Development of an automated asbestos counting software based on fluorescence microscopy

Our plain-English summary of Development of an automated asbestos counting software based on fluorescence microscopy, by M. Alexandrov, E. Ichida, T. Nishimura, et al (Environmental Monitoring and Assessment, 2015).

The study

• Authors: M. Alexandrov, E. Ichida, T. Nishimura, K. Aoki, T. Ishida, R. Hirota, T. Ikeda, T. Kawasaki, A. Kuroda
• Journal: Environmental Monitoring and Assessment
• Published: 2015;187(1):4166
• Cited by: 8 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

An emerging alternative to the commonly used analytical methods for asbestos analysis is fluorescence microscopy (FM), which relies on highly specific asbestos-binding probes to distinguish asbestos from interfering non-asbestos fibers. However, all types of microscopic asbestos analysis require laborious examination of large number of fields of view and are prone to subjective errors and large variability between asbestos counts by different analysts and laboratories. A possible solution to these problems is automated counting of asbestos fibers by image analysis software, which would lower the cost and increase the reliability of asbestos testing. This study seeks to develop a fiber recognition and counting software for FM-based asbestos analysis. We discuss the main features of the developed software and the results of its testing. Software testing showed good correlation between automated and manual counts for the samples with medium and high fiber concentrations. At low fiber concentrations, the automated counts were less accurate, leading us to implement correction mode for automated counts. While the full automation of asbestos analysis would require further improvements in accuracy of fiber identification, the developed software could already assist professional asbestos analysts and record detailed fiber dimensions for the use in epidemiological research.

In plain English

This study described the development and testing of software that automatically recognizes and counts asbestos fibers in fluorescence microscopy images, where asbestos is stained with a fluorescent probe. The authors tested it on 103 air sample filters supplied by Japan's Ministry of the Environment and two asbestos testing laboratories. Some samples were made in a chamber from relatively pure fibers, including chrysotile, amosite, crocidolite and tremolite. Others came from Japanese demolition sites or from milled asbestos-containing construction materials, and these had high dust loads. Automated counts were compared with manual counts made on the same images. The correlation coefficient was r=0.99, and correlation was especially good for samples above approximately 15–20 fibers per liter. At lower concentrations the software tended to overestimate, particularly in dusty samples, because dust particle edges were sometimes misidentified as fibers. For 27 high dust load samples with automated counts below 25 fibers per liter, correlation was r=0.64. It rose to r=0.92 after a qualified operator corrected the counts. The authors concluded that the software is not yet accurate enough for fully automated use and that its counts need correction and confirmation by qualified operators.

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: Laboratory method-development study: development of image analysis software and comparison of its automated fiber counts with manual counts on the same fluorescence microscopy images
• Who was studied: 103 air sample filters from Japan: 43 from the Ministry of the Environment, 30 from Kankyo Research (Hachioji) and 30 from Kyuden Sangyo (Fukuoka). 50 were 'low dust samples' prepared from relatively pure fibers in an air-tight chamber. 53 were 'high dust samples' collected at demolition sites in Japan or prepared by milling asbestos-containing construction materials.
• Exposure: Airborne fibers on filters. Low dust samples were prepared from chrysotile, amosite, crocidolite and tremolite, alone or in combination, some with mineral wool, and some from wollastonite and mineral wool. High dust samples contained asbestos fibers mixed with non-asbestos particles from demolition sites or milled construction materials. Samples were stained with the Asbester Air 2 kit.
• Main results:
  • Across all 103 samples (6,459 images), automated and manual counts had a correlation coefficient of r=0.99. The authors noted there were still some disparities in the accuracy of automated detection across the sample set.
  • Correlation was especially good for samples with asbestos fiber concentrations above approximately 15–20 fibers per liter (f/L).
  • At lower concentrations, automated counting tended to overestimate fiber concentrations. This was particularly noticeable for samples with high dust loads.
  • The main cause of the discrepancy was visible dust particles whose edges were sometimes misidentified as asbestos fibers.
  • For 27 high dust load samples with automated counts below 25 f/L, correlation between manual and automated counts was r=0.64. After operator correction of the automated counts it was r=0.92.
  • Fiber dimensions calculated by the software for curvy chrysotile fibers generally agreed with the estimates of human observers.
• The authors' conclusion: The authors concluded that the software's accuracy is not yet sufficient for fully automated counting. Its counts would require correction and confirmation by qualified operators, and use by non-professionals would require improvements to the fiber and particle identification algorithm.
• Limitations the authors state:
  • Only limited information was available about the sources and preparation methods of the samples, which were supplied by outside parties.
  • The main shortcoming of the software was a relatively high percentage of non-fibrous particles misidentified as fibers.
  • Correction mode, which is limited to the lower-confidence 'yellow' group fibers, is somewhat less accurate than manual counting, though faster.
  • The accuracy of the fiber identification algorithm was still below the level of well trained, experienced laboratory staff.

Topics

Medical terms in this study, explained in our Lexicon: Fiber Identification, Fiber Dimensions.

Related studies

• Differential Counting of Asbestos Using Phase Contrast and Fluorescence Microscopy — T. Nishimura, M. Alexandrov, et al. (2016)
• Evaluation of sensitivity of fluorescence-based asbestos detection by correlative microscopy — T. Ishida, M. Alexandrov, et al. (2012)
• Rapid on-site detection of airborne asbestos fibers and potentially hazardous nanomaterials using fluorescence microscopy-based biosensing — A. Kuroda, M. Alexandrov, et al. (2016)
• Synchrotron soft X-ray imaging and fluorescence microscopy reveal novel features of asbestos body morphology and composition in human lung tissues — L. Pascolo, A. Gianoncelli, et al. (2011)
• Software for Apportionment of Asbestos-Related Mesotheliomas — R.M. Ross (2016)
• Live-cell imaging of macrophage phagocytosis of asbestos fibers under fluorescence microscopy — T. Ishida, N. Fujihara, et al. (2019)

Browse more research in Environmental Monitoring and Assessment, or return to medical research.