Classification and Research of Skin Lesions Based on Machine Learning

Classification of skin lesions is a complex identification challenge. Due to the wide variety of skin lesions, doctors need to spend a lot of time and effort to judge the lesion image which zoomed through the dermatoscopy. The diagnosis which the algorithm of identifying pathological images assists doctors gets more and more attention. With the development of deep learning, the field of image recognition has made longterm progress. The effect of recognizing images through convolutional neural network models is better than traditional image recognition technology. In this work, we try to classify seven kinds of lesion images by various models and methods of deep learning, common models of convolutional neural network in the field of image classification include ResNet, DenseNet and SENet, etc. We use a fine-tuning model with a multi-layer perceptron, by training the skin lesion model, in the validation set and test set we use data expansion based on multiple cropping, and use five models’ ensemble as the final results. The experimental results show that the program has good results in improving the sensitivity of skin lesion diagnosis.     

Computer‐aided diagnosis system for the detection of bronchiectasis in chest computed tomography images

A computer‐aided diagnosis (CAD) system has been developed for the detection of bronchiectasis from computed tomography (CT) images of chest. A set of CT images of the chest with known diagnosis were collected and these images were first denoised using Wiener filter. The lung tissue was then segmented using optimal thresholding. The Pathology Bearing Regions (PBRs) were then extracted by applying pixel‐based segmentation. For each PBR, a gray level co‐occurrence matrix (GLCM) was constructed. From the GLCM texture features were extracted and feature vectors were constructed. A probabilistic neural network (PNN) was constructed and trained using this set of feature vectors. The images together with the PBRs and the corresponding feature vector and diagnosis were stored in an image database. Rules for diagnosis and for determining the severity of the disease were generated by analyzing the images known to be affected by bronchiectasis. The rules were then validated by a human expert. The validated rules were stored in the Knowledge Base. When a physician gives a CT image to the CAD system, it first transforms the image into a set of feature vectors, one for each PBR in the image. It then performs the diagnosis using two techniques: PNN and mahalanobis distance measure. The final diagnosis and the severity of the disease are determined by correlating the diagnosis determined by both the techniques in consultation with the knowledge base. The system also retrieves similar cases from the database. Thus, this system would aid the physicians in diagnosing bronchiectasis. © 2009 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 19, 290–298, 2009     

基于迁移学习与深度残差网络的滚动轴承快速故障诊断算法EI北大核心CSCD

针对现有基于深度学习的滚动轴承故障诊断算法训练参数量大,训练时间长且需要大量训练样本的缺点,提出了一种基于迁移学习(TL)与深度残差网络(ResNet)的快速故障诊断算法(TL-ResNet)。首先开发了一种将短时傅里叶变换(STFT)与伪彩色处理相结合的振动信号转三通道图像数据的方法;然后将在ImageNet数据集上训练的ResNet18模型作为预训练模型,通过迁移学习的方法,应用到滚动轴承故障诊断领域当中;最后对滚动轴承在不同工况下的故障诊断问题,提出了采用小样本迁移的方法进行诊断。在凯斯西储大学(CWRU)与帕德博恩大学(PU)数据集上进行了试验,TL-ResNet的诊断准确率分别为99.8%与95.2%,且在CWRU数据集上TL-ResNet的训练时间仅要1.5 s,这表明本算法优于其他的基于深度学习的故障诊断算法与经典算法,可用于实际工业环境中的快速故障诊断。     

基于卷积神经网络的典型农作物叶病害识别算法

目前,典型的一些农作物叶病害诊断主要依靠人工,但该方式耗时费力。针对大豆、棉花、水稻、小麦和玉米5类典型农作物的常见叶病害诊断问题,提出了一种基于卷积神经网络的典型农作物叶病害识别方法。从Plantvillage数据库以及其他网站收集典型农作物的叶病害图像,并对这些图像进行预处理,构建了含12 836张的数据集。参照AlexNet框架,构建8层卷积神经网络,采用迁移学习训练网络,最后通过测试集验证网络的识别准确率和损失值。分析不同的卷积神经网络的性能,实验结果表明：本算法对典型农作物的叶病害有良好的识别效果;迁移学习模式下,学习率为0.001时本算法在训练集的识别准确率约为99.47%,在测试集的识别准确率约为96.18%。     

