B-ALL Classification of Normal vs Malignant Cells
收藏资源简介：
从构建用于血液疾病 (例如白血病) 的计算机辅助诊断工具的角度来看，通过图像处理进行细胞分类最近引起了人们的兴趣。为了对疾病诊断和进展程度做出决定性的决定，高精度地识别恶性细胞非常重要。计算机辅助工具对于自动化细胞分割和识别过程非常有帮助。很难从显微图像中鉴定出相对于正常细胞的细胞，因为两种细胞类型在形态上看起来相似。 结果，通过显微图像分析在晚期癌症阶段检测到白血病 (血癌)，不是因为在显微镜下识别这些疾病的能力，而是因为医学领域的知识，即癌细胞开始生长不受限制的方式，因此，与正常人相比，他们的人数要大得多。 但是，重要的是要进行早期疾病诊断，以更好地治愈并提高患有癌症的受试者的整体生存率。尽管可以使用诸如流式细胞术之类的先进方法，但它们非常昂贵，并且在病理实验室或医院 (尤其是在农村地区) 中无法广泛使用。另一方面，可以以更低的成本轻松部署基于计算机的解决方案。假设医学图像处理的先进方法可以导致正常细胞与恶性细胞的识别，因此可以以经济有效的方式帮助诊断癌症。 因此，这是构建自动化分类器的努力，该分类器将克服与部署具有重复试剂成本的复杂高端机器相关的问题。它还将帮助病理学家和肿瘤学家更快地进行数据驱动的推断。
From the perspective of developing computer-aided diagnostic tools for blood disorders such as leukemia, cell classification via image processing has recently garnered significant research interest. To make decisive decisions regarding disease diagnosis and disease progression, accurately identifying malignant cells is of critical importance. Computer-aided tools greatly facilitate the automated processes of cell segmentation and recognition. It is challenging to distinguish abnormal cells from normal ones in microscopic images, as the two cell types appear morphologically similar. Consequently, leukemia (blood cancer) is typically detected in advanced cancer stages through microscopic image analysis, not due to the ability to directly identify these diseases under a microscope, but rather because of the medical knowledge that cancer cells grow uncontrollably and thus exist in far greater numbers than in healthy individuals. Nonetheless, early disease diagnosis is crucial for improving treatment outcomes and overall survival rates of cancer patients. Although advanced methods such as flow cytometry are available, they are extremely expensive and not widely accessible in pathology laboratories or hospitals, especially in rural areas. Conversely, computer-based solutions can be easily deployed at a much lower cost. It is hypothesized that advanced medical image processing methods can enable differentiation between normal and malignant cells, thereby assisting in cancer diagnosis in a cost-effective manner. Accordingly, this work aims to develop an automated classifier that overcomes the issues associated with deploying complex high-end machines that incur repeated reagent costs. It will also help pathologists and oncologists perform data-driven inferences more rapidly.
