基于文本与图像的肺疾病研究与预测EI北大核心CSCD

通过对目前现有的肺癌检测技术研究,发现大部分研究人员主要针对肺癌(Computed tomography,CT)影像进行研究,忽略了电子病历所隐藏的肺癌信息,本文提出一种基于图像与文本相结合的肺癌分类方法,从现有的基于深度学习的肺癌图像分类出发,引入了电子病历信息,使用Multi-head attention以及(Bi-directional long short-term memory,Bi-LSTM)对文本建模.实验结果证明,将电子病历信息引入到图像分类模型之后,对模型的性能有进一步的提升.相对仅使用电子病历进行预测,准确率提升了大约14%,精确率大约提升了15%,召回率提升了14%.相对仅使用肺癌CT影像来进行预测,准确率提升了3.2%,精确率提升了4%,召回率提升了4%.     

Automatic lung segmentation for large-scale medical image management

Digital medical images assist specialists in improving their diagnostic efficiency and in treating diseases. For example, the chest Computed Tomography (CT) images help in diagnosing the lung disease. The chest CT scan generates multiple images of a patient’s lung. The size of the medical imaging data has increased with the usage of medical images. In a picture archiving and communication system, large-scale medical images must be transmitted to specialists through either wired or wireless communications and retained in the archive. Hence, medical images have to be compressed, and there should be no damage to the Region of Interest (RoI) during compression. In order to protect the RoI, image segmentation is needed to detect RoI in medical images. Among the various image segmentation methods available, the method using Level-set is robust to irregular noises. However, the problems faced in using this method include manual input of the initial contour and slow performance speed. Inputting an initial contour to the Level-set that correctly fits the object’s form helps in reducing the number of repetitions. This in turn helps in improving the segmentation performance speed. However, it is difficult for a user to input an appropriate initial contour. Therefore, this paper aims at providing a method to auto-configure the initial contour in the Level-set method. Multi-resolution analysis helps in reducing the pace of the auto-configuration process of the initial contour. In addition, the volume data of a CT image is used to prevent data loss that occurs during the MRA transformation process. Studies have confirmed that the proposed method facilitates drastic improve.     

A Novel Approach based on Average Information Parameters for Investigation and Diagnosis of Lung Cancer using ANN

In this paper, an average informational parameter based approach for lung cancer detection and diagnosis has been proposed. Suggested methodology is established on average information parameters by utilizing image processing tools for lung cancer investigation. The real issue with the lung cancer diseases is the time constraint for physical diagnosis that expands the death possibilities. Henceforth essentially proposed technique is an approach that would help the medical practitioners for precise and superior decision against the lung cancer discovery. The crucial point in the proposed method is that it helps the doctors for taking a firm decision on lung cancer diagnosis. Microscopic lung images are taken for analysis and investigation by using digital image processing techniques which also recovers the quality of images that has been degraded by several reasons including random noise. The statistical parameters are implemented for lung cancer analysis. The statistical and mathematical parameters are implemented like Entropy, Standard Deviation, Mean, Variance and MSE under average information method. The statistical range of each parameter is calculated for number of iterations. The individual statistical parameter analysis with its impact on lung cancer images is carried out and finally the Artificial Neural Network is the final decision maker in lung cancer diagnosis. This paper also rejects the null hypothesis test by implementing one of the standard statistical methods.     

基于改进U-net的肺癌识别方法

目前基于深度学习的肺癌辅助诊断方法存在无法准确定位病灶的缺陷。针对该问题,在现有U-net网络结构的基础上提出一种分两步走的基于改进U-net的肺癌识别方法。利用U-net获得病灶精确位置,通过CNN分类网络对病灶进行诊断,得到原始CT图像的检测结果。实验结果表明,该方法可以对肺部病灶进行较为精确的定位,分割效果的DSC相似度指数超过80%,对肺癌病灶进行分类诊断的准确率达到90.7%。     

Analyzing CT images for detecting lung cancer by applying the computational intelligence-based optimization techniques

