基于3D卷积网络和多模态MRI的脑胶质瘤自动分割

目的:基于三维（3D）卷积神经网络和多模态MRI实现脑胶质瘤的自动分割。方法:首先对来自BraTS2020公共数据集的369例脑胶质瘤的4个模态MRI数据进行3D剪裁、重采样、去伪影、归一化的预处理。其次将MRI数据和脑胶质瘤标注信息输入到基于U-net的3D卷积神经网络模型进行训练和测试。利用相似性系数、召回率和精确率评价整体肿瘤区域、核心肿瘤区和增强肿瘤区的分割结果。结果:在74例测试数据集上,整体肿瘤区域、核心肿瘤区域和增强肿瘤区域的相似系数平均值分别为0.88、0.77和0.73,中位值分别为0.90、0.84和0.81,召回率平均值分别为0.88、0.78和0.78,中位值分别为0.90、0.84和0.84,精确率平均值分别为0.89、0.83和0.75,中位值分别为0.91、0.89和0.79。结论:基于U-net的3D卷积神经网络在多模态MRI数据集上获得了较好的分割结果,显示其在脑胶质瘤自动分割方面的潜力,可为临床诊断分级和治疗策略选择提供参考。     

基于多模态特征融合的脑瘤图像分割方法

针对目前大多数医学图像分割方法难以对多模态图像进行特征融合进而完成精确分割任务的问题,提出一种基于编码器-解码器总体架构的多模态脑瘤图像特征融合策略。首先,编码阶段利用孪生网络对不同模态数据进行特征提取,孪生网络结构参数和权值共享的特性可有效减少网络参数量;其次,在进行特征提取的编码阶段加入级间融合,保留不同模态的共性特征的同时强调其互补特征;然后,在解码阶段引入密集跳跃连接思想,最大程度结合不同尺度特征图的低级细节和高级语义信息;最后,设计混合损失函数,在网络生成的预测图受真值图监督的同时让最高级特征融合图也受同倍下采样真值图的监督。所提方法在公开数据集BraTS2019上进行实验,并用图像分割常用的5种指标进行评估。在脑瘤及水肿区域分割任务中得到平均Dice系数为0.884,阳性预测率为0.870,灵敏度为0.898,豪斯多夫距离为3.917,平均交并比达到79.1%,与较先进的算法U-Net和PA-Net相比多项指标均有提升。实验结果说明,级间融合和层间跳跃连接的加入对多模态医学图像的分割效果有所提升,在医学上对脑肿瘤磁共振图像进行病变区域分割具有重要的应用价值和理论意义。     

基于3D深度残差网络和多模态MRI的脑胶质瘤自动分级

目的:利用3D深度残差网络和多模态MRI实现对脑胶质瘤的自动分级。方法:利用BraTS2020公共数据集的293例高级别胶质瘤（HGG）和76例低级别胶质瘤（LGG）的多模态MRI数据训练和测试3D深度残差卷积网络模型。多模态MRI图像经过3D剪裁、重采样和归一化的预处理,随机分组为训练（64%）、验证（16%）和测试（20%）样本,将预处理后的多模态MRI图像和分级标注输入到网络模型进行训练、验证和测试。利用准确率（ACC）和受试者工作特征（ROC）曲线下面积（AUC）评价分级结果。结果:在59例（48例HGG和11例LGG）验证数据集上,ACC和AUC分别为0.93和0.97,在75例（62例HGG和13例LGG）测试数据集上,ACC和AUC分别为0.89和0.93。结论:3D深度残差网络在多模态MRI数据集上获得了较好的脑胶质瘤自动分级结果,可以为确定治疗方案和预测预后方面提供重要参考。     

