The Peripheral Non-small Cell Lung Cancer Subtype Diagnosis Model Research Based On CT Image

One of the most common malignant tumour is lung cancer. For decades, its incidence rate and death rate appears up-going. Early time diagnose of the lung cancer's pathological type could improve patients'treatment effect. Recently, CT-scanning as one important lung cancer diagnose measure, due to the development of pattern recognition, machine learning and image process technology, makes the computer aid diagnose to lung disease comes true. Before doctor's diagnose, using computer diagnose model to perform lung cancer classification, could decrease the error diagnose rage, then, improve the lung cancer diagnose accuracy.Firstly, we learn the Logistic theory in details, then, construct the model selection method and evaluation measures. Furthermore, we bring the CT-image feature extraction of periphery non-small cells base on Logistic regression analysis comes true. Furthermore, we select 4 features including:fringe, halo sign, holes and paraseptal emphysema as the feature input of model subtype of glandular cell cancer and squamous carcinoma.Secondly, in this paper, we learn the theory and method of the statistical Fisher discriminatory analysis carefully. Then considering the precautions when using Fisher discriminatory analysis, and established the periphery non-small cells subtype model base on the Fisher discriminatory analysis. At last, by using this model we construct model subtype experiments of glandular cell cancer and squamous carcinoma to 120 cases, the subtype accuracy of this model reached 72.5%.Finally, learn the theory and method of the Artificial Neural Network theory and the Support Vector Machine technology and its utility in medical domain, we constructed a periphery non-small cells subtype model base on the Artificial Neural Network and the Support Vector Machine technology. In the experiments of model subtype of glandular cell cancer and squamous carcinoma, the very model's subtype accuracy reached 70% and 80%, the Support Vector Machine technology higher than the Fisher discriminatory model and the Artificial Neural Network subtype model. That means under limit number of samples, the outcome of the Support Vector Machine subtype model is the best. Therefore, the very subtype model provide a reference of machine learning to aid doctor's lung cancer subtype diagnose, and also with significant means to improve the doctor's pathological diagnose accuracy.