Lung cancer is the most critical disease because it affects both men and women. Most of the time, lung cancer leads to death due to less health care and medical attention. In addition, lung cancer is difficult to identify in earlier stages due to the low-level symptoms and risk factors. To overcome the complexity, effective techniques must predict lung cancer earlier. To attain the problem statement, an lung cancer identification system is developed with the help of a meta-heuristic algorithm. The CT imageries obtained from the CIA database are analyzed step by step. The gathered image noise is removed by applying the mean filter, and the affected regions are segmented with the help of the Butterfly Optimization Algorithm-based K-Means Clustering (BOAKMC) algorithm. Afterward, various statistical features are derived, and the Supervised Jaya Optimized Rough Set related Feature Selection (SJORSFS) process is used to select the lung features. Finally, the lung cancer is identified using Autoencoder based Recurrent Neural Network (ARNN) classification algorithm, successfully recognizing the lung cancer features. Then the system's efficiency is evaluated using a MATLAB setup; here, 3000 are treated as training images and 2043 for testing images. The effective training enhances overall lung cancer prediction accuracy by up to 99.15%.     

Classification of ground glass opacity lesion characteristic based on texture feature using lung CT image

Lung cancer causes a high mortality rate in the world than any other cancers. That can be minimised if the symptoms and cancer cells have been detected early. One of the techniques used to detect lung cancer is by computed tomography (CT) scan. CT scan images have been used in this study to identify one of the lesion characteristics named ground glass opacity (GGO). It has been used to determine the level of malignancy of the lesion. There were three phases in identifying GGO: image cropping, feature extraction using grey level co-occurrence matrices (GLCM) and classification using Naïve Bayes Classifier. In order to improve the classification results, the most significant feature was sought by feature selection using gain ratio evaluation. Based on the results obtained, the most significant features could be identified by using feature selection method used in this research. The accuracy rate increased from 83.33% to 91.67%, the sensitivity from 82.35% to 94.11% and the specificity from 84.21% to 89.47%.     

基于形态学多尺度算法的肺部CT图像边缘检测

医学图像边缘检测是医学图像处理和分析的基础,传统边缘检测算子对噪声敏感,检测到的图像边缘效果较差.本文提出了一种基于形态学多尺度算法的肺部CT图像边缘检测方法.首先对形态学边缘检测算子进行改进,然后利用形态学多尺度算法检测各尺度下的图像边缘,最后采用非均匀权值方法合成最终边缘.实验结果表明:该方法在检测出肺部图像边缘的同时能够很好地抑制噪声,是一种有效的肺部CT图像边缘检测方法.     

超级计算支撑的新冠肺炎CT影像综合分析辅助系统应用

目的 新冠肺炎（COVID-19）已经成为全球大流行疾病,在全球范围数百万人确诊。基于计算机断层扫描（computed tomography,CT）数据的影像学分析是临床诊断的重要手段。为了实现快速高效高精度地检测,提出了一种超级计算支撑的新冠肺炎CT影像综合分析辅助系统构建方法。方法 系统整个处理流程依次包括输入处理模块、预处理模块、影像学分析子系统和人工智能（artifiaial intelligence,AI）分析子系统4部分。其中影像学分析子系统通过分析肺实变、磨玻璃影和铺路石等影像学典型特征检测是否有肺炎和典型新冠肺炎特征,给出肺炎影像分析结论;AI分析子系统通过构建深度学习模型来区分普通病毒肺炎与新冠肺炎,增加肺炎的筛查甄别能力。结果 系统发布以来,持续稳定地为国内外超过三十家医院与一百多家科研机构提供了新冠肺炎辅助诊断服务和科研支撑,为抗击疫情提供重要支撑。结论 本文提出的超级计算支撑的新冠肺炎CT影像综合分析辅助系统构建方法,取得了应用效果,是一种有效实现快速部署服务、对突发疫情提供高效支撑的服务方式。     