基于条件生成对抗网络的三维肝脏及肿瘤区域自动分割

肝脏计算机断层扫描成像(CT)的三维(3D)肝脏和肿瘤分割对于辅助医生的诊断及预后具有非常重要的临床价值。为了准确快速地分割肝脏及肿瘤区域,本文提出了一种基于条件生成对抗网络(cGAN)的肿瘤3D条件生成对抗分割网络(T3scGAN),同时采用了一个由粗到细的3D自动分割框架对肝脏及肿瘤区域实施精准分割。本文采用2017年肝脏和肿瘤分割挑战赛(LiTS)公开数据集中的130个病例进行训练、验证和测试T3scGAN模型。最终3D肝脏区域分割的验证集和测试集的平均戴斯(Dice)系数分别为0.963和0.961,而3D肿瘤区域分割的验证集和测试集的平均Dice系数分别为0.819和0.796。实验结果表明,提出的T3scGAN模型能够有效地分割3D肝脏及其肿瘤区域,因此能够更好地辅助医生进行肝脏肿瘤的精准诊断和治疗。     

基于TensorMixup的脑胶质瘤全自动分割

脑胶质瘤及其子区域的全自动分割对临床脑胶质瘤患者的诊断、治疗与病情监控具有重要意义。本文改进传统Mixup方法,提出TensorMixup模型,并将其应用于三维U-Net脑胶质瘤分割任务。算法核心思想包括,首先从两位患者相同模态的核磁共振脑影像中分别获取肿瘤区域所在边界框的图像序列,并从获取的图像序列中选取尺寸为128×128×128体素的图像块,然后使用一个所有元素均独立采样于贝塔分布的张量,混合图像块的信息,接着将上述张量映射为矩阵,用于混合图像块的独热编码标签序列,从而合成新图像及其标注数据,最后使用合成数据训练模型,以提高模型的分割精度。在BraTs2019数据集的测试结果显示,本文算法在完整肿瘤、肿瘤核心与增强肿瘤区域的平均Dice值依次可达91.32%、85.67%与82.20%,证明使用TensorMixup进行脑胶质瘤分割,具有可行性与有效性。     

脑肿瘤分割算法研究

由于医学成像技术的复杂性和胶质瘤表面高度的异质性,对人脑神经胶质瘤的图像分割是医学影像分析中最具挑战性的任务之一。本研究对UNet++医学图像分割网络进行了改进,改进后的网络能够融合全尺度下的粗粒度语义和细粒度语义信息。在公开的BraTS脑肿瘤分割数据集中的335例图像上进行分割实验,采用2D与3D对比分割实验综合评价改进后网络的分割性能,将分割结果与UNet,UNet++,UNet3+医学图像分割网络的结果进行对比。在Dice相似度系数(DSC),95%Hausdorff表面距离(HSD95),Sensitivity,PPV等4个指标基础上,2D对比分割实现的指标均值分别为:83.70%,1.7,88.40%,84.96%;3D对比分割实验的指标均值为:90.79%,0.242,91.23%,91.06%。实验结果表明,改进的算法使神经胶质瘤的分割结果与金标准在区域上有更多的重叠,可以更好的完成脑胶质瘤的分割。在临床应用中,可望帮助神经外科医生高效地分离脑肿瘤与人脑周围组织,从而实现快速的计算机诊疗。     

基于层次聚类的多模态磁共振脑肿瘤图像的自动分割方法

多模态磁共振脑肿瘤图像分割是医学图像应用的基础,在病理分析与手术引导等领域具有广泛的应用价值,为提升分割效率及精度,研究基于层次聚类的多模态磁共振脑肿瘤图像自动分割方法.利用Tamura的特征提取方法,以粗糙度与对比度为特征提取的定量分析指标,提取多模态磁共振脑肿瘤图像的纹理特征构建数据集;通过融合稀疏代表点的亲和传播...     

基于联合深度网络和形态结构约束的三维医学图像分割方法

医学图像自动分割具有广泛和重要临床应用价值,特别是病灶、脏器的自动分割。基于传统图像处理方法的医学图像分割仅能利用浅层结构模型的浅层特征来识别感兴趣区域,并且需要大量人工干预。而基于机器学习的分割方法在模型建模时存在局限性且缺乏可解释性。本研究提出一种基于Transformer和卷积神经网络结合形态结构约束的三维医学图像分割方法。编码器中利用卷积神经网络和Transformer构建U型网络结构提取多重特征;解码器中采用上采样并通过跳跃连接将不同层次的特征拼接在一起;加入形态结构约束模块,通过提取病灶和脏器等分割目标的形状信息,以增强模型可解释性,并采用最大池化和平均池化操作,对经过卷积神经网络得到的结果进一步提取有代表性的特征,作为形态结构模块的输入,最终提高分割结果的准确性。在公开数据集Synapse和ACDC上利用评价指标Dice相似系数(DSC)和Hausdorff距离(HD)验证所提出算法的有效性。其中,在Synapse数据集上,18例数据作为训练集,12例数据作为测试集;在ACDC数据集上,70例数据作为训练集,10例数据作为验证集,20例数据作为测试集。实验结果表明,在Sy...     

