Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis

Rahman, Md Hafizur; Hooper, Jayden K.; Wardeh, Alaa; Masilamani, Ashok Prabhu; Yockell-Lelièvre, Hélène; Ozhi Kandathil, Jayan; Khomami Abadi, Mojtaba

doi:10.3390/s25226839

This is an early access version, the complete PDF, HTML, and XML versions will be available soon.

Open AccessArticle

Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis

by

Md Hafizur Rahman

^*,

Jayden K. Hooper

,

Alaa Wardeh

,

Ashok Prabhu Masilamani

,

Hélène Yockell-Lelièvre

,

Jayan Ozhi Kandathil

and

Mojtaba Khomami Abadi

Noze, 4920 Pl. Olivia, Montreal, QC H4R 2Z8, Canada

^*

Author to whom correspondence should be addressed.

Sensors 2025, 25(22), 6839; https://doi.org/10.3390/s25226839 (registering DOI)

Submission received: 9 September 2025 / Revised: 31 October 2025 / Accepted: 4 November 2025 / Published: 8 November 2025

(This article belongs to the Special Issue Advances in Sensorized AI-Driven Intelligent Systems in Healthcare and Beyond)

Download Versions Notes

Abstract

Confounding factors in olfactory aroma data, such as high humidity levels, substantially affect sensor outputs, masking subtle volatile organic compound (VOC) patterns and hindering generalizable machine learning models. Traditional representation learning methods often require large datasets to mitigate confounder-induced variance, a resource unavailable in specialized sensor applications with limited data. This study presents Confounder-Invariant Representation Learning (CIRL), a method designed to mitigate confounding influences in data-scarce settings by leveraging explicit confounder information, such as relative humidity. CIRL enhances learned representations by reducing confounder effects, improving data purity and model robustness. Applied to three breath aroma datasets—acetone, ketosis, and peppermint-oil breath, all affected by high humidity—CIRL was integrated with standard autoencoder models. Evaluated within the same framework, CIRL improved generalization performance by 10–15% in classification accuracy across all three datasets. These results demonstrate CIRL’s potential to advance reliable artificial olfaction for applications like breath-based diagnostics in challenging real-world conditions.

Keywords: aroma sensors; aroma data; confounder invariant learning; representation learning; scarce data; relative humidity; deep learning; autoencoders; generalizability

Share and Cite

MDPI and ACS Style

Rahman, M.H.; Hooper, J.K.; Wardeh, A.; Masilamani, A.P.; Yockell-Lelièvre, H.; Ozhi Kandathil, J.; Khomami Abadi, M. Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis. Sensors 2025, 25, 6839. https://doi.org/10.3390/s25226839

AMA Style

Rahman MH, Hooper JK, Wardeh A, Masilamani AP, Yockell-Lelièvre H, Ozhi Kandathil J, Khomami Abadi M. Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis. Sensors. 2025; 25(22):6839. https://doi.org/10.3390/s25226839

Chicago/Turabian Style

Rahman, Md Hafizur, Jayden K. Hooper, Alaa Wardeh, Ashok Prabhu Masilamani, Hélène Yockell-Lelièvre, Jayan Ozhi Kandathil, and Mojtaba Khomami Abadi. 2025. "Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis" Sensors 25, no. 22: 6839. https://doi.org/10.3390/s25226839

APA Style

Rahman, M. H., Hooper, J. K., Wardeh, A., Masilamani, A. P., Yockell-Lelièvre, H., Ozhi Kandathil, J., & Khomami Abadi, M. (2025). Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis. Sensors, 25(22), 6839. https://doi.org/10.3390/s25226839

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Confounder-Invariant Representation Learning (CIRL) for Robust Olfaction with Scarce Aroma Sensor Data: Mitigating Humidity Effects in Breath Analysis

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI