基于压缩感知的融合高维多组学乳腺癌预后分析方法研究

乳腺癌预后涉及多组学数据中多个指标的综合评估,是乳腺癌治疗决策中至关重要的问题。生物多组学数据的规模和复杂性日益剧增，且数据通常呈现较高的稀疏性，传统的特征选择算法和图像重构算法无法满足日益增长的海量数据在临床诊断中的高效处理需求。本项目拟设计(1)基于压缩感知的乳腺癌预后特征选择算法;(2)基于低秩矩阵补全的乳腺癌核磁共振影像重构算法。通过利用稀疏重构技术从大量特征中选择出与乳腺癌预后相关性最高的特征，解决乳腺癌多组学数据高维度的特征冗余问题，提高乳腺癌预后的预测效果。同时利用乳腺癌MRI图像矩阵的低秩性质，通过低秩矩阵填充技术修复质量不佳的图像和填补缺失数据，提高图像的清晰度和准确性，使重构后的乳腺癌图像能够更全面地反映患者乳腺的组织结构和病变信息。本项目预期能解决乳腺癌预后预测问题中的特征冗余和图像缺失等问题，提高乳腺癌预后预测的准确率，具有重要的理论意义和应用价值。

Breast cancer prognosis, involves the comprehensive evaluation of multiple indicators in multi-omics data, is a crucial issue in breast cancer treatment decision-making. The scale and complexity of bio-multi-omics data are increasing dramatically, and the data usually show high sparsity, hence, the traditional feature selection algorithms and image reconstruction algorithms cannot meet the demand for efficient processing of the increasingly massive data in clinical diagnosis. In this project, we propose to design (1) a compressed sensing based prognostic feature selection algorithm for breast cancer; (2) a low-rank matrix complementation-based reconstruction algorithm for breast cancer MRI images complementation. By using sparse reconstruction technology to select the features with the highest correlation with breast cancer prognosis from many features, we can solve the feature redundancy problem of the high dimensionality of breast cancer multi-omics data and improve the prediction effect of breast cancer prognosis. Meanwhile, by using the low-rank property of the breast cancer MRI image matrix, the low-rank matrix complementation technique is used to repair poor-quality images and fill in the missing data, which improves the clarity and accuracy of the images so that the reconstructed breast cancer images can comprehensively reflect the histological structure of the patient's breast and lesion information. This project is expected to solve the problems of feature redundancy and missing images in the breast cancer prognosis prediction problem and improve the accuracy of breast cancer prognosis prediction, which is of great theoretical significance and application value.