Serial filter-wrapper feature selection method with elite guided mutation strategy on cancer gene expression data
Yu-Wei Song, Jie‐Sheng Wang, Yuliang Qi, Yucai Wang, Hao-Ming Song, Yi‐Peng Shang‐Guan
Abstract
Medical diagnosis is an important link in the field of health care, and they are essential for the prevention, diagnosis and treatment of diseases. Medical diagnosis is the premise and basis of disease treatment, which involves the identification of disease symptoms, the analysis of causes and the judgment of disease conditions. Among them, medical image segmentation is a key technology in medical diagnosis and clinical application, which involves the identification and isolation of areas of interest from medical image data, such as organs, tumors or other pathological structures. This process is critical for diagnosis of the disease, development of treatment plans, surgical navigation and evaluation of efficacy. However, medical image segmentation faces many challenges, including the high cost of data annotation, the complexity of images, the diversity of imaging techniques and the need for high-precision segmentation. In recent years, the development of deep learning technology has greatly promoted the progress in this field. For example, a framework called SimCVD is proposed in Ref. (You et al. 2022a ) that distills ‘‘Boundary Aware’’ knowledge in a shared potential space by contrast learning and utilizes structured distillation to improve segmentation accuracy. Ref. (Ren et al. 2024 ) describes a simple triple perspective unsupervised representation learning model (SimTrip), which combines a triple perspective architecture and a loss function to efficiently learn meaningful intrinsic knowledge from small batches of unlabeled data. In addition, Ref. (You et al. 2022b ) proposed a new adversarial Transformer model named CASTformer for 2D medical image segmentation. The model combines a pyramid structure with a class-aware Transformer module, as well as adversarial training strategies to improve segmentation accuracy. Ref. (You et al. 2024 ) reconsiders the problem of semi-supervised medical image segmentation from the perspective of variance reduction.