Inhalation of respirable coal mine dust (RCMD) can lead to chronic lung diseases, including coal worker’s pneumoconiosis (CWP) and more severe forms such as progressive massive fibrosis. After the Federal Coal Mine Health and Safety Act was passed in 1969, limits on exposure to respirable dust were set, and the prevalence of CWP abruptly decreased. However, during the last two decades, a resurgence of the disease has been reported. Many authors have argued that the increasing numbers might be related to mining practices, including the extraction of thinner coal seams, characteristics of the mineral deposits, and more powerful cutting machines. Dust particles in coal mines are usually associated with three main sources: Coal particles are produced when the coal seam is being actively extracted. Silica and silicates are generated while cutting the rock strata surrounding the coal or during roof-bolting activities. Finally, rock dust application is the primary source of highly pure carbonates. Timely information about dust composition would allow the identification of potential dust sources and pursue efforts to control dust exposure efficiently. However, this information needs to be provided promptly since dust levels are dynamically changing through the shift. Currently, monitoring technologies such as the continuous personal dust monitor allow real-time measurements, but they are limited to total dust concentration and provide no information about dust composition. More recently, the National Institute for Occupational Safety and Health (NIOSH) has been developing an end-of-shift silica monitor. Still, technologies that offer information on dust composition in a semi-continuous manner are needed. In this work, a new monitoring concept is explored that has the potential to provide near real time data on RCMD constituents. The possible use of a portable optical microscopy (OM) combined with image processing techniques is explored as the basis for a novel RCDM monitoring device. The use of OM in different fields and the rapid development of automated image analysis reveals a clear opportunity that has not been yet exploited for mine dust monitoring applications. This thesis research consisted of two primary studies. The first was an analysis of lab-generated respirable dust samples containing the main mineralogical classes in RCMD (i.e., coal, silica, kaolinite as a proxy for silicate minerals, and a real rock dust product). Samples were imaged using a polarizing microscope and analyzed using an image processing routine to identify and classify particles based on optical characteristics. Specifically, birefringence of particles was exploited to separate coal particles form mineral particles. This is an exciting result since even such a basic fractionation of RCMD would be valuable to track changing conditions at the mine production face and enable rapid decision making. The second study was conducted to explore subclassification of the mineral fraction. A model was built to explore multiple particle features, including particle size, shape, color, texture, and optical properties. However, a simple stepwise method that uses birefringence for separating coal particles first and then classifying silica particles proved most effective. One particular challenge to the silica classification was determined to be the particle loading density. Future work to further enhance the output of the algorithm and next steps were depicted. This thesis research demonstrated that OM and image processing can be used to separate mineral and coal fractions. Subclassification of silica and other minerals using optical properties such as birefringence of particles alone was successful, but showed less accuracy. A robust sampling method that accounts for particle loading density and a more complex model with additional differentiating features might enhance the results. This approach should be considered as a potential candidate for the development of new RCMD monitoring technologies. This tool could enable better tracking of dust conditions and thus better decision-making regarding ventilation, dust controls, and operator position to reduce exposure hazards. / M.S. / Inhalation of fine particles in underground coal environments can lead to chronic lung diseases, such as coal worker’s pneumoconiosis or progressive massive fibrosis (PMF), which is the most severe form of disease. During the last two decades, the rates of reported cases of PMF in underground coal miners have more than doubled. Many authors have suggested different reasons to explain this trend, including the extraction of thinner coal deposits, mining techniques, changes in mineral content, and the use of high-powered cutting equipment. However, detailed information of specific dust constituents and monitoring the variability of dust concentrations during work shifts are needed to determine possible dust sources and comprehend the more recent changing disease patterns. A dust-monitoring system that provides accurate and timely data on specific respirable coal mine dust (RCMD) constituents would enable the deployment of effective control strategies to mitigate exposure to respirable hazards. Optical microscopy (OM) has been used for a long time to analyze and identify dust particles. More recent advances in portable microscopy have allowed the microscope analysis to be implemented in the field. On the other hand, automated image processing techniques are rapidly progressing and powerful imaging hardware has become a reality in handy small devices. OM and image processing technologies offer a path for near real-time applications that have not been explored for RCMD monitoring yet. In this work, a novel monitoring concept is explored using OM and image processing to classify RCMD particles. Images from dust samples captured with a polarizing microscope were used to build a classification model based on optical properties. The method herein described showed outstanding accuracy for separating coal and mineral fractions. Additionally, the Identification of silica particles in the mineral fraction was investigated and has proved more challenging. A particular finding suggests that particle loading density in the images plays an important role in classification accuracy.