基于级联3D U-Net的CT和MR视交叉自动分割方法

目的:基于级联3D U-Net,利用配对患者头颈部数据［CT和磁共振图像（MRI）］,取得比仅CT数据更高分割精度的视交叉自动分割结果。方法:该级联3D U-Net由一个原始3D U-Net和改进的3D D-S U-Net（3D Deeply-Supervised U-Net）组成,实验使用了60例患者头颈部CT图像及MRI图像（T1和T2模态）,其中随机选取15例患者数据作为测试集,并使用相似性系数（DSC）评估视交叉的自动分割精度。结果:对于测试集中的所有病例,采用多模态数据（CT和MRI）的视交叉的DSC为0.645±0.085,采用单模态数据（CT）的视交叉的DSC为0.552±0.096。结论:基于级联3D U-Net的多模态自动分割模型能够较为准确地实现视交叉的自动分割,且优于仅利用单模态数据的方法,可以辅助医生提高放疗计划制定的工作效率。     

阿尔茨海默病患者MR图像偏差场修正和脑组织分割的研究

目的探索适用于阿尔茨海默病（Alzheimer’s disease,AD）患者MR图像脑组织的分割的方法。方法结合阿尔茨海默病患者MR图像中组织区域和边缘的特性对传统水平集进行改进,利用同态滤波对图像进行偏差场修正,增加了UNSHARP MASK处理方法,有效避免了水平集边界泄漏问题。结果标准体膜和真实数据实验证实,该改进算法分割结果优于SPM5。结论利用修正偏差场和添加UNSHARP MASK方法有可能提高AD患者MR图像脑组织分割的准确性和鲁棒性,本研究为MR图像脑组织的精确分割和进一步准确测量作了有益探索。     

基于三重注意力的脑肿瘤图像分割网络

脑肿瘤图像分割问题是脑肿瘤临床诊断和治疗脑肿瘤疾病计算机辅助诊断的基础.针对脑肿瘤MRI图像分割网络深度过深和局部与全局特征信息联系匮乏导致图像分割精度降低等问题,提出一种基于三重注意力的脑肿瘤图像分割网络.首先,借鉴残差结构,将原始图像分割网络结构的编码层和解码层中的卷积模块替换为深度残差模块,解决网络加深带来的梯度...     

3D cerebral MR image segmentation using multiple-classifier system

The three soft brain tissues white matter (WM), gray matter (GM), and cerebral spinal fluid (CSF) identified in a magnetic resonance (MR) image via image segmentation techniques can aid in structural and functional brain analysis, brain’s anatomical structures measurement and visualization, neurodegenerative disorders diagnosis, and surgical planning and image-guided interventions, but only if obtained segmentation results are correct. This paper presents a multiple-classifier-based system for automatic brain tissue segmentation from cerebral MR images. The developed system categorizes each voxel of a given MR image as GM, WM, and CSF. The algorithm consists of preprocessing, feature extraction, and supervised classification steps. In the first step, intensity non-uniformity in a given MR image is corrected and then non-brain tissues such as skull, eyeballs, and skin are removed from the image. For each voxel, statistical features and non-statistical features were computed and used a feature vector representing the voxel. Three multilayer perceptron (MLP) neural networks trained using three different datasets were used as the base classifiers of the multiple-classifier system. The output of the base classifiers was fused using majority voting scheme. Evaluation of the proposed system was performed using Brainweb simulated MR images with different noise and intensity non-uniformity and internet brain segmentation repository (IBSR) real MR images. The quantitative assessment of the proposed method using Dice, Jaccard, and conformity coefficient metrics demonstrates improvement (around 5 % for CSF) in terms of accuracy as compared to single MLP classifier and the existing methods and tools such FSL-FAST and SPM. As accurately segmenting a MR image is of paramount importance for successfully promoting the clinical application of MR image segmentation techniques, the improvement obtained by using multiple-classifier-based system is encouraging.     

