基于密集网络改进的肺结节良恶性分类模型

目的 探讨改进后的卷积神经网络模型对肺结节进行良恶性分类的准确率。方法 以分类模型密集网络（DenseNet）为基础模型,采用中间密度投影方法将肺结节的三维信息输入卷积神经网络进行训练,并针对肺结节良恶性分类问题适应性改进神经网络结构,将传统损失函数Cross Entropy Loss替换为Focal Loss,使网络能着重学习难以分辨的肺结节。结果 改进后神经网络模型对良恶性肺结节分类的准确率为89.93%,曲线下面积为0.947。结论 适应性改进后的卷积神经网络模型判断良恶性肺结节准确率较高。     

Unruptured sinus of Valsalva aneurysm with right ventricular outflow obstruction

The recommended operative management of unruptured sinus of Valsalva aneurysm consists of closure of the mouth of the aneurysm with or without aortic valve surgery. We report a case of unruptured aneurysm producing right ventricular outflow tract obstruction. Closure of the mouth of the aneurysm failed to relieve the obstruction, which was subsequently achieved by excising the aneurysmal wall overlying the outflow tract.     

Oviposition deterrent activity of three mosquito repellents diethyl phenyl acetamide (DEPA), diethyl m toluamide (DEET), and diethyl benzamide (DEB) on Aedes aegypti, Aedes albopictus, and Culex quinquefasciatus

Oviposition deterrent activity of three mosquito repellents namely diethyl phenyl acetamide (DEPA), diethyl benzamide (DEB) along with diethyl toluamide (DEET) was studied in the laboratory against Aedes aegypti, A. albopictus, and Culex quinquefasciatus mosquitoes. All the repellents evaluated at three log concentrations 0.1, 0.01, and 0.001 % in comparison with control by dual choice method. The oviposition response and oviposition activity index (OAI) was estimated by counting number of eggs laid in both the treatment and control. Mean OAI determined for A. aegypti was in the range of ?0.23 to +0.22, ?0.3 to +0.27, and +0.04 to +0.33 for DEPA, DEET, and DEB, respectively, whereas it was ?0.77 to ?0.035, ?0.92 to +0.001, and ?0.77 to ?0.07 for A. albopictus and ?0.927 to ?0.251, ?1 to ?0.41, and ?0.94 to ?0.17 for C. quinquefasciatus. The oviposition deterrent activity was observed in the order of DEET > DEPA > DEB. Similarly, the response of species toward the repellents as oviposition deterrent was in the trend of C. quinquefasciatus > A. albopictus > A. aegypti. The finding suggests potential role of mosquito repellents as oviposition deterrent.     

A method of gauging dental radiographs during treatment planning for dental implants

AIM: The goal of pre-surgical dental implant treatment planning is to position the optimum number and size of implant fixtures to achieve the best restorative results. The purpose of this article is to describe the use of radiographic imaging software to calibrate and measure anatomical landmarks to overcome inherent distortions associated with dental radiographs. The procedure along with its potential use as an adjunct to radiographic interpretation in routine clinical implant practice is presented. BACKGROUND: Diagnostic imaging is an essential component of implant treatment planning, and a variety of advanced imaging modalities have been recommended to assist the dentist in assessing potential sites for implants. Although technological advances have resulted in new imaging innovations for implant dentistry, dental radiography remains the most widely used tool for determining the quantity and quality of alveolar bone as it is a non-invasive procedure. However, the unreliable magnification factor associated with conventional radiographs remains a major problem when estimating the amount of bone available at the implant site. SUMMARY: This image measurement technique is capable of assessing the bone quantity by measuring the height and width of the alveolar crest for a specified region in a two dimensional plane in any direction related to the visible landmarks in the oral cavity. These measurements can be used by the clinician to select the type of implant and its position. Since there is no additional equipment or cost involved, the technique can be used as an important adjunct in implant practice.     

MR图像纹理特征融合诊断前列腺癌

目的 探讨三维空间自适应局部三元模式（LTP）纹理特征、常规纹理特征以及灰度统计特征融合用于诊断前列腺癌的价值。方法 从MRI中分割出前列腺外周带,提取自适应LTP融合纹理特征纹理特征、常规纹理特征和灰度统计特征,采用Adaboost算法分别获得每个特征族群以及3个族群特征融合的模型,计算对前列腺癌的诊断效能。结果 三维空间自适应LTP融合纹理特征诊断前列腺癌的AUC为0.79±0.04,敏感度为78.31%（65/83）,特异度为80.81%（80/99）,准确率为79.67%（145/182）;常规纹理特征诊断前列腺癌的AUC为0.71±0.04,敏感度为72.29%（60/83）,特异度为81.82%（81/99）,准确率为77.47%（141/182）;灰度统计特征诊断前列腺癌的AUC为0.80±0.04,敏感度为78.31%（65/83）,特异度为82.83%（82/99）,准确率80.77%（147/182）;融合特征诊断前列腺癌的AUC为0.87±0.04,敏感度为86.75%（72/83）,特异度为88.89%（88/99）,准确率为87.91%（160/182）。结论 通过融合局部三元模式特征、常规纹理特征、灰度统计特征,可有效提高诊断前列腺癌的效能。