Deep convolutional neural network: a novel approach for the detection of Aspergillus fungi via stereomicroscopy

Fungi of the genus Aspergillus are ubiquitously distributed in nature, and some cause invasive aspergillosis (IA) infections in immunosuppressed individuals and contamination in agricultural products. Because microscopic observation and molecular detection of Aspergillus species represent the most operator-dependent and time-intensive activities, automated and cost-effective approaches are needed. To address this challenge, a deep convolutional neural network (CNN) was used to investigate the ability to classify various Aspergillus species. Using a dissecting microscopy (DM)/stereomicroscopy platform, colonies on plates were scanned with a 35× objective, generating images of sufficient resolution for classification. A total of 8,995 original colony images from seven Aspergillus species cultured in enrichment medium were gathered and autocut to generate 17,142 image crops as training and test datasets containing the typical representative morphology of conidiophores or colonies of each strain. Encouragingly, the Xception model exhibited a classification accuracy of 99.8% on the training image set. After training, our CNN model achieved a classification accuracy of 99.7% on the test image set. Based on the Xception performance during training and testing, this classification algorithm was further applied to recognize and validate a new set of raw images of these strains, showing a detection accuracy of 98.2%. Thus, our study demonstrated a novel concept for an artificial-intelligence-based and cost-effective detection methodology for Aspergillus organisms, which also has the potential to improve the public's understanding of the fungal kingdom.

Mushroom Emergence Detected by Combining Spore Trapping with Molecular Techniques

Obtaining reliable and representative mushroom production data requires time-consuming sampling schemes. In this paper, we assessed a simple methodology to detect mushroom emergence by trapping the fungal spores of the fruiting body community in plots where mushroom production was determined weekly. We compared the performance of filter paper traps with that of funnel traps and combined these spore trapping methods with species-specific quantitative real-time PCR and Illumina MiSeq to determine the spore abundance. Significantly more MiSeq proportional reads were generated for both ectomycorrhizal and saprotrophic fungal species using filter traps than were obtained using funnel traps. The spores of 37 fungal species that produced fruiting bodies in the study plots were identified. Spore community composition changed considerably over time due to the emergence of ephemeral fruiting bodies and rapid spore deposition (lasting from 1 to 2 weeks), which occurred in the absence of rainfall events. For many species, the emergence of epigeous fruiting bodies was followed by a peak in the relative abundance of their airborne spores. There were significant positive relationships between fruiting body yields and spore abundance in time for five of seven fungal species. There was no relationship between fruiting body yields and their spore abundance at plot level, indicating that some of the spores captured in each plot were arriving from the surrounding areas. Differences in fungal detection capacity by spore trapping may indicate different dispersal ability between fungal species. Further research can help to identify the spore rain patterns for most common fungal species.IMPORTANCE Mushroom monitoring represents a serious challenge in economic and logistical terms because sampling approaches demand extensive field work at both the spatial and temporal scales. In addition, the identification of fungal taxa depends on the expertise of experienced fungal taxonomists. Similarly, the study of fungal dispersal has been constrained by technological limitations, especially because the morphological identification of spores is a challenging and time-consuming task. Here, we demonstrate that spores from ectomycorrhizal and saprotrophic fungal species can be identified using simple spore traps together with either MiSeq fungus-specific amplicon sequencing or species-specific quantitative real-time PCR. In addition, the proposed methodology can be used to characterize the airborne fungal community and to detect mushroom emergence in forest ecosystems.

Sequencer-Based Capillary Gel Electrophoresis (SCGE) Targeting the rDNA Internal Transcribed Spacer (ITS) Regions for Accurate Identification of Clinically Important Yeast Species

Accurate species identification of Candida, Cryptococcus, Trichosporon and other yeast pathogens is important for clinical management. In the present study, we developed and evaluated a yeast species identification scheme by determining the rDNA internal transcribed spacer (ITS) region length types (LTs) using a sequencer-based capillary gel electrophoresis (SCGE) approach. A total of 156 yeast isolates encompassing 32 species were first used to establish a reference SCGE ITS LT database. Evaluation of the ITS LT database was then performed on (i) a separate set of (n = 97) clinical isolates by SCGE, and (ii) 41 isolates of 41 additional yeast species from GenBank by in silico analysis. Of 156 isolates used to build the reference database, 41 ITS LTs were identified, which correctly identified 29 of the 32 (90.6%) species, with the exception of Trichosporon asahii, Trichosporon japonicum and Trichosporon asteroides. In addition, eight of the 32 species revealed different electropherograms and were subtyped into 2–3 different ITS LTs each. Of the 97 test isolates used to evaluate the ITS LT scheme, 96 (99.0%) were correctly identified to species level, with the remaining isolate having a novel ITS LT. Of the additional 41 isolates for in silico analysis, none was misidentified by the ITS LT database except for Trichosporon mucoides whose ITS LT profile was identical to that of Trichosporon dermatis. In conclusion, yeast identification by the present SCGE ITS LT assay is a fast, reproducible and accurate alternative for the identification of clinically important yeasts with the exception of Trichosporon species.

[MASS-SPECTROMETRY IN MICROBIOLOGICAL PRACTICE OF SCIENTIFIC CENTRE OF OBSTETRICS, GYNECOLOGY AND PERINATOLOGY]

AIM

Comparative evaluation of species identification of microorganisms by MALDI-TOF mass-spectrometry and automatic biochemical analyzer VITEK2 Compact 30.

