针对磁共振(MR)成像具有灰度不均匀、部分容积效应等缺陷,给出一种粗糙集自适应粒度的脑肿瘤MR图像分割方法,从而提高脑肿瘤分割质量.利用粗糙集自适应粒度方法选取最优分割粒度,并用粗糙集模拟目标和背景区域的上下近似,通过优化目标和背景区域的粗糙度,获得MR脑肿瘤图像分割的最佳阈值.粗糙集自适应粒度方法能够较好地提取出脑肿瘤区域.实验结果表明,该方法优于传统粗糙集分割法,且具有一定实用性和灵活性.

目的:脑核磁共振(MR)图像中普遍存在噪声的影响,传统的核模糊C-均值(Kernel Fuzzy C-means,KFCM)算法无法得到理想的脑组织分割结果,为此提出一种基于核模糊聚类优化算法的分割模型.方法:首先通过粒子群算法确定KFCM的初始聚类中心,然后利用自适应中值滤波消除图像中的噪声,最后采用该模型分别对不同的图像进行实验.结果:该方法不仅能迅速确定图像的初始聚类中心,并且有效地消除图像中的噪声.结论:与传统KFCM算法相比,提出的模型具有更高的精确度和分割效率.

目的:旨在介绍一种皮肤病移动辅助诊断系统,可为广大皮肤病医生与患者提供便捷、快速、准确的专家咨询与诊断服务.方法:系统通过移动互联网将手机拍摄的皮肤表层图像上传至云端服务器,利用机器学习算法进行图像数据挖掘和辅助诊断,诊断结果(报告)与健康指导可以发送到用户的智能终端进行显示及存储.结果:系统可为患者提供在线上传皮肤病图像、推送皮肤病诊断报告及健康知识,为患者提供更加方便快捷的移动医疗服务;同时辅助医生临床诊断工作,提高诊病准确率.结论:系统打破了皮肤病患者传统就医模式,提高了皮肤病医生的临床工作效率,促进皮肤病诊断更加便捷化、及时化和精准化,提高医疗服务质量.

Background: Mining knowledge from microarray data is one of the popular research topics in biomedical informatics. Gene selection is a significant research trend in biomedical data mining, since the accuracy of tumor identification heavily relies on the genes biologically relevant to the identified problems. Objective: In order to select a small subset of informative genes from numerous genes for tumor identification, various computational intelligence methods were presented. However, due to the high data dimensions, small sample size, and the inherent noise available, many computational methods confront challenges in selecting small gene subset. Methods: In our study, we propose a novel algorithm PSONRS_KNN for gene selection based on the particle swarm optimization (PSO) algorithm along with the neighborhood rough set (NRS) reduction model and the K-nearest neighborhood (KNN) classifier. Results: First, the top-ranked candidate genes are obtained by the GainRatioAttributeEval preselection algorithm in WEKA. Then, the minimum possible meaningful set of genes is selected by combining PSO with NRS and KNN classifier. Conclusion: Experimental results on five microarray gene expression datasets demonstrate that the performance of the proposed method is better than existing state-of-the-art methods in terms of classification accuracy and the number of selected genes.

Brain tumor segmentation on MR images has significant clinical meaning due to glioblastomas which are the most lethal form of these tumors. Compared to manual segmentation, automatic segmentation system is superior in timesaving and experience-insensitivity for doctors during clinical practice. However, its inherent contradiction is not addressed yet. i.e. imbalance of multi-class of different brain tissues. As such, we proposed a multi-class focal loss to make the loss function emphasis on bad-classified voxels in MR images. Our experiments based on the 3D UNet model proved that this method can significantly improve labeling and segmentation accuracy as compared to other loss layers.

Traditional fuzzy C-means (FCM) clustering method is not able to get the desired results of brain segmentation in the brain magnetic resonance (MR) image. In this paper, a new method of brain image segmentation based on Gaussian filtering and FCM clustering algorithm is proposed. The method adopted Gaussian filtering to remove noise and the initial cluster center was determined by the gray histograms to obtain. Using this method, 20 samples of brain MR images with 9% and 5% noise interference provided by Brain Web were segmented. The experimental results showed the proposed method was accuracy and efficiency.