Texture Features Extraction Of Chest HRCT Image Based On Granular Computing

Chronic obstructive pulmonary disease (COPD) is a group of chronic lung disease caused by chronic bronchitis and emphysema. According to statistics, the number of COPD patients is 32 million approximately in our country, about one million people died of this disease every year. Small airways disease is early COPD and reversible lesions, also is the main reason leading to airflow obstruction of COPD. So, diagnosis and treatment small airways diseases promptly is an important means to prevent them developing into COPD.High Resolution Computerized Tomography (HRCT) image is provided by Mufti-slice Spiral Computerized Tomography (MSCT). It is the best imaging technology to observe small airway disease currently. However, while massive HRCT data provides more detailed and accurate diagnostic information for radiologists, the characteristic of many tissues and wide gray distributions cause to difficulty of texture analysis and tissue segmentation. Currently, radiologists mostly make subjective judgment by experience, mainly focused on qualitative analysis of lesion, it is difficult to make accurate analysis, and quantitative research of the lesion's extent is less.To research quantitative analysis of chest HRCT images, assisting radiologists to diagnose small airway disease, the problem first of all we must solve is texture features extraction and accurate segmentation of chest HRCT image. With the continuous research of many scholars and engineer on the image engineering, the intelligent control theory which representing by genetic algorithm, fuzzy set and granular computing is applying to the medical image processing and supplying new ways and methods to solve the realistic engineering problems. Likewise, granular computing theory is trying to apply to texture features extraction and accurate segmentation of chest HRCT image.In this paper, by using HRCT data as research materials, granular computing as theoretical foundation and quantitative analysis as research objectives, we have made an in-depth study on the status quo of this field at home and abroad. The works of this dissertation are as follows:First, the characteristics of HRCT image are analyzed, HRCT signs of small airway disease and the related texture feature parameters are introduced, the tolerance granular space model of granular computing theory is researched. According to the relationship between chest HRCT image and the tolerance granular space model, a tolerance granular space model of chest HRCT image is built.Then, we study the texture features analysis and image segmentation method, introduce texture feature extraction into the HRCT analysis which is the foundation of HRCT quantitative analysis. It can provide effective methods and data analysis for the accurate diagnosis of small airway disease. For the complex tissue texture features, basing on traditional medicine image segmentation method and combing with granular computing theory, this paper presented a segmentation algorithm based on the tolerance granular space model. This algorithm uses the average gray (mean) value of region of interest (ROI) to select seed points automatically, improves the manual selection of original algorithm, and does not need to repeatedly adjust the threshold parameters. According to tolerance relation system (TR) the criteria of growth is improved, so this algorithm is more suitable for HRCT segmentation.Finally we realize texture parameter extraction of any straight line and region of interest (ROI), segment the lung tissue of chest HRCT and calculate area of lung tissue accurately by a large number of experiments and evaluations of the results. The results of experiments illustrate that we can extract texture parameter effectively. It is the more pertinence and practicality than classical texture analysis methods. And it can gain the data needed for diagnosis of small airways disease, provide the powerful data protection for further diagnosis.