Detecting Lumbar Implant and Diagnosing Scoliosis from Vietnamese X-Ray Imaging Using the Pre-Trained API Models and Transfer Learning

With the rapid growth of the autonomous system, deep learning has become integral parts to enumerate applications especially in the case of healthcare systems. Human body vertebrae are the longest and complex parts of the human body. There are numerous kinds of conditions such as scoliosis, vertebra degeneration, and vertebrate disc spacing that are related to the human body vertebrae or spine or backbone. Early detection of these problems is very important otherwise patients will suffer from a disease for a lifetime. In this proposed system, we developed an autonomous system that detects lumbar implants and diagnoses scoliosis from the modified Vietnamese x-ray imaging. We applied two different approaches including pre-trained APIs and transfer learning with their pre-trained models due to the unavailability of sufficient x-ray medical imaging. The results show that transfer learning is suitable for the modified Vietnamese x-ray imaging data as compared to the pre-trained API models. Moreover, we also explored and analyzed four transfer learning models and two pre-trained API models with our datasets in terms of accuracy, sensitivity, and specificity.     

Nodule detection in digital chest radiography: effect of system noise

Apart from the image content that is the reproduction of anatomy and possible lesions, an X-ray image also contains system noise due to the limited number of photons and other internal noise sources in the system (image plate artefacts, electronic noise, etc.). The aim of this study was to determine the extent to which the system noise influences the detection of subtle lung nodules in five different regions of the chest. This was done by conducting a receiver operating characteristic (ROC) study with five observers on two different sets of images; clinical chest X-ray images and images of a LucAl phantom at similar dose levels found in the different regions of the chest. In both image types, mathematically simulated nodules (with a full-width-at-fifth-maximum of 10 mm) were added to the images at varying contrast levels. As a measure of the influence of system noise on the detection of subtle lung nodules, the ratio between the contrast needed to obtain an area under the ROC curve of 0.80 in the system noise images to that needed in the clinical images was used. The contrast ratio between system noise images and clinical images ranged from approximately 0.02 (in the hilar region) to 0.18 (in the lower mediastinal region). The maximum difference in contrast needed for the corresponding system noise images, collected at the lowest and the highest dose represented in the anatomical image, was a factor of 2. These results indicate that probably no region in a chest X-ray image is limited by the number of quanta to the detector for the detection of 10 mm lung nodules when a radiation dose corresponding to a system with speed class 200 (leading to a detector dose of approximately 9 muGy behind the parenchyma) is used.     

Evaluation of image quality of lumbar spine images: a comparison between FFE and VGA

PURPOSE: The aim of the present study is to compare two different methods for evaluation of the quality of clinical X-ray images. METHODS: Based on fifteen lumbar spine radiographs, two new sets of images were created. A hybrid image set was created by adding two distributions of artificial lesions to each original image. The image quality parameters spatial resolution and noise were manipulated and a total of 210 hybrid images were created. A set of 105 disease-free images was created by applying the same combinations of spatial resolution and noise to the original images. The hybrid images were evaluated with the free-response forced error experiment (FFE) and the normal images with visual grading analysis (VGA) by nine experienced radiologists. RESULTS: In the VGA study, images with low noise were preferred over images with higher noise levels. The alteration of the MTF had a limited influence on the VGA score. For the FFE study, the visibility of the lesions was independent of the sharpness and the noise level. No correlation was found between the two image quality measures. CONCLUSIONS: FFE is a precise method for evaluation of image quality, but the results are only valid for the type of lesion used in the study, whereas VGA is a more general method for clinical image quality assessment. The results of the FFE study indicate that there might be a potential to lower the dose levels in lumbar spine radiography without losing important diagnostic information.     

