Abstract: Abstract: To address the issues of low information utilization and insufficient prediction accuracy in the efficacy prediction of preoperative neoadjuvant therapy for rectal cancer, a radiomics-based method for predicting efficacy was proposed. First, radiomics features were extracted from different regions of interest (ROIs) in multi-modal magnetic resonance imaging (MRI), followed by multi-modal feature fusion. Then, feature selection and dimensionality reduction were performed using methods such as the t-test. Finally, a machine learning model was constructed, and grid search was employed to optimize model parameters for predicting tumor regression grade (TRG) and pathological complete response (pCR) after total neoadjuvant therapy. Experimental results show that the proposed method improves accuracy by 9.41 percentage points compared to single-modal radiomics-based approaches and by 15.41 percentage points compared to deep learning methods, demonstrating significant advantages. These findings validate the effectiveness and superiority of the proposed method in predicting rectal cancer treatment outcomes.

Key words: rectal cancer, efficacy prediction, radiomics, machine learning, AdaBoost, multimodal fusion

摘要： 摘 要: 针对术前直肠癌全程新辅助治疗疗效预测中信息利用率低、预测精度不足的问题，本文提出了一种基于影像组学的直肠癌疗效预测方法。首先，从多模态磁共振成像（MRI）中提取多个感兴趣区域的影像组学特征，并对其进行多模态融合；其次，采用T检验等特征筛选方法进行特征筛选与降维处理；最后，构建机器学习模型，并通过网格搜索优化模型参数，预测患者接受全程新辅助治疗后的肿瘤退缩分级和病理完全缓解。实验结果表明，与单模态影像组学方法相比，本文方法的准确率提升了9.41个百分点；相较于深度学习方法，准确率提升了15.41个百分点，展现出显著优势，从而验证了该方法在直肠癌疗效预测中的有效性与优越性。

关键词: 直肠癌, 疗效预测, 影像组学, 机器学习, AdaBoost, 多模态融合