MATERIALS AND METHODS

Species identification of 18,400 isolates of microorganisms (staphylococci, streptococci, enterococci, enterobacteria, nonfermenting gram-negative bacteria, lactobacilli, anaerobes, yeast fungi, neisseriae), isolated from vagina of pregnant and non-pregnant women and from newborns, was carried out. Identification of the isolated microorganisms was carried out by automatic bacteriologic analyzer VITEK2 Compact30 (BioMerieuX, France) and MALDI-TOF-MS analysis method on AutofleXIII (Bruker Daltonics, Germany) mass-spectrometer.

RESULTS

Comparative identification of 2005 isolates of microorganisms was carried out. Sequencing of ribosomal RNA was used as a reference method. Authenticity of species identification my MALDI-TOF-MS analysis method was: for staphylococci (95.8%), enterococci (97.5%), enterobacteria (98.4%), nonfermenting gram-negative bacteria (93.6%), β-hemolytiC staphylococci (93.8%), lactobacilli (92.8%), yeast fungi (99.9%).

CONCLUSION

Introduction of MALDI-TOF-MS analysis technology into practical work of microbiological laboratories exceeds previously used methods of microbiological testing in terms of speed; cost and authenticity of identification of a wide spectrum of microorganisms.

Evaluation of Vitek 2 for identification of yeasts in the clinical laboratory

The Vitek 2 system was compared with conventional assimilation, fermentation and morphological methods for its ability to identify yeast isolates from among 151 clinical specimens and 16 known type culture or quality control strains. An unequivocal identification was obtained for 155 (92.8%) isolates, with low discrimination for nine (5.4%) and false identification for three (1.8%) isolates. All isolates of Candida albicans, Candida glabrata and Candida krusei were identified correctly. It was concluded that the Vitek 2 system offers an excellent alternative for the identification of yeasts in a clinical laboratory.

Direct and rapid discrimination of aflatoxigenic strains based on fibre-optic room temperature phosphorescence detection

An innovative analytical methodology for the rapid identification of aflatoxin-producing moulds belonging to Aspergillus genus is presented here. The procedure is based on the measurement, using a fibre-optic luminometer, of the room temperature phosphorescence (RTP) emitted by aflatoxins produced by isolated aflatoxigenic strains, cultured in a special culture medium consisting of malt extract agar modified with beta-cyclodextrin and sodium deoxycholate for RTP induction. Unequivocal detection of the presence of aflatoxins in the culture medium is achieved within the first 36 h of incubation at 32 degrees C, owing to the selectivity and sensitivity of the RTP emission, as compared with the minimum of 72 h needed using a conventional microbiological method. In a first step, the capability of aflatoxin standard solutions to emit analytically useful RTP was evaluated. In this line all experimental conditions were optimised for in vitro induction of RTP from aflatoxins. In a second step, a simple analytical test was developed and it has been evaluated for the rapid identification of aflatoxigenic strains, as a discriminating assay from non-aflatoxigenic strains based on the measurement of experimental RTP emission observed. Confirmation of aflatoxin production on the studied culture plates was accomplished by means of an HPLC/fluorescence reference method.

Phenotypic methods and commercial systems for the discrimination between C. albicans and C. dubliniensis

Candida dubliniensis is a recently described Candida species associated with oral candidosis that exhibits a high degree of phenotypic similarity to Candida albicans. However, these species show differences in levels of resistance to antimycotic agents and ability to cause infections. Therefore, accurate clinical identification of C. dubliniensis and C. albicans species is important in order to treat oral candidal infections. Phenotypic identification methods are easy-to-use procedures for routine discrimination of oral isolates in the clinical microbiology laboratory. However, C. dubliniensis may be so far underreported in clinical samples because most currently used identification methods fail to recognize this yeast. Phenotypic methods depend on growth temperature, carbon source assimilation, chlamydospore and hyphal growth production, positive or negative growth on special media and intracellular enzyme production, among others. In this review, some phenotypic methods are presented with a special emphasis on the discrimination of C. dubliniensis and C. albicans.

[Comparison of methods for the identification of the most common yeasts in the clinical microbiology laboratory]

We evaluated different methods for the routine identification of medically important yeasts. A total of 150 clinical isolates: 25 C. albicans, 25 C. tropicalis, 25 C. glabrata, 25 C. parapsilosis, 8 C. guilliermondii, 11 C. krusei and 31 Cryptococcus neoformans were tested by Yeast Biochemical Card bioMerieux Vitek (YBC), CHROMagar Candida (CHR). The addition of yeast morphology in Corn Meal agar-Tween 80 (AM) to YBC and CHR was also evaluated. The reference methods used were: API 20C, germ tube formation, AM, Christensen urea and Birdseed agar. YBC identified 135 yeasts with an overall accuracy of 90%. Sensitivity (S) and specificity (E) were: 92-98% for C. albicans and C. tropicalis; 84-99% for C. papapsilosis; 100-99% for C. glabrata; 91-100% for C. krusei; 63-98% for C. guilliermondii and 90-99% for Cryptococcus neoformans, respectively. CHR identified correctly 100% for C. albicans, 92% for C. tropicalis and 91% for C. krusei. Both methods combined with AM provided 100% S and E. We found that YBC system was appropriate for identification of yeasts isolated from human sources. CHR was effective and easy to use for identification of C. albicans, C. tropicalis and C. krusei. The routine use of AM with both methods is recommended.