Lung cancer is the most threatening tumor disease to human health. Early detection is crucial to improve the survival rate and recovery rate of lung cancer patients. Existing methods use the two-dimensional multi-view framework to learn lung nodules features and simply integrate multi-view features to achieve the classification of benign and malignant lung nodules. However, these methods suffer from the problems of not capturing the spatial features effectively and ignoring the variability of multi-views. Therefore, this paper proposes a three-dimensional (3D) multi-view convolutional neural network (MVCNN) framework. To further solve the problem of different views in the multi-view model, a 3D multi-view squeeze-and-excitation convolution neural network (MVSECNN) model is constructed by introducing the squeeze-and-excitation (SE) module in the feature fusion stage. Finally, statistical methods are used to analyze model predictions and doctor annotations. In the independent test set, the classification accuracy and sensitivity of the model were 96.04% and 98.59% respectively, which were higher than other state-of-the-art methods. The consistency score between the predictions of the model and the pathological diagnosis results was 0.948, which is significantly higher than that between the doctor annotations and the pathological diagnosis results. The methods presented in this paper can effectively learn the spatial heterogeneity of lung nodules and solve the problem of multi-view differences. At the same time, the classification of benign and malignant lung nodules can be achieved, which is of great significance for assisting doctors in clinical diagnosis.
Medical image registration plays an important role in medical diagnosis and treatment planning. However, the current registration methods based on deep learning still face some challenges, such as insufficient ability to extract global information, large number of network model parameters, slow reasoning speed and so on. Therefore, this paper proposed a new model LCU-Net, which used parallel lightweight convolution to improve the ability of global information extraction. The problem of large number of network parameters and slow inference speed was solved by multi-scale fusion. The experimental results showed that the Dice coefficient of LCU-Net reached 0.823, the Hausdorff distance was 1.258, and the number of network parameters was reduced by about one quarter compared with that before multi-scale fusion. The proposed algorithm shows remarkable advantages in medical image registration tasks, and it not only surpasses the existing comparison algorithms in performance, but also has excellent generalization performance and wide application prospects.
Objective To develop an automatic diagnostic tool based on deep learning for lumbar spine stability and validate diagnostic accuracy. Methods Preoperative lumbar hyper-flexion and hyper-extension X-ray films were collected from 153 patients with lumbar disease. The following 5 key points were marked by 3 orthopedic surgeons: L4 posteroinferior, anterior inferior angles as well as L5 posterosuperior, anterior superior, and posterior inferior angles. The labeling results of each surgeon were preserved independently, and a total of three sets of labeling results were obtained. A total of 306 lumbar X-ray films were randomly divided into training (n=156), validation (n=50), and test (n=100) sets in a ratio of 3∶1∶2. A new neural network architecture, Swin-PGNet was proposed, which was trained using annotated radiograph images to automatically locate the lumbar vertebral key points and calculate L4, 5 intervertebral Cobb angle and L4 lumbar sliding distance through the predicted key points. The mean error and intra-class correlation coefficient (ICC) were used as an evaluation index, to compare the differences between surgeons’ annotations and Swin-PGNet on the three tasks (key point positioning, Cobb angle measurement, and lumbar sliding distance measurement). Meanwhile, the change of Cobb angle more than 11° was taken as the criterion of lumbar instability, and the lumbar sliding distance more than 3 mm was taken as the criterion of lumbar spondylolisthesis. The accuracy of surgeon annotation and Swin-PGNet in judging lumbar instability was compared. Results ① Key point: The mean error of key point location by Swin-PGNet was (1.407±0.939) mm, and by different surgeons was (3.034±2.612) mm. ② Cobb angle: The mean error of Swin-PGNet was (2.062±1.352)° and the mean error of surgeons was (3.580±2.338)°. There was no significant difference between Swin-PGNet and surgeons (P>0.05), but there was a significant difference between different surgeons (P<0.05). ③ Lumbar sliding distance: The mean error of Swin-PGNet was (1.656±0.878) mm and the mean error of surgeons was (1.884±1.612) mm. There was no significant difference between Swin-PGNet and surgeons and between different surgeons (P>0.05). The accuracy of lumbar instability diagnosed by surgeons and Swin-PGNet was 75.3% and 84.0%, respectively. The accuracy of lumbar spondylolisthesis diagnosed by surgeons and Swin-PGNet was 70.7% and 71.3%, respectively. There was no significant difference between Swin-PGNet and surgeons, as well as between different surgeons (P>0.05). ④ Consistency of lumbar stability diagnosis: The ICC of Cobb angle among different surgeons was 0.913 [95%CI (0.898, 0.934)] (P<0.05), and the ICC of lumbar sliding distance was 0.741 [95%CI (0.729, 0.796)] (P<0.05). The result showed that the annotating of the three surgeons were consistent. The ICC of Cobb angle between Swin-PGNet and surgeons was 0.922 [95%CI (0.891, 0.938)] (P<0.05), and the ICC of lumbar sliding distance was 0.748 [95%CI(0.726, 0.783)] (P<0.05). The result showed that the annotating of Swin-PGNet were consistent with those of surgeons. ConclusionThe automatic diagnostic tool for lumbar instability constructed based on deep learning can realize the automatic identification of lumbar instability and spondylolisthesis accurately and conveniently, which can effectively assist clinical diagnosis.