Covid-19 Detection Using Deep Correlation-Grey Wolf Optimizer

The immediate and quick spread of the coronavirus has become a life-threatening disease around the globe. The widespread illness has dramatically changed almost all sectors, moving from offline to online, resulting in a new normal lifestyle for people. The impact of coronavirus is tremendous in the healthcare sector, which has experienced a decline in the first quarter of 2020. This pandemic has created an urge to use computer-aided diagnosis techniques for classifying the Covid-19 dataset to reduce the burden of clinical results. The current situation motivated me to choose correlation-based development called correlation-based grey wolf optimizer to perform accurate classification. A proposed multistage model helps to identify Covid from Computed Tomography (CT) scan image. The first process uses a convolutional neural network (CNN) for extracting the feature from the CT scans. The Pearson coefficient filter method is applied to remove redundant and irrelevant features. Finally, the Grey wolf optimizer is used to choose optimal features. Experimental analysis proves that this determines the optimal characteristics to detect the deadly disease. The proposed model’s accuracy is 14% higher than the krill herd and bacterial foraging optimization for severe accurate respiratory syndrome image (SARS-CoV-2 CT) dataset. The COVID CT image dataset is 22% higher than the existing krill herd and bacterial foraging optimization techniques. The proposed techniques help to increase the classification accuracy of the algorithm in most cases, which marks the stability of the stated result. Comparative analysis reveals that the proposed classification technique to predict COVID-19 with maximum accuracy of 98% outperforms other competitive approaches.     

三维多尺度交叉融合网络肺结节分类研究

计算机断层扫描影像中良、恶性肺结节的准确分类对肺癌的预防和治疗至关重要。然而,由于计算机断层扫描影像中肺结节背景的复杂性,以及良、恶性肺结节判定之间存在的不确定性,使得良恶性肺结节的准确分类成为了一项极具挑战性的工作。提出了一种深度三维多尺度交叉融合卷积神经网络实现了良恶性肺结节的精确分类。使用密集连接结构自动提取肺结节多尺度特征,为了减少特征提取过程中肺结节相关信息的丢失,对多尺度特征引入了交叉融合策略得到多尺度特征组,增强了高、低层次语义信息的表达能力,同时增强特征在网络中的传递和转移。将提取的特征组分别连接至多个softmax分类器,模拟多位经验不同医生共同决策,实现了良、恶性肺结节的精确识别。使用肺图像联合会数据集进行验证,分类准确率达到了90.96%,AUC为94.95%。     

改进卷积神经网络的COVID-19CT影像分类方法研究

针对2019年12月在中国武汉发现的新型冠状病毒,由于RT-PCR检测具有假阴性率过高且得出结果会花费大量时间等问题,研究证明计算机断层扫描（CT）已经成为了辅助诊断和治疗新型冠状病毒肺炎的重要手段之一。由于目前公开的COVID-19 CT数据集较少,提出利用条件生成对抗网络进行数据增强以获得更多样本的CT数据集,以此降低发生过拟合风险;另外提出一种基于BIN残差块的改进U-Net网络来进行图像分割,再结合多层感知器进行分类预测。通过与AlexNet、GoogleNet等网络模型进行比较,得出提出的BUF-Net网络模型性能最优,达到了93%的准确率。利用Grad-CAM技术对系统的输出进行可视化,能够更加直观地说明CT影像对于诊断COVID-19的重要作用。将深度学习技术应用到医学影像中有助于协助放射科医生获得更为有效的诊断。     

A Fourier-LDA approach for image recognition

Fourier transform and linear discrimination analysis (LDA) are two commonly used techniques of image processing and recognition. Based on them, we propose a Fourier-LDA approach (FLA) for image recognition. It selects appropriate Fourier frequency bands with favorable linear separability by using a two-dimensional separability judgment. Then it extracts two-dimensional linear discriminative features to perform the classification. Our experimental results on different image data prove that FLA obtains better classification performance than other linear discrimination methods.     

