Yadav Anupam
Department of Mathematics, National Institute of Technology Uttarakhand, Srinagar-246174, Uttarakhand India.
Department of Mathematics Dr. B.R. Ambedkar, National Institute of Technology Jalandhar, Jalandhar-144011, Punjab, India.
Comput Methods Programs Biomed. 2021 Feb;199:105893. doi: 10.1016/j.cmpb.2020.105893. Epub 2020 Dec 5.
The automatic detection and counting of white blood cells (WBCs) play a vital role in the diagnosis of hematological diseases. Computer-aided methods are prevalent in the detection of WBCs because the manual process involves several complexities. In this article, a complete automatic detection algorithm to recognize the WBCs embedded in cluttered and complicated smear images of blood is designed.
The proposed algorithm uses the ellipse detection approach to approximate the presence of WBCs in the Blood. A newly designed artificial electric field algorithm with novel velocity and position bound (AEFA-C) is employed for this purpose. The problem of detection of WBCs is transformed into an optimization problem where the random candidate solutions (ellipses) are efficiently mapped. These candidate ellipses are mapped onto the edge map of the smear image, and a complete mapping is obtained using the AEFA-C algorithm.
The effectiveness of the AEFA-C based detector is tested over the 60 smear images of the blood, having all the five types of WBCs or leukocytes. The developed algorithm obtained an overall detection accuracy of 96.90%. Further, the robustness test is performed on the same dataset which justifies that the technique can handle the different noises with the detection accuracy of 90.33%. Also, the comparative study of the proposed detection algorithm with the state-of-art detection algorithms is carried out.
The experimental results demonstrate the efficiency of the proposed scheme for the detection of the WBCs in terms of detection accuracy, stability, and robustness and its outperformance over the state-of-art algorithms.
白细胞（WBC）的自动检测与计数在血液系统疾病的诊断中起着至关重要的作用。由于手工操作过程存在诸多复杂性，计算机辅助方法在白细胞检测中较为普遍。本文设计了一种完整的自动检测算法，用于识别血液涂片图像中杂乱复杂背景下的白细胞。
所提出的算法采用椭圆检测方法来估计血液中白细胞的存在情况。为此采用了一种新设计的具有新颖速度和位置约束的人工电场算法（AEFA-C）。白细胞检测问题被转化为一个优化问题，其中随机候选解（椭圆）被有效地映射。这些候选椭圆被映射到涂片图像的边缘图上，并使用AEFA-C算法获得完整的映射。
基于AEFA-C的检测器在包含所有五种类型白细胞或白血球的60张血液涂片图像上进行了测试。所开发的算法获得了96.90%的总体检测准确率。此外，在同一数据集上进行了鲁棒性测试，结果表明该技术能够以90.33%的检测准确率处理不同噪声。同时，还将所提出的检测算法与现有检测算法进行了对比研究。
实验结果证明了所提方案在白细胞检测方面在检测准确率、稳定性和鲁棒性方面的有效性，以及其相对于现有算法的优越性。