Skin aging is the most intuitive and obvious sign of the human aging processes. Qualitative and quantitative determination of skin aging is of particular importance for the evaluation of human aging and anti-aging treatment effects. To solve the problem of subjectivity of conventional skin aging grading methods, the self-organizing map (SOM) network was used to explore an automatic method for skin aging grading. First, the ventral forearm skin images were obtained by a portable digital microscope and two texture parameters, i.e., mean width of skin furrows and the number of intersections were extracted by image processing algorithm. Then, the values of texture parameters were taken as inputs of SOM network to train the network. The experimental results showed that the network achieved an overall accuracy of 80.8%, compared with the aging grading results by human graders. The designed method appeared to be rapid and objective, which can be used for quantitative analysis of skin images, and automatic assessment of skin aging grading.
To locate the nuclei in hematoxylin-eosin (HE) stained section images more simply, efficiently and accurately, a new method based on distance estimation is proposed in this paper, which shows a new mind on locating the nuclei from a clump image. Different from the mainstream methods, proposed method avoids the operations of searching the combined singles. It can directly locate the nuclei in a full image. Furthermore, when the distance estimation built on the matrix sequence of distance rough estimating (MSDRE) is combined with the fact that a center of a convex region must have the farthest distance to the boundary, it can fix the positions of nuclei quickly and precisely. In addition, a high accuracy and efficiency are achieved by this method in experiments, with the precision of 95.26% and efficiency of 1.54 second per thousand nuclei, which are better than the mainstream methods in recognizing nucleus clump samples. Proposed method increases the efficiency of nuclear location while maintaining the location's accuracy. This can be helpful for the automatic analysis system of HE images by improving the real-time performance and promoting the application of related researches.
As the most efficient perception system in nature, the perception mechanism of the insect (such as honeybee) antennae is the key to imitating the high-performance sensor technology. An automated experimental device suitable for collecting electrical signals (including antenna reaction time information) of antennae was developed, in response to the problems of the non-standardized experimental process, interference of manual operation, and low efficiency in the study of antenna perception mechanism. Firstly, aiming at the automatic identification and location of insect heads in experiments, the image templates of insect head contour features were established. Insect heads were template-matched based on the Hausdorff method. Then, for the angle deviation of the insect heads relative to the standard detection position, a method that calculates the angle of the insect head mid-axis based on the minimum external rectangle of the long axis was proposed. Eventually, the electrical signals generated by the antennae in contact with the reagents were collected by the electrical signal acquisition device. Honeybees were used as the research object in this study. The experimental results showed that the accuracy of template matching could reach 95.3% to locate the bee head quickly, and the deviation angle of the bee head was less than 1°. The distance between antennae and experimental reagents could meet the requirements of antennae perception experiments. The parameters, such as the contact reaction time of honeybee antennae to sucrose solution, were consistent with the results of the manual experiment. The system collects effectively antenna contact signals in an undisturbed state and realizes the standardization of experiments on antenna perception mechanisms, which provides an experimental method and device for studying and analyzing the reaction time of the antenna involved in biological antenna perception mechanisms.
