DATA-CENTRIC DEEP LEARNING METHOD FOR PULMONARY NODULE DETECTION

Lung cancer is one of the most serious cancer-related diseases in Vietnam and all over world. Early detection lung nodules can help to increase survival rate patients. Computer-aided diagnosis (CAD) systems are proposed literature for early nodules. However, current CAD based on building high-quality machine learning models a fixed dataset rather than taking into account properties which very important diagnosis. In this paper, we follow direction data-centric approach nodule by proposing method improve performance CT scans. Our takes dataset-specific features (nodule sizes aspect ratios) train as well add more training data from local Vietnamese hospital. We experiment our three widely used object networks (Faster R-CNN, YOLOv3 RetinaNet). The experimental results show that improves sensitivity these up 4.24%.