Deep Learning based Automatic Radiology Report Generation

The diagnostic x-ray examination is carried out using the chest x-ray. It is the responsibility of the radiologist to analyze the x-rays and draw conclusions from them to prescribe the proper care. Obtaining comprehensive medical reports from these x-rays is frequently time-consuming. Images of the heart, lungs, airways, spine, and chest bones can be seen in a chest x-ray. A radiologist may see thousands of x-ray images in populous nations. The goal of this project is to present a collection of the best deep-learning techniques for producing medical reports from X-ray images automatically. Deep learning algorithms have been used with models to handle this difficult task and produce correct results. Therefore, a lot of work and time can be saved if a properly trained deep learning model can generate these medical reports automatically. In this research, the text report is produced using an encoder and decoder with an attention model, while the image features are obtained using a pretrained CheXnet model. The BLEU score is used to evaluate the resulting text report.