DEEP LEARNING SURVIVAL PREDICTION FOR LUNG CANCER PATIENTS

Abstract

Lung cancer is one of the most common cancers; lung cancer is a malignant tumor that seriously threatens the lives of patients. Improving survival prediction performance is meaningful for making the treatment plans and improving the survival rates of lung cancer patients. In this paper, an approach for predicting the survival of lung cancer patients is proposed based on pathological images. First, the deep learning method is used to automatically detect lung cancer cells in pathological pictures, and features of the detected lung cancer cells are extracted. In feature selection, an extraction method of topological features is given, it reflects the relationship and distribution characteristics of lung cancer cells, and the extracted topological features are used as predictive factors for survival analysis. In this paper, the extraction methods of global topological features are mainly studied; for example, the overall association, location relationship and distribution of cells, and the global topological features of lung cancer cells are extracted through the Voronoi diagram, Delaunay triangle, and minimum spanning tree methods. Finally, the Cox–LASSO method was used to predict the survival of lung cancer patients. Experimental results show that this method can improve the efficiency and accuracy of cell detection, and there is a higher ability to predict and analyze the survival of lung cancer patients.