Towards Efficient Lesion Localization Based on Template Occlusion Strategy in Intelligent Diagnosis

摘要

In recent years, Artificial Intelligence (AI) has made great achievements in medical field, and intelligent diagnosis becomes an important topic simultaneously. Recent research on the intelligent diagnosis of diseases is mainly focus on disease identification. However, how to detect lesion location is still a difficult problem which is of great significance. In medicine, lesion localization is required by the treatment of many diseases. Lesion location information can helps physicians make a further understanding of the disease, assists them in diagnosis and therapy, and increases the likelihood of the disease being cured. In this paper, we propose an efficient lesion localization method based on template occlusion (LLTO) for locating lesion areas in disease images. First, OpenCV cascade classifier is used to train candidate boxes representation model and the TensorFlow is used to train disease discrimination model. Second, to implement the lesion location task, the lesion candidate boxes are generated by the candidate boxes representation model in the test image. Third, the disease discrimination model is used to select true lesion areas from candidate boxes and integrate them to get the image marked with lesion locations. The experiment proves the efficiency and effectiveness of our method.