ESCMID eAcademy

ePosters are only viewable by those with a valid ECCMID 2021 registration.
If you already have a registration, please login to the online platform here.
If you are not yet registered, you can do so here.
Abstract
Discussion Forum (0)
Abstract number: 713

Session Type: ePosters

Session Title: ePosters

Authors(s): M. Halabi (1), J. Rau (2), N. Mauder (3)

Authors Affiliations(s): (1) Krankenhaus der Barmherzigen Schwestern Ried (KBSR), Austria, (2) Chemisches und Veterinäruntersuchungsamt Stuttgart (CVUAS), Germany, (3) Bruker Daltonik GmbH, Germany

Background:

Legionella are Gram-negative bacilli including L. pneumophila and other species, that can cause legionellosis, also known as Legionnaires' disease and a flu-like illness known as Pontiac fever, mostly caused by L. pneumophila serogroup 1. Legionella spp. are common in many environmental habitats, including aquatic systems, with at least 50 species divided over 70 serogroups. Legionella spp., and especially L. pneumophila with known 15 serogroups, are not transmissible from person to person but humans can be infected via air-borne transmission, e.g. through cooling towers or warm-water sources that produce aerosols.

The identification process of species and serogroups depend on conventional, time-consuming exclusion diagnostics (e.g. via cultivation on blood-agar) and agglutination procedures concerning serogroup identification. Therefore, new short-time techniques from cultivation to identification are highly welcome, to quickly identify the high-risk serogroup 1.

Methods:

We analysed 89 L. pneumophila isolates from Austria and Germany. Bacteria were cultivated, processed and measured as illustrated in figure Material & Methods. Measurements resulted in 3204 spectra total.

Spectra of German isolates were assigned to the corresponding classes/serogroups (SG 1 vs. SG 2-15), and six principal components were extracted using principal component analysis (PCA). Then an artificial neural network (ANN) was trained, utilizing the class-specific spectral differences (see figure Average Spectra).

The ANN-based classifier was applied on 1728 spectra of Austrian isolates, and predictions were then compared with results of the standard latex agglutination test, used as gold standard.

Results:

All spectra of the 22 Austrian isolates having serogroup 1 were correctly classified, as were all spectra of the 26 Austrian isolates having serogroup 2 to 15.

According to PCA (see figure), the effect of humidity on IR spectra is greater than the spectral difference between serogroups.

Conclusions:

The standard latex agglutination test for distinguishing serogroup 1 vs. 2-15 might – in the long run – be substituted or cross-checked with infrared spectroscopy, which could help investigations in outbreak scenarios and/or routine labs analysing water samples.

The ANN-based classifier is very well capable of distinguishing the serogroups despite of a major disturbing influence like variations in atmospheric humidity from 10% to 85%.

Keyword(s): Serotyping, Fourier-Transform Infrared Spectroscopy (FTIR), Artificial Neural Network (ANN)


COI Other: Norman Mauder is employee of Bruker Daltonics, the manufacturer of the IR Biotyper(R).
Abstract number: 713

Session Type: ePosters

Session Title: ePosters

Authors(s): M. Halabi (1), J. Rau (2), N. Mauder (3)

Authors Affiliations(s): (1) Krankenhaus der Barmherzigen Schwestern Ried (KBSR), Austria, (2) Chemisches und Veterinäruntersuchungsamt Stuttgart (CVUAS), Germany, (3) Bruker Daltonik GmbH, Germany

Background:

Legionella are Gram-negative bacilli including L. pneumophila and other species, that can cause legionellosis, also known as Legionnaires' disease and a flu-like illness known as Pontiac fever, mostly caused by L. pneumophila serogroup 1. Legionella spp. are common in many environmental habitats, including aquatic systems, with at least 50 species divided over 70 serogroups. Legionella spp., and especially L. pneumophila with known 15 serogroups, are not transmissible from person to person but humans can be infected via air-borne transmission, e.g. through cooling towers or warm-water sources that produce aerosols.