X线胸片与常规CT、HRCT对矽肺及其合并症诊断的对比分析

目的:比较X线胸片与CT、HRCT对矽肺及其合并症的诊断作用,并对不同诊断结果进行分析。方法:选取2008年1月-2013年1月我院收治的矽肺患者100例为研究对象,所有患者均行X线胸片、CT、HRCT检查,对三种影像学检查结果进行比较分析。结果:HRCT对于小阴影、肺气肿、肺间质纤维化的影像特征明显优于CT与X线胸片;而X线胸片对于大阴影的显示、肺气肿、肺间质纤维化、胸膜改变的影像特征表现略差于CT、HRCT。结论:CT对于矽肺的诊断具有一定的临床意义,特别是HRCT,而结合X线胸片能进一步对矽肺病变部位进行准确的分析。     

Developing a Recognition System for Classifying COVID-19 Using a Convolutional Neural Network Algorithm

The COVID-19 pandemic poses an additional serious public health threat due to little or no pre-existing human immunity, and developing a system to identify COVID-19 in its early stages will save millions of lives. This study applied support vector machine (SVM), k-nearest neighbor (K-NN) and deep learning convolutional neural network (CNN) algorithms to classify and detect COVID-19 using chest X-ray radiographs. To test the proposed system, chest X-ray radiographs and CT images were collected from different standard databases, which contained 95 normal images, 140 COVID-19 images and 10 SARS images. Two scenarios were considered to develop a system for predicting COVID-19. In the first scenario, the Gaussian filter was applied to remove noise from the chest X-ray radiograph images, and then the adaptive region growing technique was used to segment the region of interest from the chest X-ray radiographs. After segmentation, a hybrid feature extraction composed of 2D-DWT and gray level co-occurrence matrix was utilized to extract the features significant for detecting COVID-19. These features were processed using SVM and K-NN. In the second scenario, a CNN transfer model (ResNet 50) was used to detect COVID-19. The system was examined and evaluated through multiclass statistical analysis, and the empirical results of the analysis found significant values of 97.14%, 99.34%, 99.26%, 99.26% and 99.40% for accuracy, specificity, sensitivity, recall and AUC, respectively. Thus, the CNN model showed significant success; it achieved optimal accuracy, effectiveness and robustness for detecting COVID-19.     

Patient doses and image quality in digital chest radiology

Chest X-ray examination is one of the most frequently required procedures used in clinical practice. For studying the image quality of different X-ray digital systems and for the control of patient doses during chest radiological examinations, the standard anthropomorphic lung/chest phantom RSD 330 has been used and exposed in different digital modalities available in Slovakia. To compare different techniques of chest examination, a special software has been developed that enables researchers to compare digital imaging and communications in medicine header images from different digital modalities, using a special viewer. In this paper, this special software has been used for an anonymous correspondent audit for testing image quality evaluation by comparing various parameters of chest imaging, evaluated by 84 Slovak radiologists. The results of the comparison have shown that the majority of the participating radiologists felt that the highest image quality is reached with a flat panel, assessed by the entrance surface dose value, which is approximately 75% lower than the diagnostic reference level of chest examination given in the Slovak legislation. Besides the results of the audit, the possibilities of using the software for optimisation, education and training of medical students, radiological assistants, physicists and radiologists in the field of digital radiology will be described.     

Pattern Recognition in Photoacoustic Dataset

In photoacoustic imaging, optical absorption properties of matter are imaged by detecting the ultrasound that is produced when the material is illuminated by a laser. For medical imaging, photoacoustics is a useful tool since matter in the human body has different optical absorption properties. In this study, pattern recognition systems are used to study a set of medical images for tumor identification and extraction—to detect the specific area in which the tumor is present. The objective is to incorporate this information into real-time image acquisition systems to improve medical diagnosis. Preliminary results obtained by studying the image dataset demonstrated the interchangeability of the proposed system. A system of automatic classification was constructed, using a set of images with and without cancerous tumors to evaluate the proposed method. The training set used was manually labeled, and the test set was never seen by the training set. The results helped us determine the feasibility of the proposed system.     

Development of Deep Learning Models for Predicting the Effects of Exposure to Engineered Nanomaterials on Daphnia magna

This study presents the results of applying deep learning methodologies within the ecotoxicology field, with the objective of training predictive models that can support hazard assessment and eventually the design of safer engineered nanomaterials (ENMs). A workflow applying two different deep learning architectures on microscopic images of Daphnia magna is proposed that can automatically detect possible malformations, such as effects on the length of the tail, and the overall size, and uncommon lipid concentrations and lipid deposit shapes, which are due to direct or parental exposure to ENMs. Next, classification models assign specific objects (heart, abdomen/claw) to classes that depend on lipid densities and compare the results with controls. The models are statistically validated in terms of their prediction accuracy on external D. magna images and illustrate that deep learning technologies can be useful in the nanoinformatics field, because they can automate time‐consuming manual procedures, accelerate the investigation of adverse effects of ENMs, and facilitate the process of designing safer nanostructures. It may even be possible in the future to predict impacts on subsequent generations from images of parental exposure, reducing the time and cost involved in long‐term reproductive toxicity assays over multiple generations.     

