Electronic nose and machine learning for modern meat inspection

Abstract Objective and reliable post-mortem meat inspection is a key factor in ensuring adequate assessment and quality control of meat intended for human consumption. Early identification of issues that may impact public health and animal health and welfare, such as the presence of chemical contami...

Full description

Saved in:
Bibliographic Details
Main Authors: Ivan Shtepliuk, Guillem Domènech-Gil, Viktor Almqvist, Arja Helena Kautto, Ivar Vågsholm, Sofia Boqvist, Jens Eriksson, Donatella Puglisi
Format: Article
Language:English
Published: SpringerOpen 2025-04-01
Series:Journal of Big Data
Subjects:
Online Access:https://doi.org/10.1186/s40537-025-01151-4
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849699102872305664
author Ivan Shtepliuk
Guillem Domènech-Gil
Viktor Almqvist
Arja Helena Kautto
Ivar Vågsholm
Sofia Boqvist
Jens Eriksson
Donatella Puglisi
author_facet Ivan Shtepliuk
Guillem Domènech-Gil
Viktor Almqvist
Arja Helena Kautto
Ivar Vågsholm
Sofia Boqvist
Jens Eriksson
Donatella Puglisi
author_sort Ivan Shtepliuk
collection DOAJ
description Abstract Objective and reliable post-mortem meat inspection is a key factor in ensuring adequate assessment and quality control of meat intended for human consumption. Early identification of issues that may impact public health and animal health and welfare, such as the presence of chemical contaminants in meat, is critical. In this study, we propose a novel method to modernize meat inspection using an electronic nose combined with machine learning (ML), with focus on pig meat as a case study. We explored its potential as a complementary tool to traditional sensory evaluation and analytical methods, aiming to enhance the efficiency and effectiveness of current inspections. We employed a metal-oxide based gas sensor array of commercially available chemoresistive sensors, functioning as an electronic nose, to differentiate between various categories of 100 pig meat samples collected at a slaughterhouse based on their odor characteristics, including a urine-like smell and post-mortem aging. Using the Optimizable Ensemble model, we achieved a sensitivity of 96.5% and specificity of 95.3% in categorizing fresh and urine-contaminated meat samples. The model demonstrated robust predictive performance with a Kappa value of approximately 0.926, indicating near-perfect agreement between the predictions and actual classifications. Furthermore, our developed ML model demonstrated the ability to distinguish between nominally fresh pig meat and meat aged for one to two additional days with an accuracy of 93.5% and can also correctly identify meat aged 3–31 days or 17–31 days. Based on the consensus of preliminary decisions from each individual sensor element, the algorithm effectively determined the final status of the meat. This research lays the groundwork for practical applications within the meat inspection process in slaughterhouses and as quality assurance throughout the meat supply chain. As we continue to refine and validate this method, its potential for real-world implementation becomes increasingly evident.
format Article
id doaj-art-27cbc0e027f14f2aaa379d444c15ec5e
institution DOAJ
issn 2196-1115
language English
publishDate 2025-04-01
publisher SpringerOpen
record_format Article
series Journal of Big Data
spelling doaj-art-27cbc0e027f14f2aaa379d444c15ec5e2025-08-20T03:18:42ZengSpringerOpenJournal of Big Data2196-11152025-04-0112112110.1186/s40537-025-01151-4Electronic nose and machine learning for modern meat inspectionIvan Shtepliuk0Guillem Domènech-Gil1Viktor Almqvist2Arja Helena Kautto3Ivar Vågsholm4Sofia Boqvist5Jens Eriksson6Donatella Puglisi7Department of Physics, Chemistry and Biology, Linköping UniversityDepartment of Physics, Chemistry and Biology, Linköping UniversityDepartment of Animal Biosciences, Swedish University of Agricultural SciencesDepartment of Animal Biosciences, Swedish University of Agricultural SciencesDepartment of Animal Biosciences, Swedish University of Agricultural SciencesDepartment of Animal Biosciences, Swedish University of Agricultural SciencesDepartment of Physics, Chemistry and Biology, Linköping UniversityDepartment of Physics, Chemistry and Biology, Linköping UniversityAbstract Objective and reliable post-mortem meat inspection is a key factor in ensuring adequate assessment and quality control of meat intended for human consumption. Early identification of issues that may impact public health and animal health and welfare, such as the presence of chemical contaminants in meat, is critical. In this study, we propose a novel method to modernize meat inspection using an electronic nose combined with machine learning (ML), with focus on pig meat as a case study. We explored its potential as a complementary tool to traditional sensory evaluation and analytical methods, aiming to enhance the efficiency and effectiveness of current inspections. We employed a metal-oxide based gas sensor array of commercially available chemoresistive sensors, functioning as an electronic nose, to differentiate between various categories of 100 pig meat samples collected at a slaughterhouse based on their odor characteristics, including a urine-like smell and post-mortem aging. Using the Optimizable Ensemble model, we achieved a sensitivity of 96.5% and specificity of 95.3% in categorizing fresh and urine-contaminated meat samples. The model demonstrated robust predictive performance with a Kappa value of approximately 0.926, indicating near-perfect agreement between the predictions and actual classifications. Furthermore, our developed ML model demonstrated the ability to distinguish between nominally fresh pig meat and meat aged for one to two additional days with an accuracy of 93.5% and can also correctly identify meat aged 3–31 days or 17–31 days. Based on the consensus of preliminary decisions from each individual sensor element, the algorithm effectively determined the final status of the meat. This research lays the groundwork for practical applications within the meat inspection process in slaughterhouses and as quality assurance throughout the meat supply chain. As we continue to refine and validate this method, its potential for real-world implementation becomes increasingly evident.https://doi.org/10.1186/s40537-025-01151-4Gas sensorsMachine learningVolatile organic compoundsOdor detectionMeat chain wasteMeat quality assurance
spellingShingle Ivan Shtepliuk
Guillem Domènech-Gil
Viktor Almqvist
Arja Helena Kautto
Ivar Vågsholm
Sofia Boqvist
Jens Eriksson
Donatella Puglisi
Electronic nose and machine learning for modern meat inspection
Journal of Big Data
Gas sensors
Machine learning
Volatile organic compounds
Odor detection
Meat chain waste
Meat quality assurance
title Electronic nose and machine learning for modern meat inspection
title_full Electronic nose and machine learning for modern meat inspection
title_fullStr Electronic nose and machine learning for modern meat inspection
title_full_unstemmed Electronic nose and machine learning for modern meat inspection
title_short Electronic nose and machine learning for modern meat inspection
title_sort electronic nose and machine learning for modern meat inspection
topic Gas sensors
Machine learning
Volatile organic compounds
Odor detection
Meat chain waste
Meat quality assurance
url https://doi.org/10.1186/s40537-025-01151-4
work_keys_str_mv AT ivanshtepliuk electronicnoseandmachinelearningformodernmeatinspection
AT guillemdomenechgil electronicnoseandmachinelearningformodernmeatinspection
AT viktoralmqvist electronicnoseandmachinelearningformodernmeatinspection
AT arjahelenakautto electronicnoseandmachinelearningformodernmeatinspection
AT ivarvagsholm electronicnoseandmachinelearningformodernmeatinspection
AT sofiaboqvist electronicnoseandmachinelearningformodernmeatinspection
AT jenseriksson electronicnoseandmachinelearningformodernmeatinspection
AT donatellapuglisi electronicnoseandmachinelearningformodernmeatinspection