基于深度卷积神经网络的肝脏肿瘤检测算法研究

Research on liver tumor detection algorithm based on Deep Convolutional Neural Network

利用深度卷积神经网络对肝脏肿瘤进行检测,首先对肝脏肿瘤CT图像进行预处理,然后根据特征像素值对图像进行阈值分割,并对肿瘤区域进行标记,再使用标记好的数据集建立深度卷积神经网络模型进行训练,接着利用训练好的模型对未标记的验证数据集进行预测和验证,最后在测试数据集上测试模型的性能,根据测试结果,对肝脏肿瘤进行检测.通过对深度卷积神经网络算法、分水岭算法和连通域算法的检测结果进行比较,实验结果表明深度卷积神经网络算法在肿瘤检测方面具有最高的准确率和最大的F_(1)分数.说明深度卷积神经网络在肝脏肿瘤检测中具有卓越的性能,能够准确地识别肿瘤并减少误诊和漏诊的可能性.

Liver tumor detection is performed using deep convolutional neural networks.Initially,liver tumor CT images are preprocessed.Then,threshold segmentation is performed on the images based on feature pixel values,and the tumor regions are labeled.A DCNN model is then established using the labeled dataset for training.Subsequently,the trained model is utilized to predict and validate on an unlabeled validation dataset.Finally,the model′s performance is evaluated on a test dataset to detect liver tumors based on the test results.By comparing the detection results of the Deep Convolutional Neural Network algorithm,the watershed algorithm and the connected domain algorithm,the experimental results show that the DCNN algorithm achieves the highest accuracy and the highest F_(1) score in tumor detection.This indicates that DCNN exhibits superior performance in liver tumor detection,accurately identifying tumors and reducing the possibility of misdiagnosis and missed diagnosis.