CRADI_x-ray_classification

Propose of project

Background

X-ray chest film is listed as one of the routine physical examination items.X-ray photography, which is fast, simple and economical, has become the preferred choice for chest examination.
X-ray chest film can clearly record the general pathological changes of the lung, such as pulmonary inflammation, mass, tuberculosis, etc. X-ray photography can be used to observe the lesions in the parts with little difference in thickness and density by using different densities of human tissues. Compared with chest fluoroscopy, X-ray image is more clear, and can find subtle lesions; the objective record of image data is conducive to the review and comparison of disease diagnosis and treatment; the radiation dose of patients receiving fluoroscopy is also relatively larger.

Propose
In our study, we create a dataset with more than 70000 chest images, processed radiology report and extracted abnomal findings on the chest xray. A CNN based classification model was trained based on the training dataset. The model was then applied to different hospitals for external evaluation. According to the chest X-ray images, a variety of examination findings were detected, and the heatmap of the corresponding category of concerned areas was generated.