التعرف التلقائي على سرطان الرئة في صور التصوير المقطعي المحوسب للصدر

Automatic Lung Cancer Classification in CT-Scan Images Using CNN

The use of computed tomography (CT) with low doses is an important technique for the early detection of lung cancer. However, the errors accompanying the diagnosis are due to the workers’ lack of experience or the inefficiency of the tools and equipment used. In addition, the high costs and the increase in the patient’s exposure to radiation make it an important matter that researchers study to reduce the costs and health damages.
Recently, artificial intelligence has become a focus of great interest for researchers, especially in the medical field. Modern technology based on artificial intelligence automates tasks and provides medical professionals with accurate information based on deep learning algorithms. According to several studies, using neural networks to determine whether lung nodules are cancerous has become possible. This allows the establishment of an accurate diagnostic setting for malignancies. In the medical area, the deep learning technique has successfully classified difficult images.
In this study, we proposed to create a convolutional neural network (CNN) model, to form an efficient system based on deep learning to predict malignant tumors in lung nodules. A dataset to train the model was taken from the Kaggle platform. The model generated predicts individual lung cancer case transmission data, which forms the basis for personalized screening advice. The proposed model was trained on the dataset to detect three categories of diseases (malignant, benign, and normal). An Accuracy of 98.63% has been achieved, which is very satisfactory compared to previous studies.
The utilized platform in this study used an Hp ProBook laptop with an Intel Core i7-6600U CPU vpro clocked at 2.60 GHz with a boost up to 2.81 GHz, a 500 GB SSD drive, 16GB of RAM, and UHD graphics. It also had Windows 10 Pro, a 64-bit operating system.