Objective To achieve threedimensional (3D) contour image of boneand articular cartilage for fabricating custommade artificial semiknee joint as segment bone allograft.Methods The distal femora of human and pig were scanned with Picker 6000 spiral X-ray computed tomography with 1.0 mm thick slice. The data obtained were treated in Voxel Q image workstation for 3D reconstruction with volume rendering technique. After being downloaded to personal computer at 0.1 mm interval, the transaxial 2D image data were converted into 2D digitized contour data by using image processing software developed by the team. The 2D digitized data were inputted into image processing software of Surfacer 9.0 (Imageware Company, USA), then the 3D wire frame and solidimages of femoral condyle were reconstructed. Subsequently， based on the clinical experience and the requirement of the design of artificial knee joint, the 3Dcontour image of bone or articular cartilage was extracted from the surrounding.Results The 3D contour image of bone or articular cartilage presented was edited and processed easily for the computer aided design(CAD) of custom-madeartificial knee joint.Conclusion The 3D contour image of boneand articular cartilage can be obtained by spiral CT scanning, and the digitized data can beapplied directly to CAD of custom-made artificial joint and subsequently rapidprototyping fabricating. In addition, the reconstruction method is simple and can be applied widely to clinical implant fabricating practice of dentistry and orthopaedics.
Purpose To investigate the pattern of subretinal neovascular membrane(SRNVM)in central exudative chorioretinitis(CEC). Methods With the help of a PC microcomputer,we performed a quantitative measurement of SRNVM in 32 eyes of 32 patients with Rieger is CEC. Results SRNVM-optic disc area ratio were 0.1151plusmn;0.0842．The foveola was on the top of SRNVM in 7 cases．The other 25 of SRNVMs were scattered in macular area around foveola,and 2 of them were nasal to it．The distance between the edge of SRNVM and foveola was less than 175mu;m in 13 cases,175～300mu;m in 4 cases and more than 300mu;m in 15 cases． Conclusion To be compared with the previous data,the present results suggested that laser photocoagulation might be one of the most important therapies for SRNVM in Rieger is CEC． (Chin J Ocul Fundus Dis,1998,14:114-115)
PURPOSE:To evaluated the luminal characteristics of the elderly central retinal vessels in the anterior optic nerves. METHODS:Serial sections of 15 central retinal arteries(CRA)and 23 central retinal veins (CRA)of 18 eyes of the aged 60 to 82 years old without anatomic malformation were examined by image analysis to investigate their luminal dimensional differences at the sites of lamina cribrosa and just anterior and posterior to it. RESULTS:The average values of the mean area of the CRA in the prelaminar,laminar,retrolaminar portions were separately(12.70,17.40,18.00)times;10-3mm2 and the mean perimetric length 0.56,0.56,0.57mm.No significant difference was detected in these three sites.The average values of the mean area of the CRV were respectively(7.00,5.40,7.90)）times;10-3mm2 and the mean perimetric length 0.44,0.38,0.41mm.There were marked differences between the prelaminar value and the laminar one,and between the laminar value and retrolaminar one by comparison. CONCLUSION:The CRA has a uniform radius from prelaminar to retrolaminar positions,and tube radius of the CRV at the level of the lamina cribrosa is the least. （Chin J Ocul Fundus Dis,1997,13: 213-214 ）
In order to accurately localize the image coordinates and serial numbers of intraoperative subdural matrix electrodes, a matrix electrode localization algorithm for image processing is proposed in this paper. Firstly, by using point-by-point extended electrode location algorithm, the electrode is expanded point-by-point vertically and horizontally, and the initial coordinates and serial numbers of each electrode are determined. Secondly, the single electrode coordinate region extraction algorithm is used to determine the best coordinates of each electrode, so that the image coordinates and serial numbers of all electrodes are determined point-by-point. The results show that the positioning accuracy of electrode serial number is 100%, and the electrode coordinate positioning error is less than 2 mm. The algorithms in this paper can accurately localize the image coordinates and the serial numbers of a matrix electrode arranged in an arc, which could aid drawing of cortical function mapping, and achieve precise positioning of brain functional areas, so that it can be widely used in neuroscience research and clinical application based on electrocorticogram analysis.