi)[ www. elsevier. com/locate/jpi](homepage: www. elsevier. com/locate/jpi) High-efﬁcacy and affordable hyperspectral pancreatic tissue image analysis using near-infrared spectroscopy Zheng Tang a, Abhinav Mishra e, Benjamin Mora a, Bilal Al-Sarireh b, Olivia Irvine e, Brandon Mauri Victoria Higginbotham c, P.M. Anupama Bandaranayake e, S.H. Chandrashekhara d, Venkateswarlu Kanamarlapudi c, Debdulal Roy e,⁎ a Department of Computer Science and Mathematics, Swansea University, Swansea SA2 8PP, UK b Morriston Hospital, Heol Maes Eglwys, Morriston SA6 6NL, UK c Swansea Medical School, Swansea University, Swansea SA2 8PP, UK d All India Institute of Medical Sciences, Ansari Nagar, New Delhi 110029, Indiae Department of Chemistry, Swansea University, Swansea SA2 8PP, UK a , A R T I C L E I N F O Keywords: Machine learning Near-infrared Hyperspectral imaging Pancreatic cancer Automated tissue classiﬁcation Cancer diagnosis A B S T R A C T Pancreatic cancer remains one of the most lethal malignancies with less than 10% ﬁve-year survival rates, primarily due to late-stage diagnosis and limited early detection capabilities. Current diagnostic methods are expensive, timeintensive, and often inadequate for widespread screening applications. This study presents a novel, cost-effective approach using near-infrared (NIR) hyperspectral imaging combined with advanced machine learning for automated pancreatic tissue classiﬁcation. We have developed a comprehensive pipeline incorporating autoencoder-based spatial feature extraction, multi-method consensus outlier detection, and systematically optimized neural network classiﬁers to distinguish between cancerous and non-cancerous pancreatic tissue samples. Our methodology was evaluated on 78 tissue microarray samples, with rigorous quality control yielding a ﬁnal dataset of 69 high-quality specimens. The optimized classiﬁcation model achieved 84% balanced accuracy using leave-one-out cross-validation, representing a 10% point improvement over conventional FICA+SVM approaches (74.0%) and approaching the performance of expensive conventional histopathological methods. Key technical innovations include consensus-based outlier detection, systematic hyperparameter optimization revealing optimal single-layer architectures with ELU activation, and interpretable attention mechanisms for diagnostic decision support. The demonstrated cost-effectiveness of NIR instrumentation combined with robust classiﬁcation performance positions this approach as a promising pathway toward accessible, real-time pancreatic cancer screening tools that could signiﬁcantly impact early detection rates and patient outcomes in diverse clinical settings. Introduction Pancreatic cancer represents one of the most challenging oncological conditions in modern medicine. In 2022, there were over 510,000 new cases of pancreatic cancer globally and over 466,000 deaths worldwide, establishing it as one of the leading causes of cancer mortality globally.1 The disease is particularly devastating in the UK and USA, where it is one of the most common causes of cancer death, with a 5-year survival rate less than 10% and a 10-year survival rate less than 5%. With an average of 10,452 new cases diagnosed annually between 2016 and 2018, pancreatic cancer accounted for approximately 9600 deaths each year in the UK between 2017 and 2019.2 The stark contrast in survival rates—increasing to 20% for patients with localized disease but plummeting to merely 2% for patients with distant metastases3—underscores that the majority of patients are diagnosed at advanced stages when the cancer has grown extensively or metastasized, rendering it inoperable or resistant to treatment.4 This dire clinical reality emphasizes the critical need for rap

FAQs

What motivates the need for new pancreatic cancer screening methods?

Late-stage diagnosis and limited early detection capabilities drive poor survival rates, while existing diagnostic methods are costly, time-intensive, and insufficient for broad screening.

How does the proposed method classify pancreatic tissue?

It combines near-infrared hyperspectral imaging with a machine-learning pipeline that extracts spatial features via an autoencoder, detects outliers through consensus methods, and uses optimized neural network classifiers to distinguish cancerous vs non-cancerous tissue.

What performance was achieved and how was it validated?

The optimized model reached 84% balanced accuracy using leave-one-out cross-validation, showing improvement over conventional FICA+SVM approaches and approaching performance of expensive conventional histopathology.

Sep 22, 2026

Upload by Maya Linwood