Transition from screen-film to computed radiography in a paediatric hospital: the missing link towards optimisation

In paediatrics, the risks associated with ionising radiation should be a major concern, due to children's higher susceptibility to radiation effects. Measure entrance skin dose (ESD) in chest and pelvis X-ray projections and compare the results with the 'European guidelines on quality criteria for diagnostic radiographer images in paediatrics' in order to optimise radiological practice. ESD values were obtained using an ionisation chamber Diamentor M4 KDK (PTW) in 429 children, who underwent chest X-ray or pelvis X-ray in a Computed Radiography system. In the first phase of the study, data were collected according to protocols used in the department; in a second phase different tube voltage values were used according to patient weight. A third phase was carried out, only for chest X-ray, using the exposure parameters of phase 2, plus activating lateral ionisation chamber. Three paediatric radiologists blindly assessed image quality of chest X-ray, using a validated assessment available in the 'European guidelines on quality criteria for diagnostic radiographer images in paediatrics'. Considering all the patients submitted to chest X-ray, the average ESD was 0.22, 0.16 and 0.08 mGy, for phases 1, 2 and 3, respectively. For pelvis X-ray, the average ESD decreased from 1.18 mGy in phase 1 to 0.78 mGy in phase 2. Dose optimisation was achieved. ESD was reduced 63.6 and 33.9 % in chest and pelvis X-ray, respectively.     

CARE Dose4D技术在低剂量CT胸部体检中的应用探讨

目的：探讨CARE Dose4D技术在低剂量CT胸部体检的应用价值。方法：收集健康体检者160例,随机分为ABCD四组,各40例,参考管电流依次降低,然后评价四组图像质量,并计算有效吸收剂量（ED）。结果：BCD三组有效剂量均较A组降低,图像噪声增加,图像质量有所下降但不影响诊断。结论：采用该技术在可得到满足诊断的CT图像时大幅降低辐射剂量。     

On the Detection of COVID-19 from Chest X-Ray Images Using CNN-Based Transfer Learning

Coronavirus disease (COVID-19) is an extremely infectious disease and possibly causes acute respiratory distress or in severe cases may lead to death. There has already been some research in dealing with coronavirus using machine learning algorithms, but few have presented a truly comprehensive view. In this research, we show how convolutional neural network (CNN) can be useful to detect COVID-19 using chest X-ray images. We leverage the CNN-based pre-trained models as feature extractors to substantiate transfer learning and add our own classifier in detecting COVID-19. In this regard, we evaluate performance of five different pre-trained models with fine-tuning the weights from some of the top layers. We also develop an ensemble model where the predictions from all chosen pre-trained models are combined to generate a single output. The models are evaluated through 5-fold cross validation using two publicly available data repositories containing healthy and infected (both COVID-19 and other pneumonia) chest X-ray images. We also leverage two different visualization techniques to observe how efficiently the models extract important features related to the detection of COVID- 19 patients. The models show high degree of accuracy, precision, and sensitivity. We believe that the models will aid medical professionals with improved and faster patient screening and pave a way to further COVID-19 research.     

Reducing the absorbed dose in analogue radiography of infant chest images by improving the image quality, using image processing techniques

Radiographic inspection is one of the most widely employed techniques for medical testing methods. Because of poor contrast and high un-sharpness of radiographic image quality in films, converting radiographs to a digital format and using further digital image processing is the best method of enhancing the image quality and assisting the interpreter in their evaluation. In this research work, radiographic films of 70 infant chest images with different sizes of defects were selected. To digitise the chest images and employ image processing the two algorithms (i) spatial domain and (ii) frequency domain techniques were used. The MATLAB environment was selected for processing in the digital format. Our results showed that by using these two techniques, the defects with small dimensions are detectable. Therefore, these suggested techniques may help medical specialists to diagnose the defects in the primary stages and help to prevent more repeat X-ray examination of paediatric patients.     