Unsupervised Segmentation of Head Tissues from Multi-modal MR Images for EEG Source Localization

In this paper, we present and evaluate an automatic unsupervised segmentation method, hierarchical segmentation approach (HSA)–Bayesian-based adaptive mean shift (BAMS), for use in the construction of a patient-specific head conductivity model for electroencephalography (EEG) source localization. It is based on a HSA and BAMS for segmenting the tissues from multi-modal magnetic resonance (MR) head images. The evaluation of the proposed method was done both directly in terms of segmentation accuracy and indirectly in terms of source localization accuracy. The direct evaluation was performed relative to a commonly used reference method brain extraction tool (BET)–FMRIB’s automated segmentation tool (FAST) and four variants of the HSA using both synthetic data and real data from ten subjects. The synthetic data includes multiple realizations of four different noise levels and several realizations of typical noise with a 20 % bias field level. The Dice index and Hausdorff distance were used to measure the segmentation accuracy. The indirect evaluation was performed relative to the reference method BET-FAST using synthetic two-dimensional (2D) multimodal magnetic resonance (MR) data with 3 % noise and synthetic EEG (generated for a prescribed source). The source localization accuracy was determined in terms of localization error and relative error of potential. The experimental results demonstrate the efficacy of HSA-BAMS, its robustness to noise and the bias field, and that it provides better segmentation accuracy than the reference method and variants of the HSA. They also show that it leads to a more accurate localization accuracy than the commonly used reference method and suggest that it has potential as a surrogate for expert manual segmentation for the EEG source localization problem.     

基于DenseMedic网络的脑皮层下结构的语义分割

脑皮层下结构分割问题是神经科及其他相关疾病计算机辅助诊断和治疗的基础。通过分割和分析核磁共振图像中的脑结构,可以对自闭症谱系障碍、脑卒中、脑肿瘤等疾病进行早期诊断和治疗。为解决精准脑结构分割的问题,基于深度学习基本理论,提出一种DenseMedic网络的核磁共振图像脑皮层下结构的分割算法。首先,OreoDown方法通过较早地增大卷积核的步长增大特征感受野的增长速度,并使用不变尺寸的卷积层夹心式地恢复网络深度,使速度的增加带来有效的感受野增加;其次,DenseMedic使用DenseNet的思想实例化OreoDown框架,通过密集连接的特征提取操作来获取多尺度的上下文信息;最后,在各层中使用混合空洞卷积进一步扩大感受野,解决特征感知过于粗糙的问题。采用Dice相似度系数(DSC)、交并比(IoU)、95% Hausdorff表面距离(HSD95)和平均表面距离(ASD) 4个指标,评价神经网络的分割性能。在公开的IBSR数据集的18例图像上进行实验,算法的4个指标分别达到89.2%、80.7%、1.982和0.882;在公开的MBBrainS18数据集的7例图像上的实验显示,算法的4个指标分别达到88.7%、79.8%、1.249和0.570。实验表明,所提出的算法使脑结构的分割结果与真实结构在区域上有更多的重叠, 在轮廓上更加相似,可以更好地完成各个脑皮层下结构的分割。在临床应用中,对脑皮层下结构的精准分割将有助于准确测量相关疾病诊断的关键指标,并实现快速的计算机辅助治疗。     

Effect of Dataset Size and Medical Image Modality on Convolutional Neural Network Model Performance for Automated Segmentation: A CT and MR Renal Tumor Imaging Study

