A Diagnostic Study Of Lung Cancer Based On Chaotic Fractal Dimension And Computer Vision

AimsLung cancer is the leading cause of cancer deaths throughout the world. Through the global efforts including government agencies, academic and the pharmaceutical and diagnostic industries, substantial advances in the field of lung cancer treatments have been achieved over the past decades ranging from diagnosis and research enhanced therapies. However, the development of diagnosis and treatments are still lagging.70% of lung cancer patients to remove the tumors by surgery are impossible. As a result, lung cancer diagnosis at the earliest stage is necessary to increase the survival rates.Chaos theory is the quantitative study of unstable aperiodic behaviors in deterministic nonlinear dynamical systems.The main focus of their work has been to develop accurate and effective medical imaging. The techniques of medical image analysis have been rapidly developed and widely adopted in the field of medicine. These medical procedures seek to reveal, diagnose and examine various diseases.In this thesis, we aimed to improve the accuracy of diagnosis of the diseases through imaging studies, chaos theory and fractal dimension.Methods1. They are 86 lung cancer cases confirmed by pathology from the first affiliated hospital of Jinan University were selected along with the CT image data obtained under high resolution. Among them,77 cases of lung nodules, including 30 cases of adenocarcinoma,32 cases of squamous cell carcinoma,8 cases of small cell carcinoma,5 cases of large cell carcinoma and 2 cases of pulmonary carcinoma.9 of all the cases were non-cancer nodule, such as tuberculosis and so on. The number of male patients was 55, and that of female was 31. And the max age was 79, the min age was 31, the statistical value of age was 58.76±11.82.2. Analyze the medical imaging and input the original data into the MATLAB, and then obtain the useful information. 3. To calculate the value of fractal dimension using the box counting.4. To study the cancerous and non-cancerous nodule using the characteristics of min-valley-sum and max-vallery-sum values, MaxRho, eul number extracted from the CT image.5. The diagnostic accuray of a test to discriminate cancerous nodule from non-cancerous nodule is evaluated using Receiver Operating Characteristic (ROC) curve analysis.Results1. The different size of box was used to calculate the value of fractal dimension, and the data indicated the values of DF were different. And, the DF1, DF2, DF3, and DF4 were 1.82±0.140,1.78±0.137,1.70±0.138, and 1.64±0.140, respectively. The DF4 could be used to diagnose the lung cancer, the data has statistical significance (t=2.875, P=0.005)2. The data of MaxRho indicated that the MaxRho value of cancerous nodule was 141.79±8.332 and that of non-cancerous nodule was 86.6±15.320, The data has statistical significance (t=2.329, P=0.02).3. The characteristics of max-valley-sum and min-valley-sum can be used to diagnose cancerous and non-cancerous nodule. The max-valley-sum of cancerous nodule and non-cancerous nodule was 274.54±28.725,93.60±27.968, respectively, the data has statistical significance(t=4.513, P=0.000). The min-valley-sum of cancerous nodule and non-cancerous nodule was 326.83±30.281,125.50±30.135, respectively, the data has statistical significance (t=4.713, P=0.000).4. The number of eul of cancerous and non-cancerous nodule was 9.96±1.471,4.00±0.000, respectively, the data has statistical significance (t=4.053, P=0.000)Conclusions1. Chaos fractal dimension can be used as powerful evidence to diagnose lung cancer in spite of the instability of fractal dimension.2. The characteristics of computer vision, MaxRho, Max-valley-sum, min-valley-sum, and eul number can be used to distinguish lung cancers from the non-cancer nodules.The values of Max-valley-sum and min-valley-sum can be used to discriminate between the adenocarcinoma and squamous carcinoma when t=1.989, P=0.053 and t=1.946, P=0.058.