Intensity-based statistical features for classification of lungs CT scan nodules using artificial intelligence techniques

A computer-aided diagnostic (CAD) system for effective and accurate pulmonary nodule detection is required to detect the nodules at early stage. This paper proposed a novel technique to detect and classify pulmonary nodules based on statistical features for intensity values using support vector machine (SVM). The significance of the proposed technique is, it uses the nodules features in 2D & 3D and also SVM for the classification that is good to classify the nodules extracted from the image. The lung volume is extracted from Lung CT using thresholding, background removal, hole-filling and contour correction of lung lobe. The candidate nodules are extracted and pruned using the rules based on ground truth of nodules. The statistical features for intensity values are extracted from candidate nodules. The nodule data are up-samples to reduce the biasness. The classifier SVM is trained using data samples. The efficiency of proposed CAD system is tested and evaluated using Lung Image Consortium Database (LIDC) that is standard data-set used in CAD Systems for Lungs Nodule classification. The results obtained from proposed CAD system are good as compare to previous CAD systems. The sensitivity of 96.31% is achieved in the proposed CAD system.     

A Hybrid Artificial Intelligence Model for Skin Cancer Diagnosis

Melanoma or skin cancer is the most dangerous and deadliest disease. As the incidence and mortality rate of skin cancer increases worldwide, an automated skin cancer detection/classification system is required for early detection and prevention of skin cancer. In this study, a Hybrid Artificial Intelligence Model (HAIM) is designed for skin cancer classification. It uses diverse multi-directional representation systems for feature extraction and an efficient Exponentially Weighted and Heaped Multi-Layer Perceptron (EWHMLP) for the classification. Though the wavelet transform is a powerful tool for signal and image processing, it is unable to detect the intermediate dimensional structures of a medical image. Thus the proposed HAIM uses Curvelet (CurT), Contourlet (ConT) and Shearlet (SheT) transforms as feature extraction techniques. Though MLP is very flexible and well suitable for the classification problem, the learning of weights is a challenging task. Also, the optimization process does not converge, and the model may not be stable. To overcome these drawbacks, EWHMLP is developed. Results show that the combined qualities of each transform in a hybrid approach provides an accuracy of 98.33% in a multi-class approach on PH² database.     

CT图像肺及肺病变区域分割方法综述

计算机断层扫描（computed tomography,CT）技术能为新冠肺炎（corona virus disease 2019,COVID-19）和肺癌等肺部疾病的诊断与治疗提供更全面的信息,但是由于肺部疾病的类型多样且复杂,使得对肺CT图像进行高质量的肺病变区域分割成为计算机辅助诊断的重难点问题。为了对肺CT图像的肺及肺病变区域分割方法的现状进行全面研究,本文综述了近年国内外发表的相关文献：对基于区域和活动轮廓的肺CT图像传统分割方法的优缺点进行比较与总结,传统的肺CT图像分割方法因其实现原理简单且分割速度快等优点,早期使用较多,但其存在分割精度不高的缺点,目前仍有不少基于传统方法的改进策略;重点分析了基于卷积神经网络（convolutional neural network,CNN）、全卷积网络（fully convolutional network,FCN）、U-Net和生成对抗网络（generative adversarial network,GAN）的肺CT图像分割网络结构改进模型的研究进展,基于深度学习的分割方法具有分割精度高、迁移学习能力强和鲁棒性高等优点,特别是在辅助诊断COVID-19病例时,基于深度学习方法的性能明显优于基于传统方法的性能;介绍肺及肺病变区域分割的常用数据集和评价指标,在解决如COVID-19数据样本量少等问题时,使用GAN以合成高质量的对抗性图像用以扩充数据集,从而增加训练样本的数量和多样性;讨论了肺CT图像的肺及肺病变区域的高精度分割策略的研究趋势、现有挑战和未来的研究方向。     

基于柔性数学形态学的医学图像边缘提取

医学图像边缘提取,尤其是病灶部位的边缘提取,是医学图像处理中非常重要的预处理步骤,边缘提取的质量决定了图像的最终处理结果。人们一般习惯于用微分算子和梯度形态学算子提取边缘,但这类算子都不能很好地滤除噪声,也不能提取边缘细节。文章在阐述了数学形态学一般原理与方法及柔性数学形态学原理与性质的基础上,将柔性数学形态学用于左肺上叶周围型肺癌CT图像边缘提取。实验结果表明,这一方法比微分算子和形态学边缘梯度算子更能有效地滤除噪声并将肺部轮廓和肿瘤的大小与边缘准确地提取出来。     

