Cardiomegaly is one of the clinical findings suggesting severe cardiovascular disease, and early detection is very important because it helps in the diagnosis of serious diseases that prevent or cause the progression to severe diseases. Currently, convolutional neural networks are showing excellent performance in application fields such as image recognition and classification, and are showing high performance in the medical imaging field as they are used for disease diagnosis and classification. In this thesis, we tried to propose a new convolutional neural network(CNN) model that can classify normal heart and cardiomegaly in chest X-ray images. Before proposing a new CNN model, we tried to check whether there is any problem with the classification results when deep learning studies are performed using images with different bit numbers from medical images. In this thesis, TIFF, PNG and JPEG images converted from DICOM images are used to evaluate the effect of input image format on accuracy. Based on the experimental results, we propose a new deep learning model based on CNN. The training and test data used in this thesis were chest X-ray images of patients diagnosed with normal heart and cardiomegaly. Using CNN models, VGGNet, ResNet, InceptionNet(GoogleNet), DenseNet, and EfficientNet, we compared the results according to the input image type of images converted from DICOM images to TIFF, PNG, and JPEG images. The classification performance was confirmed by proposing a CNN model that can classify normal heart and cardiomegaly from chest X-ray images. In this thesis, through medical image research using deep learning, it was confirmed that the classification performance was not affected even if the DICOM image was converted into any format and used as an input image. In addition, the proposed CNN model showed excellent performance in distinguishing between normal heart and cardiomegaly in chest X-ray images. This will help to provide information according to the input image type to various studies that want to apply deep learning to medical images, and it is also thought to be helpful in selecting image types and learning parameters. The proposed model is expected to show useful results for classification of normal heart and cardiomegaly in chest X-ray images.