Evaluation of Three Electronic Noses for Detecting Incipient Wood Decay

Manuela Baietto, Alphus D. Wilson, Daniele Bassi, Francesco Ferrini2010 · DOI

TL;DR

This study evaluates three electronic nose (e-nose) technologies for their ability to detect incipient wood decay by analyzing volatile organic compounds (VOCs). All three e-noses successfully discriminated between healthy and decayed wood, demonstrating the potential of this technology for non-invasive, early detection of wood decay in urban trees.

What Problem Does This Paper Address?

Assessing the structural stability of urban trees is crucial for public safety, but current methods often rely on visual analysis or invasive, expensive instruments that are not always suitable for urban environments. Detecting wood decay, especially in its early (incipient) stages, is particularly challenging, as it can significantly weaken trees without obvious external signs. This paper addresses the need for a non-invasive, portable, and effective method for early detection of wood decay by exploring the use of electronic noses, which can identify volatile organic compounds (VOCs) released by decaying wood.

Methods

The study evaluated three commercially available electronic nose devices, each employing different sensing technologies:

  • AromaScan A32S: A conducting polymer-type sensor array (32 sensors) using pattern recognition and neural-net training algorithms.
  • LibraNose 2.1: A quartz crystal microbalance (QCM) system (8 sensors) that measures mass changes on quartz crystals coated with metalloporphyrines or polypyrrole films.
  • PEN3: A metal-oxide semiconductor (MOS) sensor array (10 sensors) that detects changes in conductivity due to reactions with volatile compounds.

The researchers prepared both healthy and artificially inoculated (decayed) wood samples from various tree species. They then exposed these samples to the e-noses to capture their volatile emissions. Data analysis involved constructing reference libraries, pattern recognition, principal component analysis (PCA), and linear discriminant analysis (LDA) to discriminate between healthy and decayed samples, and in some cases, to identify specific decay fungi.

Datasets

The study utilized two main types of samples:

  • Healthy Wood Samples: Collected from nine different tree species, sterilized, and used as controls.
  • Decayed Wood Samples: Produced by artificially inoculating healthy wood blocks with 11 different wood decay fungi.

Pure cultures of 24 wood decay fungi were also analyzed. The study did not introduce new benchmark datasets but rather used these prepared samples to evaluate the performance of the existing e-nose technologies.

Key Insights and Ideas

  • Early Detection is Feasible: All three e-noses demonstrated the ability to detect incipient wood decay, a critical step towards proactive tree management.
  • VOCs as a Signature: The study reinforces the idea that specific VOC profiles can serve as reliable "fingerprints" for healthy versus decaying wood, and even for different types of decay fungi.
  • Diverse Technologies, Similar Goals: The paper highlights that different e-nose technologies (conducting polymer, QCM, MOS) can achieve similar goals (discriminating healthy from decayed wood), though with varying degrees of effectiveness and analytical approaches.
  • The Power of PCA/LDA: Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) proved to be powerful tools for visualizing and quantifying the differences in volatile profiles, enabling clear separation of sample classes.
  • Host-Fungus Interaction Matters: The study observed that the specific combination of a host wood species and a decay fungus can influence the volatile signature, sometimes leading to unique or outlying data points.

Key Results

  • AromaScan A32S: Achieved 93.2% correct identification of wood decay fungi in pure culture. It also showed good performance in discriminating healthy wood (80% correctly identified) and decayed wood (95% correctly identified).
  • LibraNose 2.1: Demonstrated exceptional performance in discriminating healthy from incipient decayed woods using PCA, with principal components 1 and 2 explaining 91.0% and 7.9% of the data variance, respectively. It showed very few outlying data points, indicating high precision.
  • PEN3: Performed relatively poorly in discriminating certain fungus-host combinations but was better with others. Pair-wise PCA showed discrimination powers above 86% for decayed vs. healthy wood types in some comparisons. Its performance was significantly better (>98%) when distinguishing between different types of healthy woods.

Limitations and Future Work

  • Limited Direct Comparison: The authors note that a direct statistical comparison of the three e-noses was not possible due to differences in their analytical methods and software. Performance was assessed independently.
  • Sample Size and Variability: The AromaScan showed poorer results with small sample sizes, suggesting that increasing replications for reference library construction could improve confidence levels. Geographical variations and host-wood composition can also influence metabolite production, requiring locality-specific reference libraries.
  • Incipient vs. Advanced Decay: The study focused on incipient decay. More research is needed to assess the e-noses' performance with more advanced stages of decay, which are more common in real-world scenarios.
  • Field Testing: The current study used laboratory-prepared samples. Future work should involve field-testing these e-noses by sampling volatiles directly from living trees in urban environments.
  • Environmental Factors: Variability in ambient air humidity and temperature, as well as air pollutants, can affect e-nose readings. Future work should address these challenges through improved preconditioning of reference air and temperature regulation.
  • Miniaturization and Sensor Optimization: Reducing the size and weight of mobile e-noses and optimizing sensor design for specific applications are areas for future development.

Summary updated

BibTeX
@misc{baietto2010evaluation,
  author = {Manuela Baietto and Alphus D. Wilson and Daniele Bassi and Francesco Ferrini},
  title = {{Evaluation of Three Electronic Noses for Detecting Incipient Wood Decay}},
  year = {2010},
  doi = {10.3390/s100201062},
}
Figure