Download ##BEST## Qi Rong Image
Large-scale datasets are driving the rapid developments of deep convolutional neural networks for visual sentiment analysis. However, the annotation of large-scale datasets is expensive and time consuming. Instead, it is easy to obtain weakly labeled web images from the Internet. However, noisy labels still lead to seriously degraded performance when we use images directly from the web for training networks. To address this drawback, we propose an end-to-end weakly supervised learning network, which is robust to mislabeled web images. Specifically, the proposed attention module automatically eliminates the distraction of those samples with incorrect labels by reducing their attention scores in the training process. On the other hand, the special-class activation map module is designed to stimulate the network by focusing on the significant regions from the samples with correct labels in a weakly supervised learning approach. Besides the process of feature learning, applying regularization to the classifier is considered to minimize the distance of those samples within the same class and maximize the distance between different class centroids. Quantitative and qualitative evaluations on well- and mislabeled web image datasets demonstrate that the proposed algorithm outperforms the related methods.
download Qi Rong image
Luo-yang XUE designed the research and drafted the manuscript. Qi-rong MAO and Xiao-hua HUANG helped organize the manuscript. Jie CHEN participated in the experiments. Luo-yang XUE and Qi-rong MAO revised and finalized the paper.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
Thirty volunteers with healthy knees were examined using kinematic MRI. The knees were imaged in the transverse plane with flexion and extension angles from 0 to 40 and 40 to 0, respectively. The tibial interior and exterior rotation angles were measured, and the meniscal movement range, height change, and side movements were detected.
According to the measurement standard described by Sanfridsson et al. [7], the knee flexion angle was defined as the angle between the femoral longitudinal axis and the tibial longitudinal axis. The tibial rotation axis was defined as the vertical line from the midpoint of the anterior and posterior diameters in the tibial eminence, parallel to the tibial posterior edge. The tibial rotation angle was defined as the rotation angle of the sagittal plane through the rotation axis. In this study, layered transverse section imaging parallel to the joint surface was performed from the tibial plateau joint cartilage to the tibial tubercle plane. According to geometric principles, the tibial rotation angle was defined as the rotation angle through the line from the tibial tubercle midpoint to the upper tibiofibular joint midpoint, an accurate and convenient measurement. Overlapping images were obtained to measure the angle of the tibial rotation (Tables 1, 2, 3 and Figure 1).
In this study, the height of the meniscus was also studied. The results showed that meniscus height gradually increased with knee flexion. The meniscus undergoes morphological changes to adapt to the smaller curvature radius of the posterior femoral condyle. The meniscal matrix also contains a large number of type I collagen and proteogly can, both of which have a strong absorption ability and the capability to enhance the resistance compression of tissues and the elasticity of the meniscus [23]. The hardness of the meniscus is half that of the cartilage under compression. Therefore, the meniscus has a strong ability to disperse stress and fully absorb the concussion [23],[24]. The morphological changes of the meniscus may be related to nutrient absorption. With knee joint movement, the meniscus can present forward, backward, and sideward movements with morphological changes to adapt to the load and exert important functions. Meniscus injury or resection has adverse effects on the knee joint. Injury or resection can significantly change the load-transferring mode of the knee joint and cause an overload on the joint surface and subsequent degeneration of the joint cartilage. Kim et al. [25] found that the severity of these changes is related to the resection amount of the meniscus tissue. When treating an injury, the peripheral portions of the meniscus should be retained as much as possible and should be combined with meniscus repair, allografting, and prosthetic replacement. This method can aid in full recovery of the meniscus function.
Cross-sectional imaging techniques as computed tomography enterography (CTE) and MRE may offer additional information on transmural inflammation, stricturing and fistulising complications [5]. CTE delineate the extent and severity of bowel wall inflammation, apart from detecting extra-luminal findings [6]. However, the increased spatial resolution of CT with new multidetectors carries along with the greater dose of ionizing radiation. In fact, effective dose of radiation for CTE is nearly five times higher than small bowel follow through [7]. MRE is preferred for the evaluation of the gastrointestinal tract because of the absence of ionizing radiation, along with the similar sensitivity and diagnostic accuracy of luminal and extraluminal lesions [8,9]. Most of the recent researches focused on the inflammatory activity of involved segments using gadolinium-enhanced T1-weighted sequence for evaluation of CD. MRE signs of inflammatory activity in the bowel segments showed a good correlation with the presence and severity of endoscopic lesions [10,11]. MRE can measure intestinal thickness, enhancement after contrast injection, edema and ulceration or even more sophisticated measure as diffusion weighted imaging (DWI). DWI derives its image contrast from differences in the motion of water molecules between tissues [12]. DWI has been reported for colonic lesions in active CD, which could avoid the use of gadolinium injection [13]. Thus, the absence of gadolinium-induced nephrogenic systemic fibrosis indicates that DWI might be the proper method for long time follow-up in CD.
