The Effectiveness of the Bone Suppression Algorithm and Concentration Conversion in the Detection of Lung Tumors

Early detection of lung tumors by reading X-ray images is important for radical treatment of the disease. To decrease the risk of missed lung tumors, diagnosis support system that can provide the accurate detection of lung tumors are in high demand, and the use of artificial intelligence with deep learning is one of the promising solutions. In our research, we try to improve the accuracy of lung tumors detection by artificial intelligence using our bone suppression algorithm and the proposed concentration conversion. 604 chest X-ray images were processed using both processes and learned with a U-net deep learning model. The result showed that the combined processes successfully improved the performance of the deep learning model to identify the location of tumors (Intersection over Union) from 0.197(original images without any process) to 0.247. The combination of bone suppression and concentration conversion would be useful to improve the accuracy.