the diagnosis of chest disease depends in most cases on the complex grouping of clinical data and images. According to this complexity, the debate is increased between researchers and doctors about the efficient and accurate method for chest disease prediction. The purpose of this research is to enhance the first handling of the patient data to get a prior diagnosis of the disease. The main problem in such diagnosis is the quality and quantity of the images.In this paper such problem is solved by utilizing some methods of preprocessing such as augmentation and segmentation. In addition are experimenting different machine learning techniques for feature selection and classification.The experiments have been conducted on three different data sets. As the results showed, the recognition accuracy using SVM algorithm in the classification stage, the VGG16 model for feature extraction, and LDA for dimension reduction is 67% without using image pre-processing techniques, by applying pre-processing the accuracy increased to 89%. Using a two-layer NN the recognition accuracy is 69.3%. For the same model, the accuracy has increased with the addition of image pre-processing techniques to reach 96%.