The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) in all four different b value groups would be measured. Image noise was measured from a large area (approximately 260 cm2) outside the abdomen parenchyma and defined as the standard deviation of background signal intensity. The signal contrast and ADCs would be measured on both normal and inflammatory bowel wall when these images were properly magnified, and ROI areas would be determined as maximum. On DW images, ROIs were placed on the segments where the signal was the most significant.
Considering the optimal image quality of DWI, the b value 800 s/mm2 was adopted as scale 0. When b value 1500 s/mm2 was selected, 7 cases were evaluated as scale 1 and 29 cases as scale 0. When b value 2000 s/mm2 was selected, 11 cases were evaluated as scale 1 and the other 25 cases were evaluated as scale 2. When b value 2500 s/mm2 was selected, all cases were evaluated as scale 2 (Figure 2).
The combination of MRE and DWI could improve diagnostic sensitivity and specificity. A 34-year-old female patient suffered from active CD with multi-segmental lesions in ileum (A, B and E). On LAVA image (A) abnormal enhancement could be detected in pelvis (white arrow), which was confirmed by enteroscopy (E), While on DWI (B), no abnormal high signal intensity lesion was found in the same cross-sectional slice. A 42-year-old male patient suffered from active CD (C, D and F). On LAVA image, suspicious local bowel wall thickening (curved arrow-head) was detected in distal ileum (C), while on DWI (D), no high signal intensity was found in the same cross-sectional slice, which was confirmed by enteroscopy (F).
The combination of MRE and DWI could improve diagnostic sensitivity and specificity. A thirty-three male patient suffered from active CD. On LAVA image the jejunum (right upper abdomen) was not distended very well and it was difficult to decide whether it had inflammatory lesion because of high signal intensity both before (A) and after (B) contrast injection. While on DWI (C), abnormal high signal intensity was found in the same cross-sectional slice indicating the inflammatory lesion, which was confirmed by enteroscopy.
SBE and DBE can provide high-quality images of the small intestinal mucosa, which enables the gastroenterologists, receive integral impression and evaluation of involved small intestine immediately. Due to the lack of scoring system for enteroscopy, the severity and extent of lesions in small bowel were assessed by the CECDAI in the present study. CECDAI can diagnose CD, exclude normal mucosal breaks, identify damage induced by nonsteroidal anti-inflammatory drugs, measure disease activity and severity, assess therapeutic response for clinical trials, and determine medical management for the patient with small-bowel CD. However, CECDAI is not suitable for patients with suspicious stenosis or obstruction due to the contraindication of capsule endoscopy (CE).
In conclusion, when DWI was combined with MRE for CD activity evaluation, b value 800 s/mm2 was recommended as the optimal DWI scanning parameter because of its good image quality and high diagnostic sensitivity of lesion detection. When diagnostic accuracy of active lesion was concerned, it was concluded that the combination of DWI and MRE showed the highest sensitivity, specificity and diagnostic accuracy compared with MRE or DWI alone. At the same time, the MR score obtained from both DWI and MRE radiological characteristics had showed a significantly positive correlation with CECDAI. Among all the MR score signs, bowel wall enhancement was the optimal predictor of active inflammation.
In the article, the most interesting and simplest method discovered in our experiment for fabricating silicon nanocrystals is self-assembly growth by assistance of electron interaction, in which silicon nanocrystals rapidly grow with irradiation of electron beam on amorphous silicon film prepared by PLD and shape of silicon nanocrystals is almost sphere when crystal size is smaller with less irradiation time of electron beam. The method of electron affection could be used to replace the traditional annealing methods in preparing process of silicon nanocrystals. In the process, it was investigated that condensed structures of silicon nanocrystals are changed with different impurity atoms in silicon film, for examples oxygen or Er atoms make a stronger condensed affection than doing of nitrogen or Yb atoms in impurity, in which various localized states emission was measured. 041b061a72