The identification process of species and serogroups depend on conventional, time-consuming exclusion diagnostics (e.g. via cultivation on blood-agar) and agglutination procedures concerning serogroup identification. Therefore, new short-time techniques from cultivation to identification are highly welcome, to quickly identify the high-risk serogroup 1.

Methods:

We analysed 89 L. pneumophila isolates from Austria and Germany. Bacteria were cultivated, processed and measured as illustrated in figure Material & Methods. Measurements resulted in 3204 spectra total.

Spectra of German isolates were assigned to the corresponding classes/serogroups (SG 1 vs. SG 2-15), and six principal components were extracted using principal component analysis (PCA). Then an artificial neural network (ANN) was trained, utilizing the class-specific spectral differences (see figure Average Spectra).

The ANN-based classifier was applied on 1728 spectra of Austrian isolates, and predictions were then compared with results of the standard latex agglutination test, used as gold standard.

Results:

All spectra of the 22 Austrian isolates having serogroup 1 were correctly classified, as were all spectra of the 26 Austrian isolates having serogroup 2 to 15.

According to PCA (see figure), the effect of humidity on IR spectra is greater than the spectral difference between serogroups.

Conclusions:

The standard latex agglutination test for distinguishing serogroup 1 vs. 2-15 might – in the long run – be substituted or cross-checked with infrared spectroscopy, which could help investigations in outbreak scenarios and/or routine labs analysing water samples.

The ANN-based classifier is very well capable of distinguishing the serogroups despite of a major disturbing influence like variations in atmospheric humidity from 10% to 85%.

Keyword(s): Serotyping, Fourier-Transform Infrared Spectroscopy (FTIR), Artificial Neural Network (ANN)


COI Other: Norman Mauder is employee of Bruker Daltonics, the manufacturer of the IR Biotyper(R).
Serogroup classification of Legionella pneumophila using transmission infrared spectroscopy and machine learning
Dr. Milad Halabi
Dr. Milad Halabi
ESCMID eAcademy. Halabi M. 07/09/2021; 327748; 713
user
Dr. Milad Halabi
Abstract
Discussion Forum (0)
Abstract number: 713

Session Type: ePosters

Session Title: ePosters

Authors(s): M. Halabi (1), J. Rau (2), N. Mauder (3)

Authors Affiliations(s): (1) Krankenhaus der Barmherzigen Schwestern Ried (KBSR), Austria, (2) Chemisches und Veterinäruntersuchungsamt Stuttgart (CVUAS), Germany, (3) Bruker Daltonik GmbH, Germany

Background:

Legionella are Gram-negative bacilli including L. pneumophila and other species, that can cause legionellosis, also known as Legionnaires' disease and a flu-like illness known as Pontiac fever, mostly caused by L. pneumophila serogroup 1. Legionella spp. are common in many environmental habitats, including aquatic systems, with at least 50 species divided over 70 serogroups. Legionella spp., and especially L. pneumophila with known 15 serogroups, are not transmissible from person to person but humans can be infected via air-borne transmission, e.g. through cooling towers or warm-water sources that produce aerosols.

The identification process of species and serogroups depend on conventional, time-consuming exclusion diagnostics (e.g. via cultivation on blood-agar) and agglutination procedures concerning serogroup identification. Therefore, new short-time techniques from cultivation to identification are highly welcome, to quickly identify the high-risk serogroup 1.

Methods:

We analysed 89 L. pneumophila isolates from Austria and Germany. Bacteria were cultivated, processed and measured as illustrated in figure Material & Methods. Measurements resulted in 3204 spectra total.

Spectra of German isolates were assigned to the corresponding classes/serogroups (SG 1 vs. SG 2-15), and six principal components were extracted using principal component analysis (PCA). Then an artificial neural network (ANN) was trained, utilizing the class-specific spectral differences (see figure Average Spectra).

The ANN-based classifier was applied on 1728 spectra of Austrian isolates, and predictions were then compared with results of the standard latex agglutination test, used as gold standard.

