基于深度森林的CT图像结直肠息肉检测研究

Detection of Colorectal Polyps on CT Images Based on Deep Forest

为提高在CT图像中结直肠息肉的筛选效率,提出一种基于深度森林的结直肠息肉CT图像检测方法。通过灰度化、归一化、中值滤波、随机旋转的手段对数据集进行预处理,将处理后的数据输入一个调整后的深度森林进行预测分类,得到输出结果。实验结果表明,该模型与其他分类算法采用不同指标对比后,具有较好的分类效果,分类精度达到了99.67%,同时该模型具有较少的超参数,泛化能力强,有助于在医学影像领域辅助医生筛查疾病患者。

To improve the screening efficiency of colorectal polyps in CT images,the CT image detection method for colorectal polyps is proposed based on deep forest.The data set is first preprocessed by means of grayscale,normalization,median filtering,and random rotation,and the processed data are then entered into an adjusted deep forest for predictive classification to obtain the output.The experiment results show that the model has a better classification effect compared with other classification algorithms under different indicators,and the classification accuracy reaches 99.67%,meanwhile,the model has fewer hyper parameters and strong generalization ability,which is helpful for assisting physicians to screen patients in the field of medical imaging.