Nodule detection in digital chest radiography: part of image background acting as pure noise

There are several factors that influence the radiologist's ability to detect a specific structure/lesion in a radiograph. Three factors that are commonly known to be of major importance are the signal itself, the system noise and the projected anatomy. The aim of this study was to determine to what extent the image background acts as pure noise for the detection of subtle lung nodules in five different regions of the chest. A receiver operating characteristic (ROC) study with five observers was conducted on two different sets of images, clinical chest X-ray images and images with a similar power spectrum as the clinical images but with a random phase spectrum, resulting in an image background containing pure noise. Simulated designer nodules with a full-width-at-fifth-maximum of 10 mm but with varying contrasts were added to the images. As a measure of the part of the image background that acts as pure noise, the ratio between the contrast needed to obtain an area under the ROC curve of 0.80 in the clinical images to that in the random-phase images was used. The ratio ranged from 0.40 (in the lateral pulmonary regions) to 0.83 (in the hilar regions) indicating that there was a large difference between different regions regarding to what extent the image background acted as pure noise; and that in the hilar regions the image background almost completely acted as pure noise for the detection of 10 mm nodules.     

卷烟包装外观缺陷数据集构建及深度学习检测技术研究

目的 为了提升烟包缺陷检测的准确率,构建卷烟包装外观缺陷识别基准数据集,并开展主流深度学习模型在卷烟包装外观缺陷智能检测中的应用研究。方法 首先,从生产运行中的ZB45型细支烟硬盒包装机组采集缺陷图像,经过人工审核与筛选后获取典型的缺陷数据。然后,根据缺陷的特征与成因,将缺陷数据划分为23个类别,并逐一进行目标检测框标注。最终,形成了包含13 000余张缺陷图像的卷烟包装外观缺陷识别基准数据集,并针对烟包缺陷识别、缺陷分类、目标检测、模型迁移4项任务开展实验。结果 结果表明,数据集能够满足高准确率深度学习模型的训练需求;通过模型迁移,能够利用该数据集大幅提高不同牌号卷烟的缺陷检测效果;DenseNet模型在烟包缺陷识别与缺陷分类任务上表现较好,准确率分别达到93.70%和95.43%,YOLOv5模型在缺陷目标检测任务上mAP@0.5值达到了96.61%。结论 该数据集能够作为烟包缺陷检测领域的基准数据集,研究成果将进一步支撑卷烟包装领域的数据应用与数字化转型。     

Nodule detection in digital chest radiography: effect of nodule location

Most detection studies in chest radiography treat the entire chest image as a single background or divided into the two regions parenchyma and mediastinum. However, the different parts of the lung show great variations in attenuation and structure, leading to different amounts of quantum noise and scattered radiation as well as different complexity. Detailed data on the difference in detectability in the different regions are of importance. The purpose of this study was to quantify the difference in detectability between different regions of a chest image. The chest X ray was divided into six different regions, where each region was considered to be uniform in terms of detectability. Thirty clinical chest images were collected and divided into the different regions. Simulated designer nodules with a full-width-at-fifth-maximum of 10 mm but with varying contrast were added to the images. An equal number of images lacking pathology were included and a receiver operating characteristic (ROC) study was conducted with five observers. Results show that the image contrast needed to obtain a constant value of A(z) (area under an ROC curve) differs by more than a factor of four between different regions.     

基于X射线像增强器成像的相衬滤波模型

对铝、黄铜、不锈钢3种金属材料制作的光栅在不同射线剂量和厚度差异的条件下进行图像采集,结合系统级的联合调制评价模型研究不同材料对X射线的调制作用,结果表明:在相同的射线剂量下,无论哪种材料的光栅,厚度差异越大成像反差越大;在厚度差异相同的情况下,随着射线剂量变化,不同材料对射线的调制能力变化趋势相差甚远。对光学调制传递函数、射线剂量和光栅厚度差异值进行三维曲线拟合,建立物质对X射线的调制模型和采集模型。在像增强器成像模型的基础上提出X射线相衬增强滤波模型,通过对薄膜基准目标源进行图像采集,在噪声环境下对其进行积分滤波处理和相衬增强处理,结果压低了图像噪声的同时也强化了目标轮廓信息,并大幅地增强X射线的成像反差。