深度学习下的医学影像分割算法综述

医学影像分割是计算机视觉在医学影像处理中的一个重要应用领域,其目标是从医学影像中分割出目标区域,为后续的疾病诊断和治疗提供有效的帮助。近年来深度学习技术在图像处理方面取得了巨大进展,基于深度学习的医学影像分割算法逐渐成为该领域研究的重点和热点。叙述了计算机视觉下的医学影像分割任务及其难点,重点综述了基于深度学习的医学影像分割算法,对当前具有代表性的相关方法进行了分类和总结,介绍了医学影像分割算法常用的评价指标和数据集。对该技术的发展进行了总结和展望。     

基于Concat-UNet的食管癌肿瘤医学影像分割研究

食管癌肿瘤的诊断方式主要是医生对胸部计算机断层扫描（CT）影像进行阅片。由于医生的主观判断易受外界环境的干扰,因此诊断结果与实际结果存在偏差。基于深度学习的图像分割网络对辅助诊断食管癌肿瘤具有重要意义。因食管在整体胸部CT影像中所占的区域较小且对比度较低,传统的图像分割网络难以准确地确定食管癌肿瘤的区域。为精准分割医学CT影像中的食管癌肿瘤,提出图像分割网络Concat-UNet。基于U-Net网络,采用编码解码模式的U型对称架构对网络中的卷积模块进行改进,并引入跳跃连接和批量归一化层,将卷积模块的原始输入与提取特征后的输出进行特征融合,以增强网络的特征提取能力。在此基础上,采用BCEWithLogits与Dice损失函数相结合的方式联合训练网络。实验结果表明,相比SegNet、ERFNet、U-Net等网络,Concat-UNet在食管癌数据集上的检测精确率为91.87%,相比基准网络U-Net提升了11.64个百分点,具有较优的分割效果。     

基于生成对抗网络的乳腺癌病理图像可疑区域标记

本论文针对乳腺癌病理图像分析提出新的方法进行图像特征提取和可疑区域标记。由于深度神经网络,例如 VGG,GoogleNet,ResNet 等,均需要大量的标注样本才能完成训练,而医疗影像图像的标记成本很高,并不能为训练复杂的网络提供足够的训练数据。本论文借鉴生成对抗网络(Generative Adversarial Network, GAN) 的思想,提出基于弱监督学习的病理图像可疑区域标记网络,首先利用少量有标记的病理图像数据来训练分类模型,即判断该图像是否是乳腺癌,然后通过融合该网络提取到的具有判别力的特征来对可疑区域进行标记。由本文提出的网络在已有的国外乳腺癌病理图像数据集上达到的平均准确率为 83.8%,比基于卷积神经网络 (Convolutional Neural Network,CNN) 的分类方法在准确率上分别高 3 个百分点,说明该网络提取到的特征具有更好的判别力,不仅能够提高分类模型的准确率,还更有助于对病理图像的可疑区域进行标记。     

肺实质CT图像细化分割

目的 由于肺部CT图像中各组织结构复杂、灰度分布不均匀,造成肺实质部分难以准确分割和提取。为了提高肺实质分割的准确率,本文提出了一种基于超像素的细化分割与模糊C均值聚类相结合的自动分割算法。方法 该算法充分利用肺部CT图像的灰度、纹理特征,同时为了正确标记超像素的分类,引入一种空间邻域信息来增强空间约束进而有效地解决灰度不均匀的问题,它能够对肺实质进行分割并除去其周围的主血管,然后利用形态学知识去除肺部的分支血管。结果 在临床患有四类疾病的患者CT图像数据集上采用改进的图像特征,使得肺实质分割的准确率提高了0.8%。同时,算法准确率提高到99.46%。结论 实验结果表明,本文算法能够实现肺部CT图像肺实质的自动细化分割,结果准确适用。该算法鲁棒性好、速度快,是一种精确有效的自动肺实质分割方法。