A comparison of counting methods for hourly airborne spore concentrations

A comparison of counting methods for hourly airborne spore concentrations

J ALLERGY CLIN IMMUNOL VOLUME 111, NUMBER 2 Abstracts S91 1 Classificationof AirsporaUsing SupportVector Machines (SVM) 3 AConcentrations Comparis...

127KB Sizes 2 Downloads 51 Views

J ALLERGY CLIN IMMUNOL VOLUME 111, NUMBER 2

Abstracts

S91

1 Classificationof AirsporaUsing SupportVector Machines (SVM)

3 AConcentrations Comparison of Counting Methods for Hourly Airborne Spore

S. Kumar, S. H. Ong, S. Ranganath, T. C. Ong, F. T. Chew; National University of Singapore, Singapore, SINGAPORE. RATIONALE: Manual quantification of airspora is a slow and laborious process. Automation of this process would not only make it faster and less laborious, it may be more amendable to larger number of slides/images, and increase accuracy and consistency. Computer based classification is seen as a vital step in the automation of airspora counts via image analysis. METHODS: We evaluated the performance of two types of Support Vector Machines, SVM (polynomial or gaussian), against statistical methods such as linear discriminants and muhilayer perceptrons (MLP), to classify airspora. A data set of eleven types of airspora common to the tropical Southeast Asian region, comprising fungal spores (Curvularia sp., Dreschlera spp., Di~,mosphaeria sp., Pithomyces sp.), fern spores (Asplenium nidus, Nephrolepis auriculata, Stenochlaena palustris) and pollen (Acacia sp., Casuarina equisetifolia, Elaeis guineensis, Podocarpus sp./Pinus) was analyzed. Each data has 20 size and shape features and one textural feature (roughness). 4500 specimens were evaluated. Approximately 1500 were used for training, while the remainder was used as the "unseen" set to test the classification performance. RESULTS: Polynomial and gaussian SVM give an overall classification accuracy of 94.3% and 94.0%, respectively, on unseen airspora data sets compared to 93.2% by MLP and 89.1% by linear discriminants. We observed that linear discriminants give poor accuracy since distribution of the labeled data set is non-linearly separable. MLP, though comparable, required additional heuristics based techniques to ensure its performance on unseen data is satisfactory. CONCLUSIONS: SVM gives better classification accuracy on unseen airspora data sets compared to MLP and statistical linear discriminant.

L. Lindley ],2, E. Levetin2; t Biology, Northeastern State University, Tahlequah, OK, 2Biological Science, University of Tulsa. Tulsa, OK. RATIONALE: Air sampling can be a valuable tool for allergists, who may advise patients to stay indoors on days with high pollen or spore levels. Outdoor air sampling is commonly done with a Burkard Spore Trap. Two counting methods are generally used: the single longitudinal traverse method (SLM) and the twelve transverse traverse method (TTM). SLM is fast and provides a good approximation of the average daily spore concentration. TTM takes more time, but it gives a better approximation of the average daily concentration and senses diurnal rhythm. However, TTM is too time-consuming for routine use. The current project was undertaken to find a counting method that shows changes in spore concentrations during the day but takes less time than the TTM. METHODS: The 28-fields method is a variant of TTM in which only 28 fields are counted per traverse. To test this method, the twelve traverses per slide were counted twice. For each traverse, the entire traverse was counted, then 28 fields were recounted. The bihourly spore concentrations were then calculated and compared statistically using the Wilcoxon paired-sample test and Spearman correlation. RESULTS: Concentrations obtained from counting 28 fields correlated well (p<0.05) with concentrations from the whole traverse, especially for the more abundant spore types. CONCLUSIONS: The 28-fields method may be considered as an alternative to TTM for routine monitoring of spore concentrations. This will provide a better understanding of the daily peaks in spore concentration and enable physicians to advise patients when to avoid outdoors.

Funding: Biomedical Research Council

82

Leaf Surface Fungi as Aeroallergens

K. Dorsey, E. Levetin: Biological Science, University of Tulsa, Tulsa, OK. RATIONALE: Airborne fungal spores are well known allergens. Although it is generally assumed that fungi on leaf surfaces are major contributors to the air spora, little data are available comparing the types of fungi found on leaf surfaces with those in the atmosphere. The present study was undertaken to address this. METHODS: Air sampling was carried out with a Burkard spore trap using standard methods. Leaf samples from Ulmus and Quercus trees were aseptically collected weekly. Areas (4 cm 2) on both upper and lower leaf surfaces were separately wiped with sterile cotton swabs and then washed in 1 ml of sterile distilled water. Suspensions were plated on malt extract agar with streptomycin. Cultures were incubated at room temperature for 5-7 days and examined microscopically. RESULTS: Mean concentration of fungi on elm leaves was 48 colony forming units (CFU)/cm 2 with a range from 10-162 CFU/cm 2. The level of fungi on oak leaves was lower with a mean of 31 CFU/cm 2 and a range of 0.6-84.7 CFU/cm 2. Yeasts were the most abundant fungi on both leaves representing 67% of the total on oak and 58% on elm. The most abundant mycelial fungi were species of Phoma, followed by Cladosporium, then Alternaria. Approximately 22 different types of fungi were identified. Although yeasts and Phoma spores are not readily airborne, most other leaf-surface fungi were commonly found in the air samples. CONCLUSIONS: These data suggest that when the total leaf area of a tree is considered, leaf-surface fungi contribute significantly to airborne fungal allergens.

Funding: University of Tulsa

Funding: NIH-BRIN

4 AeroallergenSurveyofthe Texas Panhandle Using a BurkardVolumetric SporeTrap N. Ghosh 1, B. Patten l, G. Lewellen l, C. Saadeh 2, M. Gaylor2; ]Department of Life, Earth and Environmental Sciences, West Texas A & M University, Canyon, TX. 2Amarillo Center for Clinical Research/Allergy ARTS, Amarillo, TX. RATIONALE: The purpose of our analysis of pollen data is to assess and enumerate the impact of airborne pollen and mold spores on the breathing and causes of allergic rhinitis in individuals that are carried on the atmospheric oscillations of the exterior environment in the Northwest Texas Panhandle area. METHODS: The analysis of air was performed through the collection of pollen and spores through the use of the Burkard Volumetric Spore Trap. Air samples were taken every 24 hours and daily mean concentration (grains/cubic meter of air) were assessed. Aeroallergen data were correlated with daily weather conditions of maximum and minimum temperature, daily precipitation and peak wind speeds. Data was also correlated with clinical patient cases. RESULTS: The most significant allergens present during these summer months were AIternaria, ascospores, Cladosporium, Drechslera, grass (Poaceae) pollen, ragweed (Ambrosia) pollen and Pine (Pinus) pollen. Temperature was found to have an inverse relationship with mold spores. For the summer months that we observed, the most dominant pollen was grass (Poaceae), which peaked in July and then dropped off in August. In mid-August, the dominant pollen changes to ragweed (Ambrosia sp.), corresponding to the beginning of the flowering season for ragweed. CONCLUSIONS-" The effect of aeroallergens on clinical patients was apparent. The number of reported cases of allergic rhinitis increased proportionally to the increase in overall allergen counts. Most significant was the increase in number of patients corresponding with increases in mold and A. artemisifolia counts.

Funding: Killgore Research, Allergy Clinic of Dr. Saadeh