The relative and absolute amount of persons suffering from an allergy in Germany is continuously rising. Therefore, the ability to accurately measure airborne pollen concentration in the environment is an important goal for palynology, especially for improving the understanding of allergies. The search of systems and methodologies to fulfill such an ability has been unsatisfactory for practical usage to date. The main reason is the great amount of time needed to identify huge volumes of particles in airborne samples, which prevents palynologists from processing opportunely enough statistically suitable information. Furthermore, due to the individual's dynamics, the pollen concentration that affects every patient cannot be accurately associated to measurements from stationary pollen monitors. Forecasts for pollen exposure in Germany are made with such stationary systems which are not representative for the individual environment of an allergy sufferer in most cases. A portable and trackable device for capturing pollen would help to collect data about the pollen exposure around the allergy sufferer in a direct way. This allows personalized pollen profiling, which can improve the medication for the proband. It would also help to test medicine individually in his or her personal vicinity, and to find the individual pollen threshold levels. The achieved pollen information from the device combined with a location can be used to improve forecast systems by combining existing databases with the collected data. The aim of this project is to create an allergic patient profile based on reliable information of most allergenic pollen taxa concentrations in the vicinity of a patient, measured on an individual basis. Gathering such accurate individual pollen concentration data in a spatial context will benefit not only patients through more precise medication, but also will improve pollen distribution models and forecasting.
Together with our partners, we are developing a complete system of pollen collection, automatic classification, and a personal profile website. Bluestone Technology is developing a new traceable device enabling the capture of airborne-particles on exchangeable slides. The slide is then scanned through the use of a bright-field microscope to produce a high resolution image of the captured aerosols. After processing of these images by i3mainz, Health and Media will provide an expert view in which results from our classification process will be displayed for use by both professionals and the patients.
To prepare for the classification process, i3mainz utilizes image processing techniques for the extraction of pollen particle segmented images. We have both researched pollen characteristics and spoken with an expert palynologist in order to determine the correct features to extract, which can be divided into the following categories: shape, color, texture, and local features. The images of the slides are then analyzed for feature extraction, the results of which will be fed into a classification system that will automatically determine the pollen particlesâ€™ species. Furthermore, we are continuously researching best practices in classification techniques and software, and also plan to test incorporation of an expert knowledge base into the system. Because the slide will also document environmental data such as GPS-coordinates, air pressure and temperature, this information can then be associated with Â the recorded data and allows sophisticated analysis processes in geostatistics.
Towards this end, a geostatistic system is under development which uses OGC standards such as Web Feature Service (WFS), Web Mapping Service (WMS) and Web Processing Service (WPS). Within this system, both overview statistical thematic maps, and then user-based visualizations are planned.
As of the time of this writing, we designed and developed software that is capable of locating and extracting potential pollen particles out of a scanned image of a slide containing captured airborne particles. An algorithm has also been developed to outline the segmented particleâ€™s area. These boundaries are then used as the key areas for feature extraction, with our software extracting the features of color and shape (texture and local feature extraction are still under development). Also, a connection to the server infrastructure, utilizing an XML for data exchange, has been developed. Within the next two months, we will have implemented the first version of a classification system that uses these extracted features to determine the pollen species, as well as implemented the basic data exchange throughout the systemâ€™s infrastructure