A Method of Detecting Erythrocytes and Spirochaete in Darkfield Microscopy using Attention Unet
Hyunjun Park · 산업과학기술연구 논문집 · 2021
Blood circulates throughout the whole body to provide nutrition and deliver waste products to the excretory organs, so it contains various information about living things. Recently, by analyzing microorganisms such as bacteria and viruses in the blood, it is possible to simply diagnosis whether or not a patient has cancer. Therefore, there is a need for a method that can automatically detect erythrocytes and bacteria in the blood, and it has great significance in computer vision and the medical field. Therefore, in this paper, we propose a method for detecting erythrocytes and spirochaete(one of the bacterium) in blood from darkfield microscopy obtained from Kaggle. The proposed method uses Attention Unet, which improves performance by adding an attention gate to Unet, one of the well-known deep learning models. To evaluate the proposed method, we use Unet and Attention Unet to detect erythrocytes and spirochaete and create a confusion matrix to compare accuracy, precision, sensitivity, and specificity. As a result of the experiment, it was confirmed that Attention Unet showed better detection performance than the Unet.