首页|改进卷积神经网络的医学图像感兴趣区域识别

改进卷积神经网络的医学图像感兴趣区域识别

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
图像中的噪声会提高图像特征信息提取难度，影响图像识别时的细节保留效果，为此提出改进卷积神经网络的医学图像感兴趣区域识别方法。分析医学图像主要噪声来源，构建噪声模型，利用非局部均值滤波算法计算图像全部像素的加权平均值，完成图像去噪处理;通过图像求反、对比度增加和灰度调节等操作增强图像细节信息;利用局部区域特征提取方法获取图像基础纹理特征，包括灰度、平滑度与熵值等;建立具有卷积层、池化层、全连接层的卷积神经网络模型，引入区域建议网络对其改进，通过该网络确定识别的候选区域，将图像特征作为网络输入，经过不断学习迭代，输出最终感兴趣区域。实验结果表明，所提方法在提高图像质量的基础上，识别出的感兴趣区域较为完整，包含的有用信息更多。
Medical Image Region of Interest Recognition Based on Improved Convolutional Neural Network
The noise in the image will increase the difficulty of extracting image features and affect the detail re-tention effect during image recognition.Therefore,a method of recognizing the region of interest in medical images was proposed based on an improved convolutional neural network.First of all,we analyzed the main noise source of medi-cal images,and then built a noise model.After that,we used the non-local mean filtering algorithm to calculate the weighted mean of all pixels,thus completing the image denoising.Moreover,we used image inversion,contrast in-crease,and grayscale adjustment to enhance image detail information.Furthermore,we use local feature extraction to obtain basic texture features,including grayscale,smoothness and entropy.Meanwhile,we constructed a convolutional neural network model including convolution layer,pooling layer and full connection layer,and then improved by the region proposal network.In addition,we determined the candidate regions through the network.Finally,we took the image features as network input.After continuous learning and iteration,the final region of interest was output.Exper-imental results prove that the proposed method can identify more complete regions of interest including more useful in-formation while improving the image quality.

Convolutional neural networkRegion proposal networkMedical imageRecognition for region of in-terestDenoising

肖衡、潘玉霞

展开 >

三亚学院信息与智能工程学院,海南 三亚 572022

武汉理工大学计算机与人工智能学院,湖北 武汉 430070

卷积神经网络 区域建议网络 医学图像 感兴趣区域识别 去噪处理

海南省重点研发计划海南省自然科学基金三亚市高等学校及医疗机构专项科技项目海南省高等学校科研项目三亚学院校级项目

ZDYF2021SHFZ240621QN09002021GXYL58Hnky2021-52USYYB22-19

2024

计算机仿真
中国航天科工集团公司第十七研究所

计算机仿真

CSTPCD
影响因子：0.518
ISSN：1006-9348
年,卷(期)：2024.41(3)
  • 15