The aim of this study is to investigate the use of an exponential-plateau model to determine the required training dataset size that yields the maximum medical image segmentation performance. CT and MR images of patients with renal tumors acquired between 1997 and 2017 were retrospectively collected from our nephrectomy registry. Modality-based datasets of 50, 100, 150, 200, 250, and 300 images were assembled to train models with an 80–20 training-validation split evaluated against 50 randomly held out test set images. A third experiment using the KiTS21 dataset was also used to explore the effects of different model architectures. Exponential-plateau models were used to establish the relationship of dataset size to model generalizability performance. For segmenting non-neoplastic kidney regions on CT and MR imaging, our model yielded test Dice score plateaus of \(0.93\pm 0.02\) and \(0.92\pm 0.04\) with the number of training-validation images needed to reach the plateaus of 54 and 122, respectively. For segmenting CT and MR tumor regions, we modeled a test Dice score plateau of \(0.85\pm 0.20\) and \(0.76\pm 0.27\), with 125 and 389 training-validation images needed to reach the plateaus. For the KiTS21 dataset, the best Dice score plateaus for nn-UNet 2D and 3D architectures were \(0.67\pm 0.29\) and \(0.84\pm 0.18\) with number to reach performance plateau of 177 and 440. Our research validates that differing imaging modalities, target structures, and model architectures all affect the amount of training images required to reach a performance plateau. The modeling approach we developed will help future researchers determine for their experiments when additional training-validation images will likely not further improve model performance.

     

A region-based segmentation of tumour from brain CT images using nonlinear support vector machine classifier

The proposed system provides new textural information for segmenting tumours, efficiently and accurately and with less computational time, from benign and malignant tumour images, especially in smaller dimensions of tumour regions of computed tomography (CT) images. Region-based segmentation of tumour from brain CT image data is an important but time-consuming task performed manually by medical experts. The objective of this work is to segment brain tumour from CT images using combined grey and texture features with new edge features and nonlinear support vector machine (SVM) classifier. The selected optimal features are used to model and train the nonlinear SVM classifier to segment the tumour from computed tomography images and the segmentation accuracies are evaluated for each slice of the tumour image. The method is applied on real data of 80 benign, malignant tumour images. The results are compared with the radiologist labelled ground truth. Quantitative analysis between ground truth and the segmented tumour is presented in terms of segmentation accuracy and the overlap similarity measure dice metric. From the analysis and performance measures such as segmentation accuracy and dice metric, it is inferred that better segmentation accuracy and higher dice metric are achieved with the normalized cut segmentation method than with the fuzzy c-means clustering method.     

A region-based segmentation of tumour from brain CT images using nonlinear support vector machine classifier

The proposed system provides new textural information for segmenting tumours, efficiently and accurately and with less computational time, from benign and malignant tumour images, especially in smaller dimensions of tumour regions of computed tomography (CT) images. Region-based segmentation of tumour from brain CT image data is an important but time-consuming task performed manually by medical experts. The objective of this work is to segment brain tumour from CT images using combined grey and texture features with new edge features and nonlinear support vector machine (SVM) classifier. The selected optimal features are used to model and train the nonlinear SVM classifier to segment the tumour from computed tomography images and the segmentation accuracies are evaluated for each slice of the tumour image. The method is applied on real data of 80 benign, malignant tumour images. The results are compared with the radiologist labelled ground truth. Quantitative analysis between ground truth and the segmented tumour is presented in terms of segmentation accuracy and the overlap similarity measure dice metric. From the analysis and performance measures such as segmentation accuracy and dice metric, it is inferred that better segmentation accuracy and higher dice metric are achieved with the normalized cut segmentation method than with the fuzzy c-means clustering method.     

基于3D体素增强和3D alpha背景分离的多发性硬化症病灶分割方法

目的:提出一种用于T1加权像、T2加权像和流体衰减反演恢复（Flair）磁共振图像的多发性硬化症（MS）病变分割方法。方法:首先基于3D图像增强技术,将高强度MS病变区域与其他组织区域区分开来。然后利用假阳性降低方法,去除一些强度和密度不均匀的假阳性目标区域（VOI）,并利用颜色分割法去除白质之外的VOI。最后利用彩色MR技术生成3个区域,以便细化分割MS病变。结果:在CHB数据集上进行测试,得到真阳率均值为0.48,Dice相似系数均值为0.52。结论:该方法能够有效去除噪声及其他无关非病变组织,并能准确识别并分割MS病变,该方法的有效性、准确性能为后续的MS分割技术分析提供依据。同时为MS病变的预防治疗、病情跟踪提供客观、方便的诊疗方法。 【关键词】多发性硬化症;病灶分割;3D体素增强;3D alpha背景分离;颜色分割技术