White Blood Cell Classification Based on Shape and Deep Features

摘要: Classification of the white blood cells (WBCs) in smear images is essential for providing important information to physicians. In addition, manual analyzing determining various WBCs a time-consuming issue this paper, hybrid method proposed classification WBCs. Image processing (IP) and machine learning (ML) are used determine classify images. IP perspective, algorithms segment ML feature extraction employed. RGB HSV transformation, color gray tone conversion, filtering operations, thresholding, morphological processes Median adaptive histogram equalization image enhancement Otsu thresholding considered due its simplicity. Shape based features deep characterization long-short term memory (LSTM) network employed classification. A dataset containing totally 349 evaluation method. 10-fold cross validation experiments accuracy calculated accordingly. While shape produce 80.0% accuracy, obtain 82.9% accuracy. When both concatenated, 85.7% score obtained.