N-Crop Based Image Division in Deep Learning with Medical Image

Abstract

In this paper, we propose an image division technique that can solve the problem of resolution reduction due to model structure and the lack of data caused by the characteristic of medical images. To verify this technique, we compared the performance of traditional full image learning and divided image learning. As a result, it is confirmed that the image division technique can proceed X-ray image deep learning more stable and is effective in predicting tuberculosis detection with higher accuracy.