A segmentation-based sequence residual attention model for KRAS gene mutation status prediction in colorectal cancer—— 基于分割序列的残差注意力模型预测结直肠癌KRAS基因突变