Comparative Evaluation of the Effectiveness and Efficiency of Computational Methods in the Detection of Asbestos Cement in Hyperspectral Images

Our plain-English summary of Comparative Evaluation of the Effectiveness and Efficiency of Computational Methods in the Detection of Asbestos Cement in Hyperspectral Images, by G.E. Chanchí-Golondrino, M. Saba, M.A. Ospina-Alarcón (Materials, 2025).

The study

• Authors: G.E. Chanchí-Golondrino, M. Saba, M.A. Ospina-Alarcón
• Journal: Materials
• Published: 2025;18(15)
• Cited by: 1 article, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Among the existing challenges in the field of hyperspectral imaging, the need to optimize memory usage and computational capacity in material detection methods stands out, given the vast amount of data associated with the hundreds of reflectance bands. In line with this, this article proposes a comparative study on the effectiveness and efficiency of five computational methods for detecting composite material asbestos cement (AC) in hyperspectral images: correlation, spectral differential similarity (SDS), Fourier phase similarity (FPS), area under the curve (AUC), and decision trees (DT). The novelty lies in the comparison between the first four methods, which represent the spectral proximity method and a machine learning method, such as DT. Furthermore, SDS and FPS are novel methods proposed in the present document. Given the accuracy that detection methods based on supervised learning have demonstrated in material identification, the results obtained from the DT model were compared with the percentage of AC detected in a hyperspectral image of the Manga neighborhood in the city of Cartagena by the other four methods. Similarly, in terms of computational efficiency, a 20 × 20 pixel region with 380 bands was selected for the execution of multiple repetitions of each of the five computational methods considered, in order to obtain the average processing time of each method and the relative efficiency of the methods with respect to the method with the best effectiveness. The decision tree (DT) model achieved the highest classification accuracy at 99.4%, identifying 11.44% of asbestos cement (AC) pixels in the reference image. However, the correlation method, while detecting a lower percentage of AC pixels (9.72%), showed the most accurate visual performance and had no spectral overlap, with a 1.4% separation between AC and non-AC pixels. The SDS method was the most computationally efficient, running 23.85 times faster than the DT model. The proposed methods and results can be applied to other hyperspectral imaging tasks involving material identification in urban environments, especially when balancing accuracy and computational efficiency is essential.

In plain English

This was a computer-methods study. It did not study people or health effects. The authors compared five ways of finding asbestos cement in one hyperspectral image of the Manga neighbourhood in Cartagena de Indias, Colombia, covering about 0.4 square kilometres. They used 100 asbestos cement pixels, taken from places where field samples were confirmed by polarized light microscopy, and 100 pixels of other materials such as vegetation, water and metal roofs. On these sample pixels, the decision tree method scored 0.994 for accuracy, precision and recall. Across the whole image, the methods labelled between 9.716% (correlation) and 11.907% (area under the curve) of pixels as asbestos cement. On visual inspection, correlation made the fewest errors, while area under the curve mistook vegetation and water for asbestos cement. On a small test area, correlation and spectral differential similarity ran 4.92 and 23.85 times faster than the decision tree. The authors concluded that correlation, spectral differential similarity and Fourier phase similarity were valid alternatives. They noted few samples, a single image, and unknown effects of asbestos fiber type and percentage.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Comparative evaluation of five computational methods (correlation, spectral differential similarity, Fourier phase similarity, area under the curve, decision tree) for detecting asbestos cement in a hyperspectral image. No people were studied and no health outcomes were measured.
• Who was studied: One reference hyperspectral image of the Manga neighbourhood, Cartagena de Indias, Colombia, about 0.4 square kilometres (725 by 850 pixels, 616,250 pixels in total, 380 bands from 400 nm to 2500 nm). The methods were set up using 200 sample pixels: 100 asbestos cement and 100 other materials such as vegetation, water bodies and metal roofs.
• Exposure: No human exposure was studied. The target material was asbestos cement (mainly roofing). The asbestos cement sample pixels came from georeferenced sites where 1 by 1 cm field specimens were examined by polarized light microscopy. The article does not report the asbestos fiber type.
• Main results:
  • Decision tree model, five-fold cross-validation on the 200 sample pixels: classification accuracy, precision and recall were all 0.994. In the confusion matrix, 100% of non-asbestos cement pixels and 98.9% of asbestos cement pixels were classified correctly.
  • Correlation method on the sample pixels: the lowest correlation for asbestos cement pixels was 1.4% higher than the highest correlation for non-asbestos cement pixels, so the two groups did not overlap.
  • Share of all image pixels detected as asbestos cement: correlation 9.716% and spectral differential similarity 10.428%.
  • Share of all image pixels detected as asbestos cement: area under the curve 11.907% and Fourier phase similarity 11.611%.
  • Share of all image pixels detected as asbestos cement: decision tree 11.442%.
  • On visual inspection of the full image, the correlation method had the fewest classification errors and the area under the curve method the most.
  • The area under the curve method misclassified vegetation and water as asbestos cement. The Fourier phase similarity method misclassified certain metal roofing.
  • The correlation method's results were used as the visually validated reference. Pixel agreement with it was 97.59% for Fourier phase similarity and 96.97% for spectral differential similarity.
  • Speed was tested on a 20 by 20 pixel, 380-band region over 25, 50, 75 and 100 repetitions. On average, the correlation method was 4.92 times and spectral differential similarity 23.85 times faster than the decision tree.
• The authors' conclusion: The authors concluded that the correlation, spectral differential similarity and Fourier phase similarity methods can be considered valid alternatives for material detection in hyperspectral images.
• Limitations the authors state:
  • The effect of the percentage and type of asbestos fibers in the asbestos cement samples on the spectral signature is unknown, and the number of asbestos cement and non-asbestos cement samples was relatively limited.
  • There was no spatial or temporal cross-validation, because only a single hyperspectral image and constrained field data were used.
  • Open-source tools may have performance limits in large-scale operational use compared with proprietary software.

Topics

Medical terms in this study, explained in our Lexicon: Asbestos-Cement.

Related studies

• Assessment of asbestos-cement roof distribution and prioritized intervention approaches through hyperspectral imaging — D. Enrique Valdelamar Martínez, M. Saba, et al. (2024)
• Sample preparation method for visualization of nanoparticulate captured on mixed cellulose ester filter media by enhanced darkfield microscopy and hyperspectral imaging — N.M. Neu-Baker, A.C. Eastlake, et al. (2019)
• Asbestos containing materials detection and classification by the use of hyperspectral imaging — G. Bonifazi, G. Capobianco, et al. (2018)
• Mapping asbestos-cement roofing with hyperspectral remote sensing over a large mountain region of the Italian Western Alps — F. Frassy, G. Candiani, et al. (2014)
• Comparative effectiveness of second-line immune checkpoint inhibitor therapy versus chemotherapy for malignant pleural mesothelioma — R.Y. Kim, Y. Li, et al. (2021)
• An image cytometric technique is a concise method to detect adenoviruses and host cell proteins and to monitor the infection and cellular responses induced — T. Morinaga, T.T.T. Nguyễn, et al. (2017)

Browse more research in Materials, or return to medical research.