Results:

All spectra of the 22 Austrian isolates having serogroup 1 were correctly classified, as were all spectra of the 26 Austrian isolates having serogroup 2 to 15.

According to PCA (see figure), the effect of humidity on IR spectra is greater than the spectral difference between serogroups.

Conclusions:

The standard latex agglutination test for distinguishing serogroup 1 vs. 2-15 might – in the long run – be substituted or cross-checked with infrared spectroscopy, which could help investigations in outbreak scenarios and/or routine labs analysing water samples.

The ANN-based classifier is very well capable of distinguishing the serogroups despite of a major disturbing influence like variations in atmospheric humidity from 10% to 85%.

Keyword(s): Serotyping, Fourier-Transform Infrared Spectroscopy (FTIR), Artificial Neural Network (ANN)


COI Other: Norman Mauder is employee of Bruker Daltonics, the manufacturer of the IR Biotyper(R).
Abstract number: 713

Session Type: ePosters

Session Title: ePosters

Authors(s): M. Halabi (1), J. Rau (2), N. Mauder (3)

Authors Affiliations(s): (1) Krankenhaus der Barmherzigen Schwestern Ried (KBSR), Austria, (2) Chemisches und Veterinäruntersuchungsamt Stuttgart (CVUAS), Germany, (3) Bruker Daltonik GmbH, Germany

Background:

Legionella are Gram-negative bacilli including L. pneumophila and other species, that can cause legionellosis, also known as Legionnaires' disease and a flu-like illness known as Pontiac fever, mostly caused by L. pneumophila serogroup 1. Legionella spp. are common in many environmental habitats, including aquatic systems, with at least 50 species divided over 70 serogroups. Legionella spp., and especially L. pneumophila with known 15 serogroups, are not transmissible from person to person but humans can be infected via air-borne transmission, e.g. through cooling towers or warm-water sources that produce aerosols.

The identification process of species and serogroups depend on conventional, time-consuming exclusion diagnostics (e.g. via cultivation on blood-agar) and agglutination procedures concerning serogroup identification. Therefore, new short-time techniques from cultivation to identification are highly welcome, to quickly identify the high-risk serogroup 1.

Methods:

We analysed 89 L. pneumophila isolates from Austria and Germany. Bacteria were cultivated, processed and measured as illustrated in figure Material & Methods. Measurements resulted in 3204 spectra total.

Spectra of German isolates were assigned to the corresponding classes/serogroups (SG 1 vs. SG 2-15), and six principal components were extracted using principal component analysis (PCA). Then an artificial neural network (ANN) was trained, utilizing the class-specific spectral differences (see figure Average Spectra).

The ANN-based classifier was applied on 1728 spectra of Austrian isolates, and predictions were then compared with results of the standard latex agglutination test, used as gold standard.

Results:

All spectra of the 22 Austrian isolates having serogroup 1 were correctly classified, as were all spectra of the 26 Austrian isolates having serogroup 2 to 15.

According to PCA (see figure), the effect of humidity on IR spectra is greater than the spectral difference between serogroups.

Conclusions:

The standard latex agglutination test for distinguishing serogroup 1 vs. 2-15 might – in the long run – be substituted or cross-checked with infrared spectroscopy, which could help investigations in outbreak scenarios and/or routine labs analysing water samples.

The ANN-based classifier is very well capable of distinguishing the serogroups despite of a major disturbing influence like variations in atmospheric humidity from 10% to 85%.

Keyword(s): Serotyping, Fourier-Transform Infrared Spectroscopy (FTIR), Artificial Neural Network (ANN)


COI Other: Norman Mauder is employee of Bruker Daltonics, the manufacturer of the IR Biotyper(R).

By clicking “Accept Terms & all Cookies” or by continuing to browse, you agree to the storing of third-party cookies on your device to enhance your user experience and agree to the user terms and conditions of this learning management system (LMS).

Cookie Settings
Accept Terms & all Cookies