📄 中文摘要
红细胞变形是多种疾病的病理表现，其中就包括镰状细胞贫血，该疾病会导致患者反复经历疼痛发作和严重的贫血。对这类疾病患者的监测通常涉及在显微镜下观察外周血样本，这是一个耗时且需要专业人员操作的过程。此外，由于对单个红细胞的观察具有主观性，这使得诊断过程容易受到观察者经验和判断的影响，从而可能导致诊断结果的不一致性。目前，临床上迫切需要一种客观、高效且标准化的诊断辅助工具，以减轻医生工作负担，并提高诊断的准确性和一致性。利用人工智能技术，特别是深度学习和图像处理方法，可以自动化地分析红细胞形态，从而为镰状细胞贫血的诊断提供支持。通过对大量外周血图像中红细胞的形状特征进行分类，机器学习模型能够识别出正常红细胞与镰状细胞贫血患者特有的异常红细胞（如镰刀状细胞、靶形细胞等）。这种基于图像分析的诊断支持系统可以减少对人工主观判断的依赖，缩短诊断时间，并允许在资源有限的地区进行初步筛查。开发这样的系统需要构建大规模、多样化的带标签红细胞图像数据集，并设计鲁棒的图像分割和分类算法，以应对图像质量、细胞重叠和形态变异等挑战。最终目标是提供一个辅助诊断工具，能够准确识别镰状细胞贫血的生物标志物，从而改善患者管理和治疗效果。
📄 English Summary
Diagnosis Support of Sickle Cell Anemia by Classifying Red Blood Cell Shape in Peripheral Blood Images
Red blood cell (RBC) deformation is a pathological consequence of various diseases, including sickle cell anemia, which precipitates recurrent pain episodes and severe anemia. Patient monitoring for such conditions typically involves microscopic examination of peripheral blood samples, a time-consuming procedure demanding specialized expertise. Furthermore, the inherent subjectivity in observing individual RBCs introduces variability and potential inconsistencies in diagnosis, necessitating an objective, efficient, and standardized diagnostic aid. Leveraging artificial intelligence, particularly deep learning and image processing techniques, can automate the analysis of RBC morphology to support sickle cell anemia diagnosis. By classifying the shape characteristics of RBCs in numerous peripheral blood images, machine learning models can differentiate between normal RBCs and the distinctive abnormal RBCs (e.g., sickle cells, target cells) indicative of sickle cell anemia. This image-analysis-based diagnostic support system can mitigate reliance on subjective human judgment, expedite diagnosis, and enable preliminary screening in resource-limited settings. Developing such a system requires constructing large-scale, diverse, and meticulously labeled RBC image datasets, alongside designing robust image segmentation and classification algorithms capable of addressing challenges like image quality variations, cell overlapping, and morphological diversity. The ultimate objective is to deliver an assistive diagnostic tool that accurately identifies the biomarkers of sickle cell anemia, thereby enhancing patient management and treatment outcomes.