Adaptive Machine Learning for Electronic Nose‐Based Forensic VOC Classification
Advanced Science · 2025-09 · 12 (36)
DOI: 10.1002/advs.202504657 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advanced Science, online erschienen am 24.06.2025. Verfasst von Ivan Shtepliuk und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Physics, Chemistry and Biology, Linköping University, Linköping, 581 83, Sweden. Hauptthemen laut MEDLINE: Electronic Nose, Machine Learning, Volatile Organic Compounds, Odorants. Die Arbeit wurde bisher 21-mal zitiert (Crossref) und nennt 70 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Vetenskapsrådet, Internal Security Fund - Police, Faculty of Science and Engineering (LiTH), Linköping University,.
Abstract (Originalsprache)
Odor detection as forensic evidence is an emerging and transformative approach for identifying individuals in cases where conventional techniques— such as DNA profiling, fingerprints, or gas chromatography‐mass spectroscopy— are insufficient. Key challenges include distinguishing between living and deceased individuals, differentiating human from animal remains, and estimating postmortem intervals. While specially trained dogs remain the gold standard in some contexts, there is growing demand for faster, scalable, and cost‐effective alternatives. Here, a bio‐inspired electronic nose comprising 32 metal‐oxide sensors, integrated with machine learning, is presented as a non‐invasive, real‐time, and reliable tool for volatile organic compound profiling. The system accurately classifies postmortem versus antemortem human biosamples (98.1%), discriminates human from animal tissue (97.2%), and estimates postmortem intervals with high temporal resolution. This study introduces a robust AI‐driven olfactory platform for forensic scent detection, highlighting its potential to complement or replace traditional methods. By coupling volatilome analysis with sensor miniaturization and algorithmic refinement, this approach lays the groundwork for next‐generation forensic diagnostics and odor‐based biomarker discovery.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Department of Physics, Chemistry and Biology, Linköping University, Linköping, 581 83, Sweden.
- Department of Forensic Genetics and Forensic Toxicology, National Board of Forensic Medicine, Linköping, 587 58, Sweden.
ORCID-Profile: Donatella Puglisi
Förderung
- Vetenskapsrådet
- Internal Security Fund - Police
- Faculty of Science and Engineering (LiTH), Linköping University,
- Strategic Research Area in Forensic Sciences (SOFO)
- Livsmedelsverket (Swedish Food Agency)
- Vinnova (Sweden's Innovation Agency)
Zitieren
APA
Shtepliuk, I., Montelius, K., Eriksson, J., & Puglisi, D. (2025). Adaptive Machine Learning for Electronic Nose‐Based Forensic VOC Classification. Advanced Science, 12(36). https://doi.org/10.1002/advs.202504657
Vancouver
Shtepliuk I, Montelius K, Eriksson J, Puglisi D. Adaptive Machine Learning for Electronic Nose‐Based Forensic VOC Classification. Advanced Science. 2025;12(36). doi:10.1002/advs.202504657
BibTeX
@article{Shtepliuk2025,
author = {Shtepliuk, Ivan and Montelius, Kerstin and Eriksson, Jens and Puglisi, Donatella},
title = {Adaptive Machine Learning for Electronic Nose‐Based Forensic VOC Classification},
journal = {Advanced Science},
year = {2025},
volume = {12},
number = {36},
doi = {10.1002/advs.